or
k
*e>
Occurrence and Exposure
Assessment for the Final
Long Term 2 Enhanced
Surface Water Treatment Rule
-------
Office of Water (4606-M) EPA815-R-06-002 December 2005 www.epa.gov/safewater
-------
Contents
Appendices v
Exhibits vi
Acronyms ix
1. Introduction -1
1.1 Regulatory Background , -1
. 1 Total Trihalomethane Rule -3
.2 Surface Water Treatment Rule -3
.3 Total Coliform Rule -3
.4 Regulatory Negotiation Process 1-4
.5 Information Collection Rule 1-4
.6 Safe Drinking Water Act Reauthorization 1-5
.7 M-DBP Advisory Committee 1-5
.8 Stage 1 Disinfectants and Disinfection Byproducts Rule 1-6
.9 Interim Enhanced Surface Water Treatment Rule 1-8
. 10 Long Term 1 Enhanced Surface Water Treatment Rule / Filter Backwash
Recycling Rule 1-9
.1.11 Ground Water Rule 1-10
.1.12 Stage 2 M-DBP Advisory Committee 1-10
.1.13 Stage 2 Disinfectants and Disinfection Byproducts Rule 1-12
.1.14 Long Term 2 Enhanced Surface Water Treatment Rule 1-13
1.2 Purpose of This Document 1-14
1.3 Document Organization 1-14
2. Characteristics of Waterborne Pathogens of Concern 2-1
2.1 Cryptosporidium spp. and Cryptosporidium parvum 2-1
2.1.1 Description of the Species 2-2
2.1.2 Strains 2-3
2.1.3 Fate and Transport 2-4
2.1.3.1 Surface Water 2-4
2.1.3.2 Ground Water Under the Direct Influence of Surface Water 2-6
2.1.3.3 Transmission of Cryptosporidiosis 2-7
2.1.4 Health Effects 2-8
2.1.5 Persistence 2-9
2.1.6 Response To Disinfection 2-12
2.2 Giardia and Giardia lamblia 2-16
2.2.1 Description of the Species 2-16
2.2.2 Strains 2-17
2.2.3 Fate and Transport 2-17
2.2.4 Transmission of Giardiasis 2-17
2.2.5 Health Effects 2-18
2.2.6 Persistence 2-18
2.2.7 Response to Disinfection 2-20
2.3 Viruses 2-23
2.3.1 Health Effects 2-23
2.3.2 Viral Pathogens 2-24
2.3.2.1 Adenoviruses 2-24
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
2.3.2.2 Astroviruses 2-24
2.3.2.3 Caliciviruses 2-25
2.3.2.4 Hepatoviruses 2-25
2.3.2.5 Enteroviruses 2-25
2.3.2.6 Rotaviruses 2-26
2.3.3 Persistence 2-26
2.3.4 Response To Disinfection 2-27
2.4 Waterborne Disease Outbreaks (1997 - Present) 2-29
2.5 Indicators of Fecal Contamination 2-30
2.5.1 Total Coliforms 2-31
2.5.2 Fecal Coliforms 2-31
2.5.3 Eschehchia coli 2-31
2.5.4 Persistence 2-31
2.6 Summary 2-33
3. Methods for Characterizing the Occurrence of Pathogens 3-1
3.1 Data Sources 3-1
3.1.1 Pre-ICR Occurrence Data 3-1
3.1.2 ICR Monitoring Program 3-1
3.1.3 ICR Supplemental Surveys 3-2
3.1.4 Representativeness of ICR and ICRSS Plants 3-4
3.2 Analytical Methods 3-6
3.2.1 ICR Method 3-6
3.2.2 Method 1622 and Method 1623 3-7
3.2.3 ICR Method for Viruses 3-8
3.3 Data Analysis Procedures 3-8
3.3.1 Challenges in Analyzing Microbial Occurrence Data 3-8
3.3.1.1 Low Occurrence and Concentration of Microbes 3-9
3.3.1.2 Sample Variability in Laboratory Technique 3-9
3.3.2 ICR and ICRSS Recovery Studies 3-10
3.3.2.1 ICR Lab Spiking Program 3-10
3.3.2.2 ICR Performance Evaluation Study 3-10
3.3.2.3 ICRSS Matrix Spiking Program 3-10
3.3.3 Microbial Occurrence Data Analysis Techniques 3-11
3.3.3.1 Observed Data Analysis 3-12
3.3.3.2 Modeled Distributions 3-14
3.4 Conclusion 3-18
4. Occurrence of Pathogens in Source Water 4-1
4.1 Cryptosporidium 4-1
4.1.1 Observed Results 4-1
4.1.1.1 ICR Monitoring Program Results 4-2
4.1.1.2 ICRSS Results 4-5
4.1.2 Modeled Results 4-11
4.1.2.1 Modeled Results for ICR Data 4-11
4.1.2.2 Modeled Results for ICRSS Data 4-14
4.2 Giardia 4-17
4.2.1 Observed Results 4-17
4.2.1.1 ICR Monitoring Program Results 4-17
4.2.1.2 ICRSS Results 4-20
Occurrence and Exposure Assessment
for the LT2ESWTR ii December 2005
-------
4.2.2 Modeled Results for ICR Data 4-28
4.3 Viruses 4-30
4.4 Indicators 4-32
4.4.1 Observed Results 4-32
4.4.1.1 ICR Monitoring Program Results 4-32
4.4.1.2 ICRSS Results 4-34
4.5 Co-Occurrence Data Analyses 4-38
4.5.1 Turbidity Related to Occurrence of Cryptosporidium, Giardia, E. coli, and
Viruses 4-38
4.5.2 Indicators Related to Occurrence of Cryptosporidium, Giardia, and Viruses
4-42
4.5.3 Microbial Index 4-49
4.5.3.1 Expressing Plant Summary Data 4-49
4.5.3.2 Plant Source Water Designations 4-50
4.5.3.3 Microbial Index Design 4-50
4.5.3.4 Censored Coliform Data 4-51
4.5.3.5 Assessment of Microbial Index Performance 4-52
4.5.3.6 Flowing Stream Index Results 4-54
4.5.3.7 Reservoir/Lake Index Results 4-54
4.5.3.8 Conclusions of Microbial Index Analysis 4-55
4.6 Cryptosporidium and Giardia Seasonal Patterns 4-56
4,7 Source Water Occurrence—Summary 4-56
5. Treatment by Physical Removal 5-1
5.1 Removal of Cryptosporidium and Giardia 5-1
5.1.1 Conventional and Direct Filtration 5-2
5.1.2 Other Filtration Technologies 5-5
5. .2.1 Slow Sand 5-5
5. .2.2 Diatomaceous Earth (DE) 5-6
5. .2.3 Membranes 5-7
5. .2.4 Bag and Cartridge Filtration 5-7
5.1.3 Prefiltration Optimization and Filtration Characteristics 5-8
5. .3.1 Coagulation Effects 5-8
5. .3.2 Filter Breakthrough Effects 5-9
5.1.3.3 Sedimentation and Dissolved Air Flotation 5-10
5.1.3.4 Solids Contact Clarifiers 5-11
5.2 Removal of Viruses 5-12
5.3 Conclusion 5-13
6. Observed Finished Water Occurrence 6-1
6.1 Cryptosporidium 6-1
6.2 Giardia 6-3
6.3 Viruses 6-4
6.4 Indicators 6-5
6.4.1 Total Coliform 6-5
6.4.2 Fecal Coliform 6-6
6.4.3 E. coli 6-7
6.5 Finished Water Occurrence—Summary 6-8
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
7. Modeled Pre-LT2ESWTR Occurrence of Cryptosporidium in Finished Water 7-1
7.1 Source Water Occurrence 7-2
7.2 Pre-LT2ESWTR Removal of Cryptosporidium 7-9
7.2.1 Large and Medium System Triangular Distributions 7-14
7.2.1.1 Modes 7-14
7.2.1.2 Maximum 7-15
7.2.1.3 Minimum 7-15
7.2.1.4 High-End Distribution 7-15
7.2.2 Small System Triangular Distributions 7-16
7.3 Description of Monte Carlo Model Used to Predict Finished Water Occurrence .... 7-16
7.4 Estimates of Pre-LT2 Finished Water Occurrence 7-17
7.5 Comparison of EPA Estimates with Aboytes et al. (2000) 7-23
7.6 Summary 7-24
8. Population Profile for Exposure Assessment 8-1
8.1 Population Profile 8-1
8.2 Characterization of Population, Including Sensitive Subpopulations 8-2
8.3 Population Profile for Exposure Assessment—Summary 8-5
9. References 9-1
Occurrence and Exposure Assessment
for the LT2ESWTR
IV
December 2005
-------
Appendices
Appendix A Waterborne Outbreaks Caused by Microbial Agents in Public Water Systems 1991 -98 . A-l
Appendix B Modeling Microbial Source Water Occurrence B-l
Appendix C Box Plots of Observed ICR Data C-l
Appendix D Graphs of Observed ICR Supplemental Survey Data D-l
Appendix E Bayesian Analysis Cumulative Distributions E-l
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
Exhibits
Exhibit 1.1 Chronology of EPA's Drinking Water M-DBP Rulemaking Efforts 1-2
Exhibit 1.2 Stage 1 DBPR Standards 1-7
Exhibit 1.3 TOC Percent Removal Requirements for Systems Employing
Enhanced Coagulation 1-7
Exhibit 2.1 Cryptosporidium in Host Species 2-3
Exhibit 2.2 Ground Water/Surface Water Interaction 2-5
Exhibit 2.3 Physical Inactivation of Cryptosporidium Oocysts 2-11
Exhibit 2.4 Disinfectants Tested Against Cryptosporidium Oocysts 2-12
Exhibit 2.5 Effects of Environmental Conditions on the Viability of Giardia Cysts 2-19
Exhibit 2.6 Disinfectants Tested Against Giardia Cysts 2-21
Exhibit 3.1 ICR and ICRSS Comparison 3-4
Exhibit 3.2 Distributions of ICRSS Plants 3-5
Exhibit 3.3 Distributions of ICR Plants 3-5
Exhibit 3.4 Spiking Study Results 3-11
Exhibit 3.5 Bayesian Cumulative Distribution of Source Water Occurrence 3-16
Exhibit 4.1 Summary of ICR Cryptosporidium Data for Filtered Plants 4-2
Exhibit 4.2 Cumulative Distribution of Plant-Mean Cryptosporidium Concentrations for Filtered Plants
for All Source Water Types—ICR Observed Results 4-3
Exhibit 4.3 Cumulative Distribution of Plant-Mean Cryptosporidium Concentrations by Source Water
Type—ICR Observed Results 4-4
Exhibit 4.4 Summary of ICR Cryptosporidium Data for Unfiltered Plants 4-5
Exhibit 4.5 Summary of ICRSS Cryptosporidium Data 4-5
Exhibit 4.6 Cumulative Distributions of Plant-Mean Cryptosporidium
Concentrations for Total, Non-Empty, and Internal Oocysts for All Sources—ICRSS Observed
Data 4-8
Exhibit 4.7 Cumulative Distribution of Plant-Mean Cryptosporidium Concentrations for Total, Non-
Empty, and Internal Oocysts by Source Water Type—ICRSS Observed Data 4-9
Exhibit 4.8 Cumulative Distribution of Plant-Mean Cryptosporidium Concentrations for Total, Non-
Empty, and Internal Oocysts by Plant Size-ICRSS Observed Data 4-10
Exhibit 4.9 Cumulative Distributions of Total Cryptosporidium in All Source Water Types—ICR
Modeled Data 4-11
Exhibit 4.10 Summary of ICR Cryptosporidium Modeled Data for All Plants 4-13
Exhibit 4.11 Summary of ICR Cryptosporidium Modeled Data for
Filtered and Unfiltered Plants 4-13
Exhibit 4.12 Cumulative Distribution of Modeled Data—ICRSS Large Plants 4-14
Exhibit 4.13 Cumulative Distribution of Modeled Data—ICRSS Medium Plants 4-15
Exhibit 4.14 Summary of ICRSS Cryptosporidium Modeled Data 4-16
Exhibit 4.15 Summary of ICR Giardia Data for Filtered Plants 4-18
Exhibit 4.16 Summary of ICR Giardia Data for Unfiltered Plants 4-19
Exhibit 4.17 Cumulative Distribution of Plant-Mean Giardia Concentrations for All Source Water
Types—ICR Observed Results 4-19
Exhibit 4.18 Cumulative Distribution of Plant-Mean Giardia Concentrations by Source Water
Types—ICR Observed Results 4-20
Exhibit 4.19 Summary of ICRSS Giardia Data 4-21
Occurrence and Exposure Assessment
for the LT2ESWTR vi December 2005
-------
Exhibit 4.20 Cumulative Distribution of Plant-Mean Giardia Concentrations for Total, Non-Empty,
Internal, and Internal >1 Cysts—ICRSS Data 4-22
Exhibit 4.21 Cumulative Distribution of Plant-Mean Giardia Concentrations for Total and Non-Empty
Cysts by Source Water Type—ICRSS Observed Data 4-23
Exhibit 4.22 Cumulative Distribution of Plant-Mean Giardia Concentrations for Internal and >1 Internal
Structure Cysts by Source Water Type—ICRSS Observed Data 4-24
Exhibit 4.23 Cumulative Distribution of Plant-Mean Giardia Concentrations for Total and Non-Empty
Cysts by Plant Size—ICRSS Observed Data 4-26
Exhibit 4.24 Cumulative Distribution of Plant-Mean Giardia Concentrations for Internal and >1 Internal
Structure Cysts by Plant Size—ICRSS Observed Data 4-27
Exhibit 4.25 Cumulative Distribution of Total Giardia in All Source Water Types—1CR Modeled Data
4-28
Exhibit 4.26 Summary of ICR Modeled Giardia Data 4-29
Exhibit 4.27 Summary of ICR Giardia Modeled Data for Filtered and Unfiltered Plants 4-29
Exhibit 4.28 Summary of ICR Virus Results, Filtered Plants 4-30
Exhibit 4.29 Summary of ICR Virus Results, Unfiltered Plants 4-31
Exhibit 4.30 Cumulative Distribution of Plant-Mean Virus Concentrations for All Source Water
Types—ICR Observed Results 4-31
Exhibit 4.31 Summary of ICR Coliform Data for Filtered Plants 4-32
Exhibit 4.32 Summary of ICR Coliform Data for Unfiltered Plants 4-33
Exhibit 4.33 Summary of ICRSS Coliform Data 4-34
Exhibit 4.34 Cumulative Distribution of Plant-Mean Coliform Concentrations by Plant Size—ICRSS
Observed Data 4-36
Exhibit 4.35 Cumulative Distribution of Plant-Mean Coliform Concentrations by Source Water
Type—ICRSS Observed Data 4-37
Exhibit 4.36 ICR Total Cryptosporidium, Total Giardia, Viruses, and E. coli vs. Turbidity 4-39
Exhibit 4.37 ICRSS Total Cryptosporidium and Total Giardia vs. Turbidity 4-41
Exhibit 4.38 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. E. coli 4-43
Exhibit 4.39 ICRSS Total Cryptosporidium and Total Giardia vs. E. coli 4-44
Exhibit 4.40 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. Fecal Coliform 4-45
Exhibit 4.41 ICRSS Total Cryptosporidium and Total Giardia vs. Fecal Coliform 4-46
Exhibit 4.42 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. Total Coliform 4-47
Exhibit 4.43 Total Cryptosporidium and Total Giardia vs. Total Coliform 4-48
Exhibit 4.44 Summary of Microbial Index Approach 4-51
Exhibit 4.46 Effect of Different Trigger Levels on Microbial Index Sensitivity (Flowing Streams) .. 4-54
Exhibit 4.47 Effect of Different Trigger Levels on Microbial Index Sensitivity (Reservoir/Lake) ... 4-55
Exhibit 4.48 ICR Monthly Mean Cryptosporidium Concentrations—Modeled Data 4-56
Exhibit 5.1 Cryptosporidium and Giardia Removal Efficiencies 5-1
Exhibit 5.2 Data Comparing Sedimentation and Dissolved Air Flotation Removal of Cryptosporidium
5-11
Exhibit 6.1 Summary of Volumes Sampled and Analyzed 6-1
Exhibit 6.2 Summary of Sample Results 6-2
Exhibit 6.3 Summary of Cryptosporidium-Pos\t\\e Sample Results 6-2
Exhibit 6.4 Summary of G/ard/'a-Positive Sample Results 6-4
Exhibit 6.5 Summary of Virus Positive Results 6-5
Exhibit 6.6 Summary of Total Coliform Positive Results 6-6
Exhibit 6.7 Summary of Fecal Coliform Positive Results 6-7
Exhibit 6.8 Summary of E. coli Positive Results 6-8
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
Exhibit 7.1 Comparison of ICR, ICRSSL, and 1CRSSM Median Curves Across 1,000 Modeled Source
Water Occurrence Distributions 7-4
Exhibit 7.2 ICR Filtered Systems
Summary of 1,000 Modeled Source Water Distribution Curves 7-5
Exhibit 7.3 ICR Unfiltered Systems
Summary of 1,000 Modeled Source Water Distribution Curves 7-5
Exhibit 7.4 ICR Supplemental Survey—Large Systems
Summary of 1,000 Modeled Source Water Distribution Curves 7-6
Exhibit 7.5 ICR Supplemental Survey— Medium Systems
Summary of 1,000 Modeled Source Water Distribution Curves 7-6
Exhibit 7.6 Comparison of Percentiles and Bounds (ICR Filtered Systems) 7-8
Exhibit 7.7 Triangular Removal Distributions for Medium and Large Systems 7-10
Exhibit 7.8 Triangular Removal Distributions for Small Systems 7-11
Exhibit 7.9a Cumulative Probability Low-End Distributions for Large and Medium Systems 7-12
Exhibit 7.9b Cumulative Probability High-End Distributions for
Large and Medium Systems 7-12
Exhibit 7.10a Cumulative Probability Low-End Distributions for Small Systems 7-13
Exhibit 7.1 Ob Cumulative Probability High-End Distributions for Small Systems 7-13
Exhibit 7.11 All Data Sets, Small Systems
Comparison of Median Finished Water Distribution Curves 7-18
Exhibit 7.12 All Data Sets, Large Systems
Comparison of Median Finished Water Distribution Curves 7-18
Exhibit 7.13 ICR Filtered Data, Small Systems
Median Curve and 90-Percent Confidence Bounds 7-19
Exhibit 7.14 ICR Filtered Data, Large Systems
Median Curve and 90-Percent Confidence Bounds 7-19
Exhibit 7.15 ICRSSL Data, Small Systems
Median Curve and 90-Percent Confidence Bounds 7-20
Exhibit 7.16 ICRSSL Data, Large Systems
Median Curve and 90-Percent Confidence Bounds 7-20
Exhibit 7.17 ICRSSM Data, Small Systems
Median Curve and 90-Percent Confidence Bounds 7-21
Exhibit 7.18 ICRSSM Data, Large Systems
Median Curve and 90-Percent Confidence Bounds 7-21
Exhibit 7.19 ICR Unfiltered Data, Large Systems
Median Curve and 90-Percent Confidence Bounds 7-22
Exhibit 7.20. Summary of Finished Water Occurrence Distributions
by Data Source and System Size 7-23
Exhibit 8.1 Estimates of Population Served By System Size (2000) 8-2
Exhibit 8.2 Estimates of Immunocompromised Population
Served By Surface Water and GWUDI Systems By System Size 8-3
Exhibit 8.3 Estimates of U.S. Population 65 and Older
Served By Surface Water and GWUDI Systems by System Size 8-4
Exhibit 8.4 Estimates of U.S. Population under Age 5
Served by Surface Water and GWUDI Systems by System Size 8-4
Occurrence and Exposure Assessment
for the LT2ESWTR viii December 2005
-------
Acronyms
AIDS Acquired Immunodeficiency Syndrome
AUX 1 Auxiliary 1
AWWARF American Water Works Association Research Foundation
CC Cell Culture
COWP Cryptosporidium outer wall protein
CPE Comprehensive Performance Evaluation
CT Concentration x Time
CWS Community Water System
DAPI Diamidino-2-phenylindole
DBPR Disinfection Byproducts Rule
DBFs Disinfection Byproducts
DIG Differential Interference Contrast
EPA Environmental Protection Agency
ESWTR Enhanced Surface Water Treatment Rule
FACA Federal Advisory Committee Act
FeCl3 Ferric Chloride
FBRR Filter Backwash Recycle Rule
FS Flowing Stream
GAC Granular Activated Carbon
GWR Ground Water Rule
GWUD1 Ground Water Under the Direct Influence of Surface Water
HAAs Haloacetic Acids
HAAS The Sum of 5 Haloacetic Acids
ICR Information Collection Rule
ICRSS Information Collection Rule Supplemental Surveys
IDSO Infectious Dose Causing Disease in 50 Percent of the Population
IESWTR Interim Enhanced Surface Water Treatment
LSP Lab Spiking Program
LT2ESWTR Long Term 2 Enhanced Surface Water Treatment Rule
M/DBP Microbial/Disinfection Byproduct
MCL Maximum Contaminant Level
MRDLGs Maximum Residual Disinfectant Levels Goals
MRDLs Maximum Residual Disinfectant Levels
NODA Notice of Data Availability
NPDWR National Primary Drinking Water Regulation
NTNCWS Nontransient Noncommunity Water System
NTU Nephelometric Turbidity Units
OST Office of Science and Technology
PCR Polymerase Chain Reaction
PWS Public Water System
QA/QC Quality Assurance and Quality Control
RL Reservoir/Lake
SA-11 Simian Rotavirus
SDWA Safe Drinking Water Act
SDWIS Safe Drinking Water Information System
SWTR Surface Water Treatment Rule
TCR Total Coliform Rule
Occurrence and Exposure Assessment
for the LT2ESWTR
IX
December 2005
-------
TOC Total Organic Carbon
TRAP-CI Thromobospondin-related adhesive protein Cl
TTHM Total Trihalomethanes
TWO Technical Work Group
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
1. Introduction
The United States Environmental Protection Agency (EPA or Agency) Office of Ground Water
and Drinking Water (OGWDW) is developing interrelated drinking water regulations to control microbial
pathogens and disinfectants/disinfection byproducts in drinking water. These rules are collectively
known as the microbial/disinfection byproducts (M-DBP) rules.
The Safe Drinking Water Act (SDWA) Amendments of 1996 require EPA to develop rules to
balance the public health risks from pathogens and DBFs. The Stage 1 Disinfectants and Disinfection
Byproducts Rule (Stage 1 DBPR) and the Interim Enhanced Surface Water Treatment Rule (IESWTR),
the first set of M-DBP rules under the SDWA Amendments, were promulgated in December 1998. The
Stage 1 DBPR and the IESWTR were the culmination of a 6-year (1992-1998) rule development process
that included regulatory negotiations with representatives of the water industry, environmental and public
health groups, and local, State, and Federal government agencies. The Amendments also require EPA to
publish a Stage 2 DBPR.
To support rule development, EPA expanded its microbial and DBP research program and
entered into collaborative efforts with other agencies and the water industry to collect data. This data
collection effort included the Information Collection Rule (ICR) and the ICR Supplemental Surveys
(ICRSS). In addition, under a joint effort between EPA and the National Rural Water Association
(NRWA), NRWA state chapters conducted a survey of byproduct and treatment information at small
public water systems (PWSs).
The purpose of the ICR was to provide microbial and DBP occurrence and treatment information
from large PWSs to support M-DBP regulations. EPA has worked with stakeholders under the Federal
Advisory Committee Act (FACA) to develop the Stage 2 DBPR and Long Term 2 Enhanced Surface
Water Treatment Rule (LT2ESWTR). These rules are being developed concurrently, using occurrence
data from the ICR and other available sources, to ensure that microbial protection is maintained or
enhanced while exposure to DBFs is reduced.
This occurrence and exposure assessment provides background information for the LT2ESWTR.
This chapter presents an overview of the regulatory background and describes the purpose of this
document.
1.1 Regulatory Background
Exhibit 1.1 presents a brief chronology of EPA's rulemaking activities on microbiological
contaminants, disinfectants, and DBPs in drinking water, starting with the Total Trihalomethane Rule
promulgated in 1979 (USEPA 1979). Following Exhibit 1.1 is a brief description of rulemaking efforts.
Occurrence and Exposure Assessment
for the LT2ESWTR
l-l
December 2005
-------
Exhibit 1.1 Chronology of EPA's Drinking Water M-DBP Rulemaking Efforts
Year
1979
1989
1989
1992
1994
1996
1997
1998
1999
2000
2001
2002
2003
2005
Regulation
Total Trihalomethane Rule (TTHM)
Surface Water Treatment Rule (SWTR)
Total Coliform Rule (TCP)
Negotiated Rulemaking
Stage 1 Disinfectants and Disinfection Byproduct Rule (DBPR)
Interim Enhanced Surface Water Treatment Rule (IESWTR)
Information Collection Rule (ICR)
Information Collection Rule
Safe Drinking Water Act (SDWA)
Microbial and Disinfectants/Disinfection Byproduct Federal Advisory
Committee {M-DBP FACA)
Stage 1 DBPR and IESWTR Notice of Data Availability (NODA)
(November)
Stage 1 DBPR Notice of Data Availability (NODA) (March)
Stage 1 Disinfectants and Disinfection Byproduct Rule (DBPR)
Interim Enhanced Surface Water Treatment Rule (IESWTR)
Stage 2 M-DBP FACA
Long Term 1 Enhanced Surface Water Treatment Rule
(LT1ESWTR)
Filter Backwash Recycle Rule
Ground Water Rule
Filter Backwash Recycle Rule
Long Term 1 Enhanced Surface Water Treatment Rule
(LT1ESWTR)
Long Term 2 Enhanced Surface Water Treatment Rule
(LT2ESWTR)
Stage 2 DBPR
LT2ESWTR
Stage 2 DBPR
Action
Promulgated
Promulgated
Promulgated
Initiated
Proposed
Proposed
Proposed
Promulgated
Reauthorized
Established
Presented new data and
M-DBP FACA
recommendations
Presented new health
effects data
Promulgated
Promulgated
Established
Proposed
Proposed
Proposed
Promulgated
Promulgated
Planned Proposal
Planned Proposal
Planned Promulgation
Planned Promulgation
Occurrence and Exposure Assessment
for the LT2ESWTR
1-2
December 2005
-------
1.1.1 Total Trihalomethane Rule
In November 1979, EPA promulgated an interim maximum contaminant level (MCL) of
0.10 milligrams per liter (mg/L) for total trihalomethanes (TTHMs): the sum of chloroform,
bromodichloromethane, dibromochloromethane, and bromoform. Compliance is calculated as a running
annual average of TTHMs (USEPA 1979). The TTHM Rule was based on the need to reduce exposure to
DBFs while maintaining disinfection to address microbial risks (USEPA 1979). Certain DBPs have been
shown to cause cancer in laboratory animals. The interim TTHM standard applies only to community
water systems that use surface and/or ground water, serve at least 10,000 people, and add a disinfectant
during any part of the treatment process. At their discretion, States may extend coverage to smaller PWSs
(USEPA 1979).
1.1.2 Surface Water Treatment Rule
Between 1979 and 1989, no new regulations related to disinfection or disinfection byproducts
were promulgated, however, interim regulations promulgated in 1975 regulating coliform bacteria and
turbidity remained in effect. The turbidity limits from the 1975 interim rule still apply to a few water
systems that have not installed filtration. In 1989, in response to the requirements of the 1986 SDWA,
EPA promulgated the Surface Water Treatment Rule (SWTR) (USEPA 1989a), which established
maximum contaminant level goals (MCLGs) of zero for Giardia lamblia, viruses, and Legionella. The
rule applies to water systems that treat surface water or ground water under the direct influence of surface
water. The SWTR specifies treatment technique requirements for filtered and unfiltered water treatment
systems that are intended to protect against the adverse health effects of exposure to Giardia lamblia,
viruses, Legionella, and many other pathogenic organisms. Briefly, those requirements include the
following:
• Maintenance of a disinfectant residual in the distribution system
• Removal and/or inactivation requirements of 3 logs (99.9 percent) for Giardia and 4 logs
(99.99 percent) for viruses
Combined filter effluent performance standards of five nephelometric turbidity units (NTU)
as a maximum and 0.5 NTU in at least 95 percent of the measurements taken each month,
based on samples collected at a 4-hour monitoring interval for treatment plants using
conventional treatment or direct filtration (with separate standards for other filtration
technologies)
• Watershed protection and other requirements for unfiltered systems.
1.1.3 Total Coliform Rule
In 1989, EPA promulgated the Total Coliform Rule (TCR) (USEPA 1989b) to provide protection
from microbiological contamination in the distribution system. Prior to the TCR, the interim regulations
required compliance with a MCL based on coliform bacteria density. The TCR established a MCLG of
zero for total and fecal coliform bacteria, and a MCL based on the percentage of positive samples
collected during a monthly compliance period. Total coliforms are bacteria that are used as an indicator
of water treatment effectiveness and distribution system integrity. Fecal coliform bacteria are generally
considered indicators of possible fecal contamination. Under the TCR, no more than 5 percent of
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
distribution system samples collected in any month may contain coliform bacteria. The number of
samples to be collected in a month is based on the population served. The location and frequency of
sampling is based on a system-specific sampling plan that provides representative coverage throughout
the distribution system. Combined, the SWTR and the TCR are intended to address risks associated with
pathogens that might be found in source waters or associated with distribution systems.
1.1.4 Regulatory Negotiation Process
Prompted by an interest in balancing health risks from microbial pathogens and DBFs, in 1992
EPA initiated a negotiated rulemaking to address public health concerns associated with disinfectants,
DBPs, and microbial pathogens. The negotiators included representatives of State and local health and
regulatory agencies, public water systems, elected officials, consumer groups, and environmental
organizations. The main concern in developing the rules was to ensure that when utilities changed
existing treatment to comply with new requirements for DBPs, they would not compromise microbial
protection. Hence, the negotiators agreed that EPA should propose a microbial rule with the DBPR.
Early in the rulemaking process, the Negotiating Committee determined that sufficient plant-
specific information on how to optimize the use of disinfectants, while concurrently minimizing pathogen
and DBF exposure risk, was not available. Nevertheless, the Negotiating Committee recommended that
EPA propose a DBPR to extend coverage to all community water systems (CWSs) and nontransient
noncommunity water systems (NTNCWSs) that use disinfectants. CWSs are public systems that serve at
least 15 service connections used by year-round residents or that regularly serve at least 25 year-round
residents. NTNCWSs are systems that serve at least 25 of the same persons for more than 6 months of the
year, where those persons are not full-time residents (e.g., colleges, schools, office buildings).
As a result of the negotiations, the Committee recommended that EPA develop three sets of rules.
These rules include a two-staged disinfection byproducts rule, a companion microbial rule, and an
information collection rule to gather data on microbial and DBP occurrence.
1.1.5 Information Collection Rule
EPA promulgated the ICR, a monitoring and data-reporting rule, on May 14, 1996 (USEPA
1996a). The ICR data collection results provide EPA and stakeholders with additional information on the
national occurrence in drinking water of key influent water quality parameters, disinfectants and
disinfection byproducts, and disease-causing microorganisms, including Cryptosporidium, Giardia, and
viruses. The ICR database also provides engineering data describing how PWSs currently control for
such contaminants and information on treatment applications used to reduce DBPs and their precursors.
The ICR data collection focused on large PWSs, which serve populations of 100,000 persons or
more. About 300 PWSs operating 501 treatment plants participated in this extensive data collection.
Over an 18-month period, these PWSs monitored influent water quality parameters affecting DBP
formation and DBP levels in the treatment plant and the distribution system. PWSs also provided
operational data and descriptions of their treatment plant design. The surface water systems monitored
for bacteria, viruses, and protozoa. Surface and ground water systems conducted monitoring to determine
the applicability of treatment study requirements. In addition, ground water systems serving between
50,000 and 100,000 persons were required to perform applicability monitoring. A subset of PWSs
performed treatment studies, using either granular activated carbon (GAC) or membrane filtration
processes, to evaluate DBP precursor removal and control of DBPs. The systems that were required to
Occurrence and Exposure Assessment
for the LT2ESWTR 1-4 December 2005
-------
perform treatment studies were selected based on applicability monitoring. All ICR systems began
monitoring for treatment study applicability in September 1996. The remaining occurrence monitoring
began in July 1997 and was completed in December 1998.
The ICR dataset is the basis for many analyses (for example, Crypfosporidium occurrence in
source waters) described in this document that support the development of the LT2ESWTR and the
Stage 2 DBPR.
1.1.6 Safe Drinking Water Act Reauthorization
In 1996, Congress reauthorized the Safe Drinking Water Act. The 1996 SDWA Amendments
include provisions related to the SWTR and the Enhanced Surface Water Treatment Rule. Those
provisions established a deadline of November 1998 for the promulgation of both the Stage 1 DBPR and
the IESWTR. The Amendments also set deadlines of November 2000 for the LT1ESWTR and May 2002
for the final Stage 2 DBPR. No mandatory deadline was established for LT2ESWTR. However, to
ensure a proper balance between microbial and DBP risks, EPA believes it is important to finalize the
Stage 2 DBPR in conjunction with the LT2ESWTR.
1.1.7 M-DBP Advisory Committee
In May 1996, EPA initiated a series of public meetings to exchange information on issues related
to the development of the IESWTR and the Stage 1 DBPR. EPA established the M-DBP Advisory
Committee under the FACA on February 12, 1997, to collect, share, and analyze new information and
data, as well as to build consensus on the regulatory implications of this new information. The M-DBP
Advisory Committee comprised 20 members representing EPA, State and local public health and
regulatory agencies, local elected officials, drinking water suppliers, chemical and equipment
manufacturers, and public interest groups. The M-DBP Advisory Committee agreed that the Stage 1
DBPR and IESWTR should:
1) Include proposed MCLs for total trihalomethanes, haloacetic acids, and bromate
2) Require enhanced coagulation and enhanced softening (with an adjustment based on new
data)
3) Require microbial profiling to ensure that DBP control does not compromise microbial
protection
4) Continue to give credit for complying with disinfection requirements (see section 1.1.2)
5) Establish stricter turbidity limits
6) Establish a Cryptosporidium MCLG and requirements for removing Cryptosporidium
7) Use a multiple barrier approach
8) Strengthen existing sanitary survey requirements (USEPA 1997)
Occurrence and Exposure Assessment
for the LT2ESWTR 1-5 December 2005
-------
1.1.8 Stage 1 Disinfectants and Disinfection Byproducts Rule
The Stage 1 DBPR (USEPA 1998a) sets maximum residual disinfectant levels (MRDLs) and
MRDL goals (MRDLGs) for chlorine, chloramine, and chlorine dioxide, and MCLs for chlorite, bromate,
and two groups of DBFs: TTHMs and five haloacetic acids (HAAS). HAAS refers to the sum of mono-,
di-, and trichloroacetic acids, and mono- and dibromoacetic acids. MCLGs were set for chlorite, bromate,
di- and trichloroacetic acids, and each individual trihalomethane. These standards are listed in Exhibit
1.2.
In addition, systems that use surface water or ground water under the direct influence of surface
water (GWUDI) and employ conventional treatment or softening must remove a specified percentage of
organic materials, measured as total organic carbon (TOC), unless they meet high source water quality
standards. This is important because these organic materials, or precursors, react with disinfectants to
form DBFs. Precursors can be removed through treatment techniques (enhanced coagulation or enhanced
softening), which are described in the Stage 1 DBPR (USEPA 1998a). Exhibit 1.3 lists the percentage
reduction of TOC required for various influent TOC and alkalinity levels.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-6
December 2005
-------
Exhibit 1.2 Stage 1 DBPR Standards
Disinfectants
Chlorine
Chloramine
Chlorine Dioxide
DBFs
Bromate
Chlorite
TTHMs
Bromoform
Bromodichloromethane
Chloroform
Dibromochloromethane
HAAS
Monochloroacetic acid
Dichloroacetic acid
Trichloroacetic acid
Monobromoacetic acid
Dibromoacetic acid
MRDLG (mg/L)
4.0 (as CI2)
4.0 (as CI2)
0.8 (as CIO2)
MRDL (mg/L)
4.0 (as CI2)
4.0 (as CI2)
0.8 (as ClOj)
MCLG (mg/L)
0
0.8
-
0
0
N/A
0.06
-
-
0
0.3
-
-
MCL (mg/L)
0.010
1.0
0.080
--
-
-
-
0.060
-
-
--
--
-
Note: The Stage 1 DBPR included a MCLG of zero for chloroform. The MCLG was challenged and the U.S. Court of
Appeals for the District of Columbia Circuit issued an order vacating the zero MCLG (U.S. Court of Appeals, DC
Circuit 2000). On May 30, 2000, EPA removed the MCLG for chloroform from its National Primary Drinking Water
Regulations (NPDWRs) (USEPA 2000a). EPA is proposing a new MCLG for chloroform in the Stage 2 DBPR
(USEPA 2003).
Exhibit 1.3 TOC Percent Removal Requirements for Systems Employing
Enhanced Coagulation
Source Water TOC (mg/L)
>2.0-4.0
>4.0-8.0
>8.0
0-60
35%
45%
50%
Source Water Alkalinity
>60-120
25%
35%
40%
(mg/L as CaCO5)
>1201
1*? Q£
I w /O
25%
30%
1 Requirements apply only to systems using enhanced softening.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-7
December 2005
-------
1.1.9 Interim Enhanced Surface Water Treatment Rule
The IESWTR serves two purposes: to improve the control of pathogens, specifically
Cryptosporidium, in drinking water, and to address risk trade-offs with DBFs. Key provisions established
in the IESWTR include the following:
• A maximum contaminant level goal of zero for Cryptosporidium
• 2 log (99 percent) Cryptosporidium removal requirements for systems that use conventional
or direct filtration
• Strengthened performance standards for combined filter effluent turbidity and individual
filter turbidity
Disinfection benchmark provisions to ensure continued levels of protection against pathogens
while facilities take the necessary steps to comply with the new DBF standards (referred to as
risk trade-off)
• Inclusion of Cryptosporidium in the definition of GWUDI and in the watershed control
requirements for unfiltered PWSs
• Requirements for covers on new finished water reservoirs
• Sanitary surveys of all surface water systems, regardless of size
The IESWTR specifies requirements for turbidity levels in all surface water systems that use
conventional treatment or direct filtration, serve 10,000 or more persons, and are required to filter. The
turbidity levels in combined filtered water must be no greater than 0.3 NTU in at least 95 percent of
samples taken each month; turbidity must not exceed 1.0 NTU at any time. In addition, systems must
monitor individual filters and provide an exceptions report to the State monthly. Exceptions to be
reported include the following:
• Any individual filter with a turbidity level greater than 1.0 NTU in 2 consecutive
measurements taken 15 minutes apart.
• Any individual filter with a turbidity level greater than 0.5 NTU after 4 hours of filter
operation, based on 2 consecutive measurements taken 15 minutes apart. If no obvious
reason for abnormal filter performance can be identified, a filter profile must be produced
within 7 days of the exceedance.
• An assessment by the system of any individual filter that has turbidity levels greater than 1.0
NTU in 2 consecutive measurements taken 15 minutes apart in each of 3 consecutive months.
• A Comprehensive Performance Evaluation by the State, or by a third party approved by the
State, of any individual filter that has turbidity levels greater than 2.0 NTU in 2 consecutive
measurements taken 15 minutes apart in each of 2 consecutive months.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-8
December 2005
-------
1.1.10 Long Term 1 Enhanced Surface Water Treatment Rule / Filter Backwash Recycling Rule
In 2001, EPA promulgated the Filter Backwash Recycling Rule (FBRR) and in January 2002,
promulgated the LT1ESWTR to increase protection of finished drinking water supplies from
contamination by Cryptosporidium and other microbial pathogens (USEPA 200la; USEPA 2002). These
rules apply to PWSs that use surface water or ground water under the direct influence of surface water.
The LT1 ESWTR extends protection against Cryptosporidium and other disease-causing microbes to the
11,500 small surface water systems that serve fewer than 10,000 persons annually. The FBRR, which
applies to all systems that recycle filter backwash, thickener supernatant, and liquids from dewatering,
regardless of size, establishes filter backwash requirements for certain surface water systems. The filter
backwash requirements will reduce the potential risks associated with recycling contaminants removed
during filtration. In the 1996 SDWA Amendments, Congress required the Agency to promulgate both
rules. A brief description of these rules follows.
FBRR Provisions - These apply to all systems that recycle regardless of population served.
• Recycling systems are required to return certain recycle streams through the processes of a
system's existing conventional or direct filtration system or at an alternative location
approved by the State.
• AH recycling systems must notify the State that they practice recycling and must submit a
plant schematic and recycle flow information.
LT1 ESWTR Provisions - These apply to systems that are required to filter and that serve fewer than
10,000 persons:
1) Cryptosporidium Removal
All systems must achieve 2 log (99 percent) removal of Cryptosporidium by meeting
turbidity requirements.
2) Turbidity
Conventional and direct filtration systems must comply with specific combined filter effluent
turbidity requirements and individual filter turbidity requirements; the requirements are
identical to those in the 1ESWTR.
3) Disinfection Benchmarking
Public water systems are required to develop a disinfection profile to ensure that if changes
are made to the disinfection practices in order to comply with the Stage 1 DBPR, current
microbial inactivation treatment is maintained.
4) Other Requirements
Finished water reservoirs for which construction begins after the effective date of the rule
must be covered.
Unfiltered systems must comply with updated watershed control requirements that add
Cryptosporidium as a pathogen of concern.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-9
December 2005
-------
1.1.11 Ground Water Rule
In May 2000, EPA proposed a targeted risk-based regulatory strategy for all ground water
systems (USEPA 2000b). The requirements provide a meaningful opportunity to reduce the public health
risk associated with the consumption of water contaminated with pathogens from fecal contamination for
a substantial number of people served by ground water sources.
The strategy addresses risks through a multiple-barrier approach that relies on five major
components:
• Periodic sanitary surveys of ground water systems requiring the evaluation of eight elements
and the identification of significant deficiencies.
• Hydrogeologic assessments to identify wells sensitive to fecal contamination.
• Source water monitoring for systems drawing from sensitive wells without treatment or with
other indications of risk.
• Compliance monitoring to ensure disinfection treatment is reliably operated (where used).
• Correction of significant deficiencies and fecal contamination through one of the following
actions:
- Eliminate the source of contamination
- Correct the significant deficiency
- Provide an alternative source of water
- Provide a treatment which achieves at least 99.99 percent (4 log) inactivation or removal
of viruses
1.1.12 Stage 2 M-DBP Advisory Committee
In March 1999, EPA reconvened a M-DBP Advisory Committee to develop recommendations on
issues pertaining to the development of the Stage 2 DBPR and LT2ESWTR. The Committee consisted of
organizational members representing EPA, State and local public health and regulatory agencies, local
elected officials, Indian Tribes, drinking water suppliers, chemical and equipment manufacturers, and
public interest groups. The Committee evaluated recent health effects information and the potential
benefits of a Stage 2 DBPR and LT2ESWTR. The Committee considered new information from the ICR
and other data sources on the occurrence of DBPs and pathogens as well as the treatment performance and
costs of various technologies. Technical support for the Committee's discussions was provided by a
technical workgroup (TWO) established by the group.
Despite the evaluation of a large amount of data, the Committee recognized that substantial
uncertainty remains regarding the nature and magnitude of risk associated with DBPs and pathogens in
drinking water. In light of this uncertainty, the Committee recommended steps, based on the extensive
analysis discussed in this document and in EPA's economic analysis, to address the areas of greatest
concern without placing an undue burden on public water systems.
In September 2000, the Committee signed the Agreement in Principle— a full statement of the
consensus recommendations of the group. The agreement was published by EPA in a December 2000
Occurrence and Exposure Assessment
for the LT2ESWTR 1-10 December 2005
-------
Federal Register notice (USEPA 2000c). The Agreement is divided into Parts A and B, as summarized
below.
Part A
Stage 2 DBPR
• MCLs for TTHM and HAAS will remain at 0.080 and 0.060 mg/L, respectively.
• Compliance with MCLs for TTHM and HAAS will be based on the locational running annual
average (LRAA), in two phases of the rule.
• In Phase 1 of the rule, systems must comply with TTHM and HAAS MCLs of 0.080 and
0.060 mg/L as a running annual average (RAA) and 0.120 and 0.100 calculated as a LRAA at
sample location.
• In Phase 2, compliance with TTHM and HAAS MCLs of 0.080 and 0.060 mg/L is calculated
as a LRAA for each of the new monitoring locations identified in the Individual Distribution
System Evaluation (IDSE).
• Systems will carry out an IDSE to select compliance monitoring sites that best capture the
highest TTHM and HAAS levels. The studies will be based either on system-specific
monitoring or other system specific data that provides equivalent or better information on site
selection.
• MCL for bromate will remain at 0.010 mg/L.
Occurrence and Exposure Assessment
for the LT2ESWTR
l-ll
December 2005
-------
LT2ESWTR
Additional treatment requirements for Cryptosporidium will be based on the results of source
water monitoring.
Systems that are required to provide additional treatment choose technologies from a
'toolbox' of options.
The monitoring burden for small systems will be reduced through the use of indicators to
determine which systems must monitor for Cryptosporidium.
Systems will conduct future monitoring to determine if source water quality has changed
following completion of the initial monitoring.
Unfiltered systems will provide at least 2 logs of Cryptosporidium inactivation, and unfiltered
systems will meet overall inactivation requirements with a minimum of two disinfectants.
Systems will cover all uncovered finished water reservoirs unless the reservoir effluent is
treated to achieve 4 logs of virus inactivation or the State/Primacy Agency determines that
existing risk mitigation is adequate.
EPA will develop guidance and criteria to facilitate the use of UV light for compliance with
drinking water disinfection requirements.
PartB
Beginning in January 2001, as part of the 6-year review of the Total Coliform Rule, EPA will
initiate a stakeholder process to address distribution system requirements related to
significant health risks.
• The Committee recommends that EPA develop a national water quality criteria under the
Clean Water Act for microbial pathogens for stream segments designated by States/Tribes for
drinking water use.
These recommendations reflect the Committee's emphasis on targeted, risk-based rulemaking.
They incorporate substantial initial monitoring to identify systems with the highest potential risk.
Additional treatment steps are required only where systems exceed limits on locational average DBP
concentrations or source water Cryptosporidium occurrence levels (USEPA 2000c).
1.1.13 Stage 2 Disinfectants and Disinfection Byproducts Rule
EPA proposed the Stage 2 DBPR in Fall 2003. The rule includes MCLGs for chloroform,
monochloroacetic acid (MCAA) and trichloroacetic acid (TCAA). The rule also sets MCLs and
monitoring, reporting, and public notification requirements for TTHM—the sum of chloroform,
bromodichloromethane, dibromochloromethane, and bromoform—and HAAS—the sum of mono-, di-,
and trichloroacetic acids and mono- and dibromoacetic acids. The rule also revises the monitoring
requirements for chlorite and chlorine dioxide. The document includes the best available technologies
(BATs) upon which the MCLs are based.
Occurrence and Exposure Assessment
for the LT2ESWTR 1-12 December 2005
-------
• The DBF standards are based on a LRAA compliance calculation, as opposed to a RAA.
With the LRAA, each sample point must be in compliance with the standards, rather than all
sample points combined.
• Systems must perform an IDSE, which will refocus their sampling plans on points within the
distribution system that best represent the highest concentrations of TTHMs and HAAS.
EPA believes the implementation of the Stage 2 DBPR will reduce the levels of DBFs in drinking
water supplies and will reduce inequities in exposure across distribution systems, resulting in reduced risk
of reproductive and developmental health effects and cancer.
The Stage 2 DBPR applies to public water systems that are C WSs and NTNCWSs that add a
primary or residual disinfectant other than ultraviolet light or deliver water that has been treated with a
primary or residual disinfectant other than ultraviolet light. In addition, the revised monitoring
requirements for chlorite and chlorine dioxide apply to transient noncommunity water systems
(TNCWSs).
1.1.14 Long Term 2 Enhanced Surface Water Treatment Rule
The LT2ESWTR provisions follow the recommendations of the M-DBP FACA Advisory
Committee Agreement in Principle. Key provisions include the following:
• Source water monitoring for Cryptosporidium, with reduced monitoring requirements for
small systems.
Additional Cryptosporidium treatment for some filtered systems, based on source water
Cryptosporidium concentrations.
* At least 2 log inactivation of Cryptosporidium by all unfiltered systems, using at least 2
disinfectants.
« Disinfection profiling and benchmarking to assure continued levels of microbial protection
while PWSs take the necessary steps to comply with new disinfection byproduct standards.
• Covering, treating, or implementing a risk management plan for uncovered finished water
reservoirs.
• Criteria for a number of treatment and management options (i.e., the microbial toolbox) that
PWSs may implement to meet additional Cryptosporidium treatment requirements. The
LT2ESWTR will build upon the treatment technique requirements of the Interim Enhanced
Surface Water Treatment Rule and the Long Term 1 Enhanced Surface Water Treatment
Rule.
The LT2ESWTR will apply to all PWSs using surface water or GWUDI as a source.
EPA believes that implementation of the LT2ESWTR will significantly reduce levels of
Cryptosporidium in finished drinking water in systems with high Cryptosporidium levels in their source
waters. This will substantially lower incidence of endemic cryptosporidiosis associated with drinking
Occurrence and Exposure Assessment
for the LT2ESWTR
1-13
December 2005
-------
water. In addition, the treatment technique requirements of this document are expected to increase the
level of protection from exposure to other microbial pathogens (e.g., Giardid).
1.2 Purpose of This Document
The purpose of this document is to summarize available information on Cryptosporidium and
Giardia and use it with statistical models to estimate the occurrence of these pathogens in drinking water
systems. This document emphasizes Cryptosporidium and Giardia because they are important
waterborne pathogens that can survive for months in the environment, are resistant to chlorine
disinfection (Giardia is less resistant than Cryptosporidium), and—particularly for Cryptosporidium—a
small number of oocysts can cause infection.
Many studies support the fact that Cryptosporidium oocysts are ubiquitous in the water
environment and some types are infectious to humans and many animals. Rivers, lakes, streams, and
ground water are potential sources of drinking water contamination. This document characterizes the FCR
and ICR Supplemental Survey data on occurrence of Cryptosporidium oocysts and Giardia cysts in the
source waters of large systems and models those data to characterize occurrence in source waters
nationwide.
This document uses information from scientific articles related to: (1) the occurrence, health
effects, persistence, and transmission of—and the analytical methods for detecting—Cryptosporidium and
Giardia; (2) monitoring studies (from the ICR and ICRSS); and (3) statistical modeling. Its foundation is
the Cryptosporidium and Giardia Occurrence and Exposure Assessment for the Interim Enhanced Surface
Water Treatment Rule (EPA 1998c). In addition, EPA conducted literature searches to identify articles
and studies to provide supplemental information regarding occurrence and exposure of Cryptosporidium,
Giardia, and other waterborne pathogens.
1.3 Document Organization
This document is organized into eight chapters. A description of each chapter follows.
• Chapter 2—Waterborne Pathogens of Concern: This chapter summarizes the characteristics
of common waterborne pathogens, Cryptosporidium, Giardia, and viruses. Their response to
treatment and health effects of specific strains are discussed. Recent waterborne disease
outbreaks are described.
• Chapter 3—Characterization of Pathogen Occurrence: This chapter describes the data sources
and subsequent analysis methods for the source water microbial occurrence. The challenges
encountered in analyzing such occurrences are also discussed.
• Chapter 4—Source Water Occurrence Data: This chapter presents the data collected to
document the national occurrence of pathogens in source water. Analyses of co-occurrence
and interactions of source water constituents are also presented.
• Chapter 5—Treatment by Physical Removal: This chapter reviews treatment effectiveness of
existing and new filtration technologies.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-14
December 2005
-------
Chapter 6—-Observed Finished Water Occurrence: This chapter presents the data describing
pathogen occurrence in finished water.
Chapter 7—Pre-LT2ESWTR Occurrence Estimates: This chapter provides estimates of pre-
LT2ESWTR finished water occurrence of pathogens based on treatments applied.
Chapter 8—Population Profile: This chapter presents the estimated populations served by
public water systems by system size and source of supply. These populations are used to
estimate the potential exposure to pathogens in drinking water.
Occurrence and Exposure Assessment
for the LT2ESWTR
1-15
December 2005
-------
2. Characteristics of Waterborne Pathogens of Concern
This chapter summarizes the characteristics of the waterborne pathogens Cryptosporidium,
Giardia, and viruses. Sections 2.1, 2.2, and 2.3 describe the health effects, variety of strains, person-to-
person transmission, and responses to disinfection for these three pathogens. The pathogens discussed are
those for which EPA has designated treatment requirements (USEPA 1989a, 1998b) and that were studied
as part of the ICR. EPA also designated treatment requirements for Legionella, a bacterial pathogen, but
Legionella was not monitored under the ICR and is not included here. Sections 2.1 and 2.2, on
Cryptosporidium and Giardia, contain descriptions of individual species and their persistence under
various environmental conditions. Section 2.3 discusses viruses. Section 2.4 describes recent outbreaks
of waterborne disease and their subsequent health implications. Section 2.5 describes the use of coliforms
as indicators of contamination.
2.1 Cryptosporidium spp. and Cryptosporidium parvum
Cryptosporidium is a major concern as a waterborne pathogen because its resistant oocyst form is
relatively unaffected by commonly used disinfection methods. For example, 2.5 milligrams per liter
(mg/L) of chlorine inactivated Giardia after 5 minutes (Jarroll et al. 1981), while it took 80 parts per
million of chlorine 90 minutes to inactivate Cryptosporidium (Korich et al. 1990) (starting concentrations
of organisms differed for each study). Cryptosporidium can persist in the environment for extended
periods of time. Several biological and epidemiological factors contribute to the potential for waterborne
transmission of cryptosporidiosis, including the following (adapted from Casemore 1990):
• Completion of the life cycle in a single host species (i.e, the presence or absence of other
species is not necessary for reproduction)
• Excretion of oocysts in large numbers, enhanced by the fact that Cryptosporidium can
reproduce within the host (autoinfection)
• Immediate infect! vity of excreted oocysts (no external "ripening" required)
• Relative resistance of oocysts to most environmental extremes and to common disinfectants
• Low infective dose in humans (DuPont et al. 1995; Chappell et al. 1999; Okhuysen et al.
1999)
• Excretion of oocysts by asymptomatic (displaying no symptoms) carriers
• Long-term excretion of oocysts by animals and chronically ill patients (e.g.,
immunocompromised individuals)
• Ubiquitous geographic distribution
• Lack of specific therapeutic cure
Occurrence and Exposure Assessment
far the LT2ESWTR
2-1
December 2005
-------
2.1.1 Description of the Species
Members of the genus Cryptosporidium are taxonomically classified in the phylum Apicomplexa,
order Eucoccidiorida, suborder Eimeriorina, and family Cryptosporidiidae (Payer et al. 1997;
O'Donoghue 1995). All members of the Apicomplexa are parasitic, and some of them are extremely
important as disease agents (Levine 1985). Within the phylum Apicomplexa are several related genera
referred to collectively as coccidia. Most coccidia are small protozoa that complete their life cycles
intracellularly (within the digestive tract epithelium, liver, kidney, blood cells, or other tissues of the
host). Cryptosporidium species infect epithelial surfaces, particularly those of the intestines. These
species are found in a wide range of vertebrates, including humans (Payer et al. 1997).
There is uncertainty about the taxonomy (i.e., classification) of species within the genus
Cryptosporidium. Until 1980, classification was based on the assumption that a particular species
infected only one type of animal (i.e., each host species harbored a separate species of Cryptosporidium)
(Payer and Ungar 1986). The "single host/single species" assumption is no longer accepted (Tzipori
1985). For example, C. parvum has been found to infect both humans and cattle. Hence, other more
appropriate taxonomy schemes have been suggested. Molecular characterization techniques provide
considerable evidence of genetic variations between isolates of a single species of Cryptosporidium from
different species of hosts, and increasing evidence suggests that a series of host-adapted genotypes/strains
exist (Morgan and Thompson 1998). Khramtsov et al. (1997) discovered a double-stranded RNA in C.
parvum sporozoite cytoplasm that was not present in any other Cryptosporidium species examined. This
RNA could be used as an identifier for C. parvum. Payer et al. (1997, 2000) identified 10 species of
Cryptosporidium, listed in Exhibit 2.1 (isolated from five mammalian, two avian, four reptilian, and one
fish species), and refer to several unnamed species isolated from a variety of hosts. The C. andersoni
species infecting cattle (see Exhibit 2.1) is not the same as the C. parvum strain that infects cattle; it more
closely resembles C. muris.
Depending on the classification scheme in use, virtually all Cryptosporidium infections in
humans are caused by C. parvum. Other species such as C.felis, C. meleagridis, and other species,
however, may infrequently infect humans, especially immunocompromised people (those with weak
immune systems) (Pieniazek et al. 1999; Morgan et al. 2000a; Pedraza-Diaz et al. 2001).
Xiao (1999) presented another classification scheme based on the 18S rRNA gene. This scheme
separates C. parvum and related species into intestinal and gastric genotypes that also exhibit other
differences, such as infectiousness versus non-infectiousness to humans. Xiao indicated that
Cryptosporidium includes at least four distinct species: C. parvum, C. baileyi, C. serpentis, and C. muris;
he states that C. wrairi, C. felis, C. meleagridis, and C. saurophilum are actually strains of C. parvum.
Champliaud et al. (1998) noted that although C. meleagridis is morphologically different from C. parvum
and does not normally infect humans, the DNA sequences compared by polymerase chain reaction (PCR)
amplification were very similar for each species. Therefore, other sequences need to be found to help
distinguish between the two.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-2
December 2005
-------
Exhibit 2.1 Cryptosporidium in Host Species
Named species of Cryptosporidium proposed as valid
Cryptosporidium species Host species
C. andersoni
C. baileyi
C. tolls
C. meleagridis
C. muris
C. nasorum
C. parvum
C. saurophilum
C. serpentis
C. wrairi
Bos taurus (domestic cattle)
Gallus gallus (domestic chicken)
Fe//s catis (domestic cat)
Meleagris gallopavo (turkey)
Mus musculus (house mouse)
Naso literatus (fish)
Homo sapiens (humans) +• >100 other mammals
Eumeces schneideri (skink)
Elaphe guttata (cornsnake)
Elaphe subocularis (rat snake)
Sanzinia madagascarensus (Madagascar boa)
Caw'a porcellus (guinea pig)
Source: Adapted from Fayer et al. 1997; Payer et ai. 2000.
Tzipori and Griffiths (1998) reviewed the difficulties associated with dividing Cryptosporidium
into species. These include (1) the lack of clearly defined and fully characterized reference strains for
comparative studies to define distinguishing phenotypic (physically observable) and genotypic
(genetically determined) parameters; (2) the ability of Cryptosporidium to infect a variety of cells, tissues,
organs, and vertebrate species; (3) and the often conflicting results of cross-transmission experiments.
The authors suggested that Cryptosporidium should be considered as a genus that consists of a wide
spectrum of isolates whose differences—-in host origin, site of infection, and oocyst size—are not as
important as virulence and genetic attributes, which have not yet been fully characterized. Tzipori and
Griffiths suggest that Cryptosporidium species be defined not simply by physical or genetic
characteristics, but by linking physical traits (particularly virulence attributes) and genetic markers.
2.1.2 Strains
Considerable genotypic and phenotypic differences exist among and within C. parvum isolates
(Tzipori and Griffiths 1998). Variations in infectivity to other animals, pathogenicity, antigenicity,
protein banding, isoenzyme typing, and genes that code for RNA and certain proteins have been
recognized. Xiao et al. (1999) believe, based on such characteristics, that several species that infect
different animals are actually strains of C. parvum.
Okhuysen et al. (1999) evaluated the virulence of three strains of C. parvum in humans. A vast
difference in the virulence of the isolates was observed among subjects with presumed infection (diarrhea
and/or oocyst excretion): the infectious dose causing disease in 50 percent of the population (ID50) was
87 oocysts for the Iowa (calf) strain; 9 oocysts for the TAMU (horse) isolate; and 1,042 oocysts for the
Occurrence and Exposure Assessment
for the LT2ESWTR
2-3
December 2005
-------
UCP isolate (calf)- For subjects who only excreted oocysts, the IDJO was 74.5,125, and 2,788 oocysts for
the Iowa, TAMU, and UCP isolates, respectively. Eighty-six percent of the volunteers who received the
TAMU isolate developed diarrhea, compared to 52 and 59 percent of those receiving the Iowa and UCP
isolates, respectively.
Several studies support the existence of at least two genotypes within the species C. parvum:
genotype 1, infecting humans only and genotype 2, infecting humans and livestock (Sulaiman et al. 1998;
Spano et al. 1998; Pieniazek et al. 1999; Xiao et al. 1998). Widmer et al. (2000), however, did succeed in
infecting piglets with genotype 1 and found that virulence increased with each passage. The human
genotype was also detected in a dugong (Morgan et al. 2000b). McLauchlin et at. (1999) suggest that the
two genotypes of C. parvum represent two reproductively isolated populations that are, or behave as,
different species within the genus Cryptosporidium based on the different chromosome location of the
Cryptosporidium outer wall protein (COWP) and thrombospondin-related adhesive protein Cl (TRAP-
Cl). Molecular analysis using TRAP-C2 sequencing for C. parvum isolates also differentiated the
isolates into two genotypes representing different animal transmission cycles (Peng et al. 1997; Sulaiman
et al. 1998). In a review of Cryptosporidium taxonomy, Morgan et al. (1999) also suggested separating
the two genotypes based on a lack of similarity between the rDNA ITS gene. They also recommended
that the pig, dog, and marsupial Cryptosporidium genotypes be assigned species names, based on genetic
differences even larger than those between C. parvum and C. wrairi, which infects guinea pigs and has
been accepted as a separate species. The authors noted the need to determine how genetic variations in
Cryptosporidium arise, through studies of reproduction and population, in order to more accurately define
what constitutes a separate species.
Contradicting the two-genotype theory, a study by Di Giovanni et al. (1999) identified seven
distinct C. parvum heat shock protein (hsplQ) genotypes. Reproducible differences between the C.
parvum LA-1 (the laboratory control strain) hsplO sequence and the C. parvum KSU-1 hsplO reference
sequence suggested that hsplO sequences may be useful for differentiating strains of C. parvum. The
authors also found that the hsplQ sequences of raw water or backwash samples of C. parvum matched
those of LA-1 in some cases and KSU-1 in others. In other samples, the hsplQ sequences differed from
both the laboratory and reference strains.
2.1.3 Fate and Transport
The fate and transport of pathogens in the environment are major issues with respect to the
exposure of humans to waterborne pathogens. The ability of microorganisms to survive in the
environment permits their transport either by water, food, or personal contact with a human host (Hurst
1997). The human exposure pathway of concern to EPA is drinking water from surface water and from
ground water supplies that are under the influence of surface water. Surface water sources include lakes,
rivers, reservoirs, and cisterns. Ground water supplies are extracted for drinking water by vertical and
horizontal wells, infiltration galleries, and springs.
2.1.3.1
Surface Water
When more rain falls than can be absorbed immediately by the soil or soil cover, water will pond
on the surface. With increasing rainfall, the water will flow to a lower level on the surface, such as a
river, lake, or reservoir, as shown in Exhibit 2.2. As water travels, it may pick up contaminants on the
Occurrence and Exposure Assessment
for the LT2ESWTR 2-4 December 2005
-------
soil surface (e.g., Cryptosporidium oocysts from deposited fecal matter). These particles are then
suspended in the surface water and can be transported as runoff. The microorganisms (including parasitic
protozoa) associated with the soil can be transported as individual organisms, aggregates of organisms, or
within an aggregate of soil particles and organisms. Partly for this reason, incidences of waterborne
cryptosporidiosis tend to vary temporally, with a higher prevalence during the warmest, wettest months
(Current 1986). In most areas of North America, Cryptosporidium occurrence generally becomes a
concern in surface waters during the spring, when rains increase runoff and many newborn animals,
which are more susceptible to C. parvum, are present in the environment. However, cryptosporidiosis
cases in the United States are most prevalent in late summer (Wolfson et al. 1985). The seasonal pattern
of cryptosporidiosis cases is suggestive of recreational water transmission in late summer and fall.
The character (topography, plant cover) and uses (urban, farming) of a watershed also influence
the occurrence or concentration of Cryptosporidium in surface water (Hansen and Ongerth 1991). For
example, one survey found that a mountainous, forested watershed with little or no human activity had
the lowest surface water oocyst concentrations and oocyst production, while downstream sample sites
influenced by dairy farming and urban runoff had oocyst concentrations and production rates almost 10
times higher than the upstream sites (Hansen and Ongerth 1991). In contrast, modeling by Walker and
Stedinger (1999) suggests that dairy oocyst loads are minor compared with treated wastewater oocyst
loads from humans.
Exhibit 2.2 Ground Water/Surface Water Interaction
Cryptosporidium may also directly enter surface water via waterfowl. Canada geese fed large
doses of C. parvum have been shown to pass intact oocysts that then caused severe illness in newborn
mice (Graczyk et al. 1997), although the geese did not develop cryptosporidiosis. Oocysts were also
found in goose feces collected in the environment (Graczyk et al. 1998). Canada geese, some of which no
longer migrate, could cause contamination of surface water sources and finished water reservoirs.
Cryptosporidium also can be transported through soil and ground water (Mawdsley et al. 1996;
Hurst 1997). Movement of C. parvum through soil and ground water is affected by sedimentation and
Occurrence and Exposure Assessment
for the LT2ESWTR
2-5
December 2005
-------
filtration of the surrounding soil and aquifer matrix (Brush et al. 1999; Harter et al. 2000). Harter et al.
measured oocyst pore velocities of 16.8 meters per day (m/day) in a coarse sand laboratory column.
Adsorption of oocysts to soil and aquifer matrix particles affects filtration. Adsorption depends
on the magnitude and distribution of the electrical charge on the organism and the surrounding soil and
aquifer matrix. Walker et al. (1998) summarized research on the effect of oocyst charge on surface
transport of Cryptosporidium. They reported that, because oocysts are approximately the size of clay/silt
particles, the amount of kinetic energy needed to entrain and suspend oocysts in overland flow may be
quite small. Studies of oocyst sorption to simplified analogs of soil mineral fractions and hydrophobic
materials suggested that the effective diameter of the oocyst may be the same as the oocyst itself in the
absence of positively charged particles. However, Brush et al. (1998) reported that the charge distribution
can be altered by the purification method and that the hydrophobicity can change as oocysts age.
The buoyancy of oocysts also affects their fate and transport in environmental media. Oocysts
that are not bound to particles may have a tendency to float. Researchers have found that 65 to 84 percent
of oocysts floated up into the supernatant of a homogenized fecal matter/distilled water mixture after 18
and 23 hours (Swabby-Cahill et al. 1996). Also, after six 10-minute centrifugations at 1,500 rotations per
minute (rpm), 16 percent of the oocysts were detected in the top third of the supernatant, suggesting some
resistance to settling (Swabby-Cahill et al. 1996). Cryptosporidiwn oocysts have a very low density
(about 1.05 grams per cubic centimeter (g/cm3)) and a very low settling rate (2 millimeters (mm) per hour
or less), which suggests that sedimentation without coagulation may not be an effective means of oocyst
removal (Gregory 1994). Rose et al. (1997) and Sreter and Szell (1998) also noted the low sedimentation
rate for oocysts. Medema et al. (1998) found that oocysts attached to wastewater effluent particles settled
more quickly than those that were freely suspended and that sedimentation velocity increased with
particle size. In source waters, many oocysts are likely to be adsorbed to organic or other suspended
material and would probably settle more quickly than free-floating oocysts (Medema et al. 1998).
2.1.3.2 Ground Water Under the Direct Influence of Surface Water
Some ground water—extracted for drinking by wells, infiltration galleries, or springs—is
regulated as surface water. Infiltration galleries are collection devices characterized by buried perforated
pipe in which water collects and is directed towards pumps. They are often used with shallow ground
water sources. Ground water that is considered to be under the direct influence of surface water
(GWUDI) is usually immediately adjacent to surface water or to the discharge point of a spring. These
ground water supplies are considered especially vulnerable to contamination by parasitic protozoa.
GWUDI may be contaminated by direct infiltration of oocysts from the surface as a result of rain. More
commonly, however, ground water is contaminated by the action of pumping wells (see Exhibit 2.2).
Given sufficiently high pumping rates, wells can reverse the direction of ground water flow. In this case,
surface water is induced to flow from a river, lake, or reservoir into the adjacent ground water, where it
may be extracted by one or more pumping wells. If the surface water is contaminated with oocysts, the
adjacent ground water may also become contaminated. Because of the potential for contamination,
GWUDI is regulated as if it were surface water.
Surface water sediments and the aquifer matrix material are believed to play significant roles in
minimizing oocyst transport to water supply wells. Not enough is known, however, about the
hydrogeology of aquifer matrices and sediments to determine their significance in preventing
contamination. Also, as discussed in section 3.3.1, problems with detection methods complicate the
Occurrence and Exposure Assessment
for the LT2ESWTR
2-6
December 2005
-------
accuracy of data collection. Therefore, whether the oocysts are present and not recovered or simply not
present in hydrogeologic settings such as alluvial aquifers is not known. Furthermore, little information is
available to elucidate which hydrogeologic settings are sensitive to oocyst contamination, because ground
water flow and oocyst transport through fractures or dissolution conduits can effectively bypass the
protective action of most of the aquifer matrix. In one study examining riverbank filtration, oocysts were
recovered at a well 200 feet from the Ohio River (Arora et al. 2000).
Harvey et al. (1995) modeled the transport of protozoa in ground water systems, using free-living
flagellates (protozoa with flagella, tail-like organelles used for locomotion) (2 to 3 micrometers (urn) in
size) and microspheres (0.7 to 6.2 urn in size). The authors observed the movement of flagellates and
microspheres through a column of sediment containing layers of varying grain size. They noted that
physical straining was particularly important in porous media, such as coarse sands, with grain diameter
greater than 100 um. Adsorption effects appeared to be related to the size of the microspheres. The
largest microspheres (2.8 to 6.2 um) were not significantly transported, but 83 percent of the
concentration of 1.7-um-size microspheres added to the column were present in the effluent.
To examine the mechanisms by which oocysts may be transported through soil to ground water,
Mawdsley et al. (1996) studied transport of Cryptosporidium parvum oocysts through three soil types
using 35-centimeter-long cores of each soil type. C. parvum transport was greater in a silty loam and a
clay loam soil than in a loamy sand soil. Because these results contradict other evidence that suggests
clay soils exhibit greater adsorption and smaller micropores than sandy soils, factors other than adsorption
and micropore size appear to have influenced the oocyst movement. The authors cited their use of intact
soil cores to maintain the natural soil structure and macropores, and concluded that the rapid flow of
water through macropores, which are representative of natural field conditions, largely bypasses the
filtering and adsorptive effects of the soil and greatly increases the risk of this pathogen's transport to
ground water (Mawdsley et al. 1996).
Hancock et al. (1998a) investigated the correlation of Cryptosporidium and Giardia occurrence in
GWUDI. They found no correlation for either pathogen between the distance of the ground water source
from adjacent surface water and occurrence. However, Hancock et al. (1999) reported that microbiota in
eight major groups were indicative of Cryptosporidium and Giardia contamination of ground water.
Cryptosporidium were found in seven of 149 vertical wells and five of 14 horizontal wells.
2.1.3.3 Transmission of Cryptosporidiosis
Cryptosporidiosis is primarily transmitted by fecal-oral transmission through direct contact (e.g.,
hand-to-hand) with an infected person or through contact with a fecally contaminated item (e.g., door
knobs or sink faucets). It is also commonly spread by swimming or playing in contaminated water and
sometimes through contaminated food. Cryptosporidiosis resulting from contaminated drinking water is
relatively rare, but large outbreaks have occurred as a result of drinking water contamination. People
infected with Cryptosporidium in an outbreak associated with a particular source, such as a day care
center or a contaminated swimming pool, can infect others, such as family members, who were not in
contact with the original source. This is called secondary transmission.
Secondary spread of Cryptosporidium infection among individuals may occur in homes, daycare
centers, hospitals, and urban environments, anywhere people are highly concentrated (Baxby et al. 1983;
Brown et al. 1989; Ribeiro and Palmer 1986; Heijbel et al. 1987; Melo Cristino et al. 1988; Payer et al.
Occurrence and Exposure Assessment
for the LT2ESWTR 2-7 December 2005
-------
1990; Casemore et al. 1997). Juranek (1995) explains that children in diapers are at especially high risk
for direct transmission of cryptosporidiosis because of intimate play or careless diaper changing practices.
Infection rates greater than 60 percent in urban day care centers have been reported (Payer and Ungar
1986). Asymptomatic family members living with infected individuals are sometimes found to excrete
small numbers of oocysts. Outbreaks are often associated with confirmed secondary cases among family
members and other people who had recent contact with infected individuals. For example, in 1989, an
outbreak of cryptosporidiosis occurred at a daycare center in Atlanta. Forty-nine percent of the children
and 13 percent of the staff members who submitted stool samples were found to be infected with
Cryptosporidium (Tangermann et al. 1991). Most of the cases likely were transmitted through the fecal-
oral route.
Documented hospital cross-infection with C. parvum, such as from patient to staff, is further
evidence of human-to-human transmission (Tzipori et al. 1983; Crawford and Vermund 1988; Casemore
et al. 1997). Cryptosporidium has been found in sputum and in vomit (Tzipori et al. 1983). Nosocomial
cryptosporidiosis infections (those contracted in hospitals) have been reported in both hospital staff and
patients (Juranek 1995). Fecal-oral exposure during sexual contact has been implicated as a transmission
(exposure) route for direct-contact transmission in homosexual males with acquired immunodeficiency
syndrome (AIDS). Respiratory cryptosporidiosis has also been reported in AIDS patients, sometimes in
the absence of diarrhea (Mifsud et al. 1994; Clavel et al. 1996; Dupont et al. 1996). Symptoms may
include fever, pneumonia, bronchitis, sore throat, and difficulty breathing.
2.1.4 Health Effects
Members of the genus Cryptosporidium are parasites of the intestinal tracts of fishes, reptiles,
birds, and mammals. C. parvum is the species commonly associated with self-limiting infections in
healthy humans characterized by mild to severe diarrhea, dehydration, stomach cramps, and/or a slight
fever, all generally lasting less than 2 weeks. Infection in unhealthy humans (especially those who are
immunocompromised) typically is more severe and may persist for months and result in death. A
Cryptosporidium outbreak in Milwaukee caused at least 46 deaths of individuals with AIDS and 8 deaths
among people with underlying medical conditions (Hoxie 1997). A few other Cryptosporidium species
(e.g., C. felts) have been found to infect immunocompromised humans, (Piezanek et al. 1999).
Cryptosporidium infection may result from ingestion of oocysts in food or water contaminated by
feces, or by person-to-person contact. In the gastrointestinal tract of suitable hosts, four sporozoites
excyst from each oocyst and parasitize epithelial cells. Thick-walled oocysts that are shed from the host
in fecal material are resistant to environmental stressors and may persist until they enter a new host by the
oral route. Approximately 20 percent of the oocysts produced in the gut have been shown to fail to form
a thick oocyst wall and instead release "thin-walled oocysts." These thin-walled oocysts allow the
excystation (release) of sporozoites within the gut and result in accelerated infection of new cells without
infection from external sources (Kansas State University 1997). This process of reproduction within the
same host is called autoinfection. Infection may or may not result in clinical symptoms of
cryptosporidiosis, for which no therapeutic cure is available.
The first symptoms of cryptosporidiosis appear 2 to 10 days after a person becomes infected. The
symptoms may include profuse, nonbloody, watery diarrhea that generally resolves spontaneously within
2 weeks; however, variability in clinical symptoms exists. Diarrhea! symptoms generally are not
distinguishable from those caused by other common enteric pathogens. Other symptoms reported by
Occurrence and Exposure Assessment
for the LT2ESWTR 2-8 December 2005
-------
individuals afflicted with cryptosporidiosis include abdominal cramps, vomiting, lethargy, and general
malaise (USEPA 1998d).
Human volunteer studies in which subjects ingest known doses of C. parvum (also called
challenge studies) have been conducted to assess the infectivity and dose-response of C. parvum in
humans (DuPont et al. 1995). In this study, infectivity was defined as causing excretion of oocysts
(regardless of whether symptoms were present). Sixty-two percent (18 of 29) of healthy subjects who
ingested became infected. The oocyst dose ranged from 30 to 1 million. Eleven of the 18 infected
individuals had enteric symptoms; seven of these individuals had diarrhea and, by clinical definition,
cryptosporidiosis. Only 1 of 5 subjects receiving a dose of 30 oocysts became infected, while 14 of 16
who received doses of 300 or more oocysts did, and all of those receiving doses of 1,000 or more oocysts
did. The ID50 for the Iowa strain of C. parvum was calculated at 132 oocysts in humans, compared with
an ID50 of 60 oocysts in neonatal mice; however, the test strain of C. parvum in this case was adapted to a
mouse model before challenge studies began, which may account for the disparity in ID50 values. The
mean and median incubation periods for cryptosporidiosis in the study were 9.0 and 6.5 days,
respectively. Infected humans developed clinical enteric symptoms that were associated with excretion of
oocysts, although 1 of the 11 subjects who did not pass oocysts passed a single soft stool on day 10 and
exhibited enteric symptoms on days 23 through 31. Symptoms of clinical illness included abdominal
pains, cramps, and diarrhea (in six subjects); nausea (in six subjects); vomiting (in one subject); and
moderate dehydration (in one subject).
Follow-up studies indicate that the number of excreted oocysts and the pattern and duration of
oocyst shedding vary widely among immunocompetent individuals (Chappell et al. 1996). In the
volunteer challenge study, high variability in shedding patterns was observed. Oocysts were observed
intermittently in consecutive stool samples, implying that production of oocysts is not uniform and may
be influenced by unknown factors. Thus, when single stool samples are submitted for diagnostic analysis,
the test for a patient with cryptosporidiosis may be negative.
Chappell et al. (1999) conducted additional volunteer studies to determine whether
Cryptosporidium infectivity varied with prior exposure to Cryptosporidium, as determined by positive
tests for anti-C. parvum serum immunoglobulin G. Study participants ingested doses of 500 to 500,000
oocysts. Infection and diarrhea correlated with higher oocyst doses. The authors determined that the
Cryptosporidium IDSO for previously exposed individuals was 1,880 oocysts, 20 times higher than the IDSO
for non-exposed people. In addition, oocysts were detected in the feces of only 54 percent of test subjects
with symptoms. These results suggest that previous Cryptosporidium exposure provides some immunity
to further infection at low oocyst doses.
2.1.5 Persistence
Several factors influence oocyst survival. This section presents the findings from several studies
describing oocyst inactivation due to temperature and dessication.
The survival pattern of oocysts suggests that, once an initial contamination has occurred, water
can remain a source of viable oocysts for days (Heisz 1997; Lisle and Rose 1995). Lisle and Rose
reported a duration of 176 days to produce die-off rates of 96 percent in tap water and 94 percent in river
water. After 2 days, a realistic contact time in most water distribution systems, only 37 percent of the
oocysts were nonviable.
Occurrence and Exposure Assessment
for the LT2ESWTR 2-9 December 2005
-------
Olson et al. (1999) compared oocyst survival in different media at temperatures likely to occur in
the natural environment. They examined survival in -4°, 4°, and 25° C in distilled water, soil, autoclaved
(sterilized) soil, and feces. Unlike Giardia, which died off quickly at low temperatures, Cryptosporidium
oocyst survival was best at -4°C, with close to 50 percent of oocysts remaining viable for 12 weeks in all
media except feces. Survival was lowest at 25°C, but oocysts were still viable at six weeks in all media.
Survival rates were best in water and worst in feces. Viability was determined by dye exclusion tests.
Cordell and Addiss (1994) noted that oocyst survival decreases in extreme temperatures and arid
conditions. Laboratory studies show that Cryptosporidium oocysts stored in airtight containers can
remain viable for 8 to 9 months, and excystation seems to occur soon after exposure to air (Tzipori 1983).
Robertson et al. (1992) reported that air drying an oocyst suspension at room temperature for 4 hours
eliminated viability. Oocysts in fecal material are protected from desiccation, however, so their viability
in the environment is prolonged (Rose et al. 1997).
Inactivation of oocysts is generally determined by assessing the viability or infectivity of the
oocyst. Viability means the organism is alive, but with many methods, one cannot be absolutely sure this
is the case. Viability can be estimated (with varying results, as described below) by testing an organism's
capability to exhibit metabolic activity or respond to biochemical stimuli through in vitro excystation,
infection of cell lines, changes in parasite morphology observed by light microscope, and uptake or
exclusion of fluorescent dyes. Infectivity is the ability of an organism to complete its life cycle within a
host and is the only way to be sure an organism is alive.
The correlation between viability and infectivity is important for assessing potential exposure.
Campbell et al. (1992) and Neumann et al. (2000) demonstrated that the fluorescence intensity from
nucleic acid binding dyes correlates with viability and infectivity for untreated oocysts, although some
controversy surrounds the use of fluorescent dyes to assess viability (Robertson et al. 1998). The
controversy arises because some oocysts that are not shown by such assays to be viable may appear viable
after exposure to a trigger, such as acid. Oocysts that are shown to be viable through other methods, such
as excystation or in vitro cultured cell infectivity, may not necessarily be able to infect live animals
(Neumann et al. 2000). In addition, correlation may be affected by chemical treatment. Bukhari et al.
(2000) found slight inactivation of C. parvum after treatment with ozone, as determined by dye
permeability assays and excystation. But they found much higher inactivation when they assessed the
same ozonated C. parvtim via mouse infectivity.
Many studies describe inactivation in terms of "log inactivation" or "log reduction." Each unit of
log removal is a factor of 10. Therefore, "a disinfectant achieved 1 log inactivation" means 10 percent of
the original number of viable oocysts were present after treatment. In other words, 90 percent of the
oocysts have been inactivated. Two log inactivation means that the viable oocyst concentration is 1
percent of the concentration present before treatment, or that 99 percent of oocysts have been inactivated.
Determining the amount of inactivation is difficult due to the problems with assessing viability.
Exhibit 2.3 shows the results of several studies of physical Cryptosporidium inactivation by
dessication and exposure to extreme temperatures. The studies in Exhibit 2.3 are not strictly comparable,
because the researchers did not assess inactivation in the same manner—they began with different
concentrations of oocysts and may have processed the samples differently. The results are thus presented
as a brief summary of the range of conditions found to inactivate Cryptosporidium.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-10
December 2005
-------
Exhibit 2.3 Physical Inactivation of Cryptosporidium Oocysts
Agent
Conditions
Results
Test
Reference
Note: Ex = Excystation, 1 = Infectious, Nl = Noninfectious, DEP = Dielectrophoresis. In vivo testing
performed in mice.
Heat
Heat
Heat
Heat
Heat
Freezing
Freezing
Freezing
Drying
Drying
12TC, 10 minutes
50-55PC, 5 minutes
45°C, 20 minutes
60°C, 6 minutes
59.70C, 5 minutes
64.2°C, 5 minutes
67.5°C, 1 minute
72.4°C, 1 minute
71 .7°C, 5-1 5 seconds
-196°C, 10 minutes
-20°C. 3 days
-70°C. 1 hour
-20°C, 8 hours; 1 day
-15°C, 24 hours; 1 week
-10°C, 1 week
Liquid nitrogen
-22°C, &32 days
Air dried, 2 hours
Air dried, 4 hours
Air dried in feces, 1-4 days
Protein
changes
Nl
Nl
Nl
I
Nl
I
Nl
Nl
Nl
Nl
Nl
I;NI
l;Nl
I
100% reduced
98% reduced
97% reduced
1 00% reduced
Nl
DEP
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
In vivo
Ex/dyes
Ex/dyes
Ex/dyes
Ex/dyes
In vivo
Archer etal. 1993
Blewett 1989
Anderson 1985
Payer 1994
Harp etal. 1996
Sherwood et al. 1982
Payer and Nerad 1996
Robertson et al. 1992
Robertson et al. 1992
Anderson 1986
Source: Adapted from Payer et al. 1997.
Temperature is a key factor affecting survival (Rose 1997). Although most of the temperatures
shown in Exhibit 2.3 are above what Cryptosporidium would be exposed to in nature, they indicate the
extent to which oocysts can survive adverse conditions. For instance, Payer and Nerad (1996), testing
oocysts frozen at -10°, -15°, -20°, and -70°C for 1 to 168 hours, demonstrated that oocysts of C. parvum
in water can be both viable and infective after freezing, although survival rates decrease with decreasing
temperature. Blewett (1989) observed a 92-percent reduction in oocyst viability (assessed by excystation)
following exposure to a temperature of 55°C for 5 minutes. Harp et al. (1996) tested the effect of
pasteurization on infectivity of oocysts in water and milk and confirmed that exposure to temperatures of
71.7°C for 5 to 15 seconds is sufficient to destroy the infectivity of C. parvum oocysts in water and milk.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-11
December 2005
-------
2.1.6 Response To Disinfection
Most Cryptosporidium oocysts are removed through the filtration of surface water, along with
coagulation, flocculation, and sedimentation. For the remaining oocysts, disinfection becomes important
for inactivation. Disinfectants commonly used to treat drinking water, however, are not very effective for
inactivating oocysts. Exhibit 2.4 summarizes the results of several studies regarding the effectiveness of
disinfectants, including ultraviolet (UV) radiation, on Cryptosporidium oocyst inactivation. In general,
chlorine/hypochlorite required the largest concentrations and longest contact times. UV radiation is now
feasible as a disinfectant; smaller doses are needed for inactivation than initially thought. The studies in
Exhibit 2.4 may not be directly comparable, because they did not assess inactivation in the same way.
Also, each began with different concentrations of oocysts and may have processed their samples
differently. The results are provided to summarize the effectiveness of different levels of disinfectant
under various water conditions. More detail on selected studies follows Exhibit 2.4.
Exhibit 2.4 Disinfectants Tested Against Cryptosporidium Oocysts
Disinfectants
Conditions
Results
Test
Reference
Note: Ex = Excystation, 1 = Infectious, Nl = Noninfectious. NR=No marked reduction. In vivo testing
performed in mice.
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine followed
by chloramine
Hypochlorite
Hypochlorite
Hypochlorite
Hypochlorite
Chlorine dioxide
Chlorine dioxide
2.5 mg/L, 30 min.
5 mg/L, 30 min.
867 - 51 18 mg/L, 24 hours
16,000 mg/L, 12 hours
28,000 mg/L, 24 hours
80 ppm, 90 minutes
2.5 -7. 5 mg/L CI2, 45
minutes; 4:1 NH3 to CI2,
2-3 hours
2.8%, 30 minutes, 25°C
1%, 30 minutes, 22°C
1 %. 30 minutes, 37°C
3%, 18 hours
5.25%, 2 hours, 20°C
0.007 mg/L, 16 minutes
0.22 mg/L, 30 minutes
1.3 mg/L, 1 hour
NR
72.5% to 88.1%
reduction
90% reduction
1 00% reduction
90% reduction
7 - 75%
reduction
89% reduction
55% reduction
69% reduction
I
I
97% reduction
94.3% reduction
92.7% reduction
Ex
Ex
Ex
Ex
In vivo
Ex
Ex
In vivo
In vivo
Ex
In vivo
Ex
Quinn and Berts
1993
Ransome et al.
1993
Smith etal. 1990
Korichetal. 1990
Oppenheimer et al.
1997
Sundermann et al.
1987
Blewett1989
Campbell et al.
1992
Payer 1995
Peeterset al. 1989
Korich et al. 1 990
Occurrence and Exposure Assessment
for the LT2ESWTK
2-12
December 2005
-------
Exhibit 2.4 Disinfectants Tested Against Cryptosporidium Oocysts
Disinfectants
Chlorine dioxide
Chlorine dioxide
Chlorine dioxide
Chloramine
Monochloramine
Monochloramine
Ozone
Ozone
Ozone
Ozone
Ozone followed by
chloramine
Mixed-oxidant
solution
UV
UV
UV
Conditions
4.03 mg/L, 15 minutes
1.52mg/L, 10-60 min, pH 6
1.52 mg/L, 10-60 min, pH 8
1.4 mg/L, 120 min., 1°C
1.2 mg/L, 120 min., 22°C
4.7 mg/L, 30 min. , 1°C
4.5 mg/L, 30 min., 22"C
3%, 24 hours
BO ppm, 90 minutes
0.066 mg/L, 48 hours
3.76 mg/L, 24 hours
0.3 -2.3 mg/L, 5-15
minutes, 3-22°C
0.4-2.4 mg/L, 22°C, 3-30
minutes, pH 6-8
0.36 - 2.2 mg/L, up to 42
minutes, 20°C
0.3-0.4 mg/L, 2 minutes
0.8 - 4 mg/L O3 12-30
minutes;
0.5 - 2.5 mg/L chloramine,
30-120 minutes
5 mg/L, 4 or 24 hours
80 mJ/cm2
120 mJ/cm2
8748 mJ/cm2
41-246 mJ/cm2
19-159mJ/cm2
Results
96% reduction
50-80% reduced
80-95% reduced
68.4% reduction
99.2% reduction
74.8% reduction
98.4% reduction
I
90% reduction
76.8% reduction
80.5% reduction
21-99.998%
reduction
68.4-99.997%
reduction
2-100%
reduction
NR-90+%
49.9% to 96%
inactivation
99.9%
inactivation
90% reduced
99% reduced
100% reduced
Nl, 10.7-98.9%
reduced
Nl, 1,24.7-95.1%
reduced
Test
Ex
Ex, cell
culture
In vivo
In vivo
Ex
Ex
Ex
In vivo
In vivo
Ex
Dye, ex,
in vivo
In vivo
In vivo
Ex
Ex
Ex/dyes
Ex/dyes
Hn vivo
Reference
Ransome et al.
1993
LeChevallier et al.
1996
Lietal. 1998
Pav!asek1984
Korich et al. 1 990
Ransome et al.
1993
Finch etal. 1993a
Gyureketal. 1999
Renneckeretal.
1999
Bukhari et al. 2000
Oppenheimer et al.
1997
Venczeletal. 1997
Ransome et al.
1993
Campbell et al.
1995
Bukhari etal. 1999
Occurrence and Exposure Assessment
for the LT2ESWTR
2-13
December 2005
-------
Exhibit 2.4 Disinfectants Tested Against Cryptosporidium Oocysts
Disinfectants
UV
uv
Pulsed UV
UV
Pulsed light
Conditions
3-9 rnj/crn2, medium
pressure UV
1 1-20 mJ/cm2 medium
pressure
3-9 mJ/cm2, low pressure
16-33 mJ/cm2, low
pressure
3 mJ/cm2
0.25-9.5 mJ/cm2
0.5-6 mJ/cm2 medium
pressure
1 mJ/cm2
Results
99.96- >99.99%
inactivation
>99.998%
inactivation
99.9-99.97%
inactivation
99.995-
>99.999%
inactivation
99.8%
inactivation
0-99.95%
reduced
0-99.95%
reduced
1 00% reduced
Test
In vivo
In vivo
Cell
culture-
PCR
Same
In vivo
Reference
Clancy et al. 2000
Shin et al. 2000a
Mofidietal. 1999
Dunnetal. 1995
Source: Adapted from Payer et al. 1997
Finch et al. (1997) evaluated the effects of several disinfection methods on the inactivation of
Cryptosporidium, noting that chlorine and monochloramine alone at practical plant levels are not
effective. Current (1986) also concluded in his review of Cryptosporidium biology that chlorine and
sodium hypochlorite {chlorine bleach) in typically used concentrations are poor disinfectants for
Cryptosporidium, although full-strength bleach (5.25 percent sodium hypochlorite) destroyed oocyst
infectivity to mice after 10 minutes.
Korich et al. (1990) exposed Cryptosporidium parvum to ozone, chlorine dioxide, chlorine, and
monochloramine and tested viability using a comparison of excystation and mouse infectivity. Ozone and
chlorine dioxide inactivated oocysts more effectively than did chlorine and monochloramine. Ozone at 1
mg/L for 5 minutes produced greater than 90 percent inactivation. Chlorine dioxide (1.3 mg/L)
inactivated 90 percent after 1 hour, and chlorine (80 mg/L) and monochloramine (80 mg/L) required
90 minutes for 90-percent inactivation.
Oppenheimer et al. (1997) studied the relationship between Cryptosporidium oocyst inactivation
and CT, the product of disinfectant residual (C) and disinfectant contact time (T), for disinfectants applied
to a wide range of source waters. The disinfection practices investigated included ozonation, addition of
chlorine followed by chlorine, addition of chlorine followed by chloramines, and addition of ozone
followed by chloramines. The data suggest no appreciable biocidal effect, either for chlorine alone or for
chlorine followed by chlorine. No more than 0.5 log inactivation was achieved using chlorine followed
by chloramines, even at impracticably high CT levels. However, chloramination immediately following
ozonation provided some enhancement of inactivation.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-14
December 2005
-------
Recent research on chlorine dioxide inactivation of Cryptosporidium has demonstrated moderate
effectiveness at feasible concentrations (considering disinfection byproducts) and reasonable contact
times. LeChevallier et al, (1996) showed that chlorine dioxide was moderately successful at disinfecting
Cryptosporidium. They achieved up to about 1.3 log of disinfection at pH 6 when the treatment was
carried out at 20°C, and less than 1 log if disinfection occurred at pH 8 or at 10°C. Li et al. (1998)
investigated temperature effects on Cryptosporidium inactivation at pH 6 and found that log inactivation
increased three- to four-fold as temperature was raised from 1 to 22°C.
Of all the disinfectants used in water treatment plants, ozone is the most effective in inactivating
Cryptosporidium in terms of the short contact time and lower residual concentration needed to achieve a
significant level of inactivation. Disinfection effectiveness, however, decreases with increasing pH.
Finch et al. (1997) reported a 1.5 log inactivation at 22°C and pH 6 and 0.5 log at pH 8. Gyiirek et al.
(1999) developed models for predicting inactivation of oocysts up to 3 logs, concluding that ozone was an
effective disinfectant for Cryptosporidium, but that results would vary with raw water source. One
problem with ozone disinfection is that at low levels the observed inactivation varies depending on the
method used to determine viability. Dye assays showed little difference between viability of ozonated
oocysts and controls at low doses and contact times, while mice infectivity tests showed a significant
increase in inactivation (Bukhari et al. 2000).
Venczel et al. (1997) evaluated an electrolytically produced mixed-oxidant solution (containing
free chlorine, chlorine dioxide, ozone, hydrogen peroxide, and other short-lived oxidants) for inactivating
Cryptosporidium parvum oocysts. The disinfection efficacy of the mixed-oxidant solution was compared
with that of free chlorine. The mixed-oxidant solution was considerably more effective, with a 5 mg/L-
dose of mixed oxidants producing a greater than 3 log inactivation of C. parvum oocysts (Iowa strain) in
4 hours. The same dose of free chlorine produced no measurable inactivation in 4 or 24 hours.
UV radiation works by damaging the DNA, RNA, or both of an organism, preventing it from
reproducing. Four types of U V treatment have been developed to inactivate microorganisms. In the first,
low-pressure radiation, constant UV radiation of a particular wavelength (254 nanometers (nm)) is
applied to water passing through a tube. In advanced UV treatment, similar UV lamps are used, but the
process is combined with a filter that temporarily traps organisms, lengthening the time they are exposed
to UV radiation. Medium-pressure radiation uses lamps that emit radiation of wavelengths between 200
and 300 nm. In pulsed UV treatment, water flows through a chamber filled with lamps that provide
flashes of high-intensity radiation of multiple wavelengths, including visible wavelengths.
Turbidity, color, and dissolved salts all prevent UV energy from penetrating the water and
therefore affect the amount of UV radiation required to disinfect a volume of water. UV light is generally
not used for disinfecting turbid water because of interference; it is used to disinfect ground water,
however, which is not prone to turbid conditions. Unlike chlorine, UV light has no residual disinfecting
capability and cannot prevent recontamination (USEPA 1999b), but it causes minimal disinfection
byproduct (DBF) formation.
Several independent data sets now exist for inactivation studies with Cryptosporidium parvum
using either animal infectivity or cell culture infectivity endpoints (mouse—Clancy et al. 2000; Bukhari et
al. 1999; and cell culture — Shin et al. 1999; Shin et al. 2000a; Mofidi et al. 1999). All the data
conservatively suggest that a pulsed UV dose of 20 millijoules per square centimeter (mJ/cm2) will result
in at least a 3 log inactivation of Cryptosporidium parvum. In fact, in many studies, a dose of 10 mJ/cm2
was so effective the assay detection limit was reached.
Occurrence and Exposure Assessment
for the LT2ESWTR 2-15 December 2005
-------
Bukhari et al. (1999) reported that medium-pressure UV was effective at inactivating C. parvum
up to 3.9 logs at a dose of 19 mJ/cm2 in a 200-gallon per minute (gpm) pilot system, using a mouse
infectivity assay. This medium-pressure U V work demonstrated that a continuous wave UV source could
perform as well as a pulsing source.
UV irradiation from a monochromatic low-pressure UV source for inactivation of C. parvum was
subsequently investigated by Shin et al. (1999,2000b), who demonstrated that a dose as low as 4 mJ/cm2
resulted in a greater than 3 log inactivation. Thus, monochromatic UV appeared to be as efficient at
inactivation of C, parvum as medium-pressure and pulsed UV sources.
Collectively, these studies—using low-pressure (monochromatic), medium-pressure
(polychromatic) and pulsed (polychromatic) U V systems on the bench and pilot scale in a variety of water
matrices—clearly indicate that UV light is effective in inactivating C. parvum. Cryptosporidium is
perhaps the pathogen most studied for UV effectiveness, and studies continue.
2.2 Giardla and Giardia lamblia
Giardia is found worldwide, and giardiasis is one of the most prevalent intestinal diseases of
humans (Meyer 1990; Craun 1986). Giardia infections are also common among domestic animals, such
as cats, dogs, birds, horses, rabbits, sheep, cattle, and goats, as well as other mammals and birds (Meloni
et al. 1995). Meloni et al. documented evidence of zoonotic transmission (transmission of disease from
animals to humans) in their study of genetic variation of Giardia, based on the fact that the same
genotypes were detected in humans and animals.
Members of the genus Giardia are flagellated, single-celled, binucleate protozoa that exist as
parasites in the intestinal tract of virtually every class of vertebrates. These protozoa have a two-stage life
cycle, the trophozoite (vegetative form) and the cyst (dormant, resistant form). Trophozoites of Giardia
lamblia (synonyms: G. duodenalis and G. intestinalis) inhabit the upper small intestine of the vertebrate
host (Filice 1952). The trophozoites are 9 to 21 urn long and 5 to 15 um wide, and cysts are 10 to 15 urn
long and 7 to 10 um wide (Daly 1983; USEPA 1998e).
2.2.1 Description of the Species
Host specificity and morphological characteristics have been used to distinguish species of
Giardia (Meloni et al. 1995). Both approaches, however, have limitations—the same Giardia species
may be found in different host species, and Giardia that appear identical may be different genetically.
Additional research to provide a taxonomic interpretation of genetic variation is currently underway.
Thompson and Lymbery (1996) examined genetic variability in Giardia and Cryptosporidium. They
noted the lack of understanding of within-host interactions among genetically different parasites, both
within the same species of parasite and among different species of the same genus.
Sil et al. (1998) cloned and characterized the ribosomal RNA genes from an Indian isolate of G.
lamblia to develop a method to differentiate Giardia from other enteric pathogens. Of the gene regions
studied, all were found to be genus-specific and not strain-specific.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-16
December 2005
-------
2.2.2 Strains
Studies ofGiardia have identified genetic variations both among and within species. Hopkins et
al. (1997) compared small subunit ribosomal RNA sequences from 13 human and 9 dog isolates of
Giardia duodenalis, which revealed 4 distinct genetic groups. Groups 1 and 2 contained all of the human
isolates, whereas groups 3 and 4 consisted entirely ofGiardia samples recovered from dogs. A genetic
basis for the differences observed between the groups was supported by sequence analysis of nine in vitro
cultured isolates that were placed into the same genetic groups established by enzyme electrophoresis.
G. lamblia strains have been separated into two groups: group A, also called Polish, and group B,
also called Belgian (Faintlia et al. 1999). Ey et al. (1998) studied variant surface protein genes of genetic
groups I and IF within group A. They identified a third type of gene found exclusively in group II
isolates.
2.2.3 Fate and Transport
The factors that influence the transport ofGiardia in the environment are the same as those
affecting Cryptosporidium (see section 2.1.4): adsorption, filtration, and sedimentation. The other main
feature affecting transport ofGiardia, especially in soil and aquifer materials, is its size. Giardia cysts
are 10 to 15 um in length and 7 to 10 um in width, larger than the 4-to 6-um diameter Cryptosporidium
oocyst. The cyst's larger size potentially restricts movement through some soils and aquifer materials,
except in the presence of natural pathways such as macropores, fractures, and conduits. As with
Cryptosporidium, Giardia cysts in feces deposited on soil surfaces are readily transported by surface
runoff during rainfall into surface water and, in some hydrogeologic settings, to ground water.
2.2.4 Transmission of Giardiasis
The most common method ofGiardia transmission in the United States is through the fecal-oral
route, particularly in day care centers. Overturf (1994) reviewed studies of giardiasis in daycare centers
and reported that the occurrence ofGiardia infections among children in daycare ranges from 17 to 90
percent. A study in Wisconsin reported the rates of Giardia infection among children were 17 to 47
percent, with rates among tested staff and household contacts of 9 to 35 percent and 5 to 18 percent,
respectively (Overturf 1994). Because Cryptosporidium and Giardia are distributed worldwide, data
from daycare centers in Salamanca, Spain, are applicable to the question of human transmission.
Rodriguez-Hernandez et al. (1996) studied 170 children younger than 4-years-old who regularly attended
daycare centers. Giardia was the most frequently identified parasite (found in 25.3 percent, or
43 children); 10 percent of children (17) had Cryptosporidium parasites. Children infected at daycare
centers may transmit giardiasis to other family members.
Giardia can also be transmitted within hospitals or other institutions. Transmission has also been
documented through sexual contact (USEPA 1998e). Giardia can be transmitted through drinking water
or recreational water. Between 1965 and 1996, 108 outbreaks occurred in public water systems; 15
outbreaks occurred in association with recreational water use (USEPA 1998e).
Occurrence and Exposure Assessment
for the LT2ESWTR 2-17 December 2005
-------
2.2.5 Health Effects
Giardia cysts enter a host through ingestion. They excyst in the stomach, subdivide, and attach
to the intestinal wall. The incubation period can last from 2 to 12 days, after which Giardia can lead to
weight loss and dehydration and may cause any of the following symptoms: diarrhea, abdominal cramps,
headaches, nausea, vomiting, and low-grade fever. For the average healthy adult, the symptoms last
approximately 2 weeks and, if untreated, can become chronic or cause intermittent diarrhea. For
immunocompromised individuals, the disease can last for months.
In 1997,25,389 cases of giardiasis were reported by states to the Centers for Disease Control
(CDC) nationwide (Fumess et al. 2000). This number is thought to underestimate the actual number of
cases, since most cases of diarrhea are not reported. Furness et al. estimate that anywhere from 500,000
to 2.5 million giardiasis cases occur annually. However, it is unlikely that these cases are associated with
drinking water contamination. Although states do not report the source of infection to CDC, the authors
hypothesize that most cases are due to recreational water use, since most cases occur during the summer.
2.2.6 Persistence
Environmental conditions contributing to the persistence of Giardia cysts are similar to those
described for Cryptosporidium in section 2.1.5. Surface water sources are more likely to be contaminated
with Giardia than are ground water sources (Craun 1990). Marginally treated or untreated surface water
supplies pose a high risk of transmitting Giardia because cysts can survive for several months in cold
water, and relatively few Giardia cysts are required for an infective dose (Craun 1990). Exhibit 2.5
summarizes some representative examples of the effects of environmental conditions on the persistence
and viability of Giardia cysts. Some of the conclusions of these studies are described below.
The occurrence of Giardia cysts in water is well documented (Madore et al. 1987; Craun 1990;
Hancock et al. 1998b; LeChevallier and Norton 1995), as is persistence. Giardia cysts survive relatively
long periods in water, particularly at temperatures below 20°C; above 20°C, cyst inactivation is rather
rapid. Evidence suggests that Giardia cysts in fresh water survive best at 4 to 8°C (Wickramanayake
1985; Jakubowski 1990). Kayed and Rose (1987) reported survival of protozoan cysts in water in the
laboratory for more than 140 days. Jarroll et al. (1984) reported that cysts did not survive when Giardia
were exposed for 24 hours to artificial sea water at 4°C or to air-drying at 4°C or 21 °C. Johnson et al.
(1997) studied the survival of Giardia, Cryptosporidium, and other enteric pathogens in marine waters.
Using excystation as the viability assay, they demonstrated a 3 log reduction of cysts in marine waters
after 3 hours in direct sunlight; but cysts in the dark required 77 hours to show a 3 log reduction (Johnson
etal. 1997).
t
Occurrence and Exposure Assessment
for the LT2ESWTR 2-18 December 2005
-------
Exhibit 2.5 Effects of Environmental Conditions on the Viability of Giardia Cysts
Environmental
Conditions
Conditions
Effect
Viability
Assay
Reference
Note: Ex = Excy station, 1 = Infectious, N! = Noninfectious, DEP - Dielectrophoresis. In vivo testing
performed in mice.
Liquefied feces
Distilled water
Tap water
Distilled water
Environmental
waters (lake
and river)
Minneapolis tap
water
Artificial sea
water
Distilled water
Distilled water
Soil
Autoclaved Soil
Feces
Stored at 4°C
Stored at 37°C
Stored at 8°C
100°C
-13°C
G. muris cysts in
fecal pellets stored
in water at 5-7°C
Same stored in
water at 5-7°C
Same stored in tap
water at 20-28 °C
24 hr at 4°C
1-20+ days at
-6°C
5°C
20°C
37°C
-4°C
4°C
25°C
Same
Same
Same
Infective for 1 yr
Viable <;4 days
Viable 77 days
100% reduction
>99% reduction after 14
days
100% viability at 7 days;
17-1 00% at 28 days;
0% at 56 days
Viable 2-3 months
Loss of viability within
3 days; 0-1 7% at
7 days; 0% at 14 days
No cysts survived
99.2- >99.8% reduced
2-9% reduced
6- >99.8% reduced
92- >99.8% reduced
Viable <1 wk.
80% reduced after 1 1
wks. (Ex), noninfective
after 1 1 wks.
Viable 4 wks. (Ex),
noninfective after 2 wks.
Viable/infective 1-9 wks
Viable/infective 1-9 wks
Viable/infective 1-9 wks
In vivo,
counterimmuno-
electrophoresis of
feces for antigen
Ex, dye
Same
Same
Same
Dye, in vivo, and
cyst morphology
by microscopy
Same
Same
Ex
Ex
Dye, in vivo
Craft 1982
Bingham et al.
1979
deRegnier et
al.,1989
Jarroll et al.
1984
Wickramana-
yake et al.
1985
Olson et al.
1999
Occurrence and Exposure Assessment
for the LT2ESWTR
2-19
December 2005
t
-------
Environmental
Conditions
Marine waters
Air
Conditions
3 hr in sunlight
77 hr in the dark
Cysts exposed to
air drying at 4°C or
21°Cfor24hr
Effect
99.9% reduction
99.9% reduction
No cysts survived
Viability
Assay
Ex
Ex
Ex
Reference
Johnson et al.
1997
Jarroll et al.
1984
2.2.7 Response to Disinfection
Giardia cysts are not as resistant to disinfection as Cryptosporidium oocysts; thus, treatment
designed to inactivate oocysts will effectively inactivate Giardia cysts. Korich et al. (1990) reported that
C. parvum oocysts are 30 times more resistant to ozone and 14 times more resistant to chlorine dioxide
than Giardia cysts exposed to the same disinfectant under the same conditions. Data collected by Owens
et al. (1994) indicated that Cryptosporidium parvum was 10 times more resistant to ozone than G. marts
protozoa. Oppenheimer et al. (1997) noted that the temperature-adjusted simulated CT values for ozone
inactivation of oocysts were 5 to 20 times the CT values for Giardia inactivation listed in the Surface
Water Treatment Rule Guidance Manual.
Jarroll (1988) and Jakubowski (1990) reviewed Giardia cyst disinfection. Some of the studies
they reviewed and data from additional studies are summarized in Exhibit 2.6. These studies may not be
comparable because they did not all use the same methods for assessing inactivation. Each also began
with different concentrations of cysts and may have processed samples differently. The results are
presented to provide a brief summary of the disinfectant levels found to inactivate Giardia.
Many of these studies are further described in the text following Exhibit 2.6. Some of the studies
described in the text are not listed in Exhibit 2.6 because insufficient data on the experimental conditions
or results were available.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-20
December 2005
-------
Exhibit 2.6 Disinfectants Tested Against Giardia Cysts
Disinfectant
Conditions
Results
Viability
Assay
Reference
Note: Ex = Excy station, 1 = Infectious, Nl = Noninfectious, DEP = Dielectrophoresis. In vivo testing
performed in mice.
Chlorine
Chlorine
Chlorine
Ozone
Ozone
1 mg/L at 5°C for 10
minutes
1.5mg/Lat25°Cfor10
minutes
2 mg/L at 5°C for 60
minutes
2.5 mg/L at 15°C for 10
minutes
0.3 - 2.5 mg/L at 0.5°C
to 5.0'C at pH 6 to pH 8
0.3 mg/L, 6.5 hours, pH 9
3.6 mg/L, 3 hours, pH 9
16.3 mg/L, 1 hour, pH 9
0.15 mg ozone/L for 0.97
minute at 25°C, or 0.48
mg ozone/L for 0.95
minutes at 5°C
0.1 8 mg ozone/L for
1.3 minutes at 25°C, or
0.70 mg ozone/L for
2.5 minutes at 5°C
0.02 -1.3 mg/L, 0.25-5
minutes, 22°C
35% (at pH 6) to 56%
(at pH 8} cyst
(G. lamblia) survival
No cyst survival
No cyst survival
No cyst survival at
pH 6, but 1 .8%
survival at pH 7
Mean CT to produce
99.9% to 99.99%
inactivation: 185 to
280
To produce >99.99%
inactivation: 220 to
290
99% inactivation
99% inactivation
99% inactivation
99% inactivation of
G. lamblia cysts
99% inactivation of
G. muris cysts
1.4%-99.996%
inactivation of G.
muris cysts
Ex
In vivo
Ex
Ex
Ex
Dye, ex, in
vivo
Jarrolletal. 1981
Hibleretal. 1987
Rubin etal. 1989
Wickramanayake
etal. 1985
Labatiuk et al.
1991
Occurrence and Exposure Assessment
for the LT2ESWTR
2-21
December 2005
-------
Disinfectant
Ozone
UV
UV
UV
UV
Conditions
0.26-0.82 mg/L for 5
min., pH 6.85
1.1 0-2 .52 mg/L for 2-1 3
min., pH 6.7
63 mJ/cm2
Up to 85 mJ/cm2
Up to 95 mJ/cm2
10-40mJ/cm2
2 mJ/cm2
Results
99.92- >99.9992%
inactivation of G.
muris cysts
99.5-99.996%
inactivation of G.
lamblia
Less than 90%
reduction
94.99- 99.87%
reduction of G. muris
0-68.3% reduction of
G. muris
99-99.9% reduction
>99.99% reduction
Viability
Assay
In vivo
In vivo
Modified ex
In vivo
Ex
In vivo
In vivo
Reference
Finch etal. 1993b
Rice and Hoff
1981
Finch and
Belosevic 1 999
Campbell 2000
Shin et al. 2000a
Chemical disinfection with chlorine is dependent on temperature and pH. Hibler et al. (1987)
determined the chlorine CT required to inactivate Giardia cysts at 0.5°C to 5.0°C. Reaction rates
decrease with temperature, which contributes to higher CT values necessary to inactivate cysts at lower
temperatures. Loss of biocidal activity occurs at pH above 7.5. CT values were determined for final
chlorine concentrations of 0.3 to 2.5 mg/L. The data were inconsistent for concentrations above 2.5
mg/L. The authors did not recommend using concentrations above 2.5 mg/L to compensate for a shorter
contact time, because the cyst wall may be able to resist extreme adverse environmental conditions for a
short period of time.
Ozone proved to be much more efficient at inactivating Giardia cysts than chlorine, based on a
CT comparison with results from Jarroll et al. (1981) by Wickramanayake et al. (1985). Increasing
temperature from 5°C to 25°C decreased the CT requirements of ozone necessary to achieve a 99-percent
inactivation; this temperature effect was similar to findings for chlorine by Hibler et al. (1987). Labatiuk
et al. (1991) and Owens et al. (1994) reported that ozone effectively inactivates Giardia muris cysts. G.
muris and G. lamblia did not exhibit significantly different responses to ozone in a comparative study by
Finch etal.(1993b).
Najm et al. (1998) evaluated the effect of the turbidity level on ozone inactivation of Giardia
muris cysts in natural water samples. They observed that cyst inactivation was higher in a low-turbidity
sample (1.3 nephelometric turbidity units (NTU)) than in samples having turbidity levels of 11 NTU and
19NTU(Najmetal. 1998).
Limited studies have been completed on the effectiveness of UV light in inactivating Giardia
cysts. As with Cryptosporidium, initial studies of the UV inactivation of Giardia were done using
viability assays (excystation); these studies probably overestimated the dose required to inactivate
Occurrence and Exposure Assessment
for the LT2ESWTR
2-22
December 2005
-------
Giardia. Rice and Hoff (1981) reported UV doses of 63 mJ/cm2 were necessary for less than 1 log
inactivation, based on excystation viability assays. Although these doses are not extremely high, they
were considered impractical at the time because UV equipment available at the time was designed for
smaller doses. In addition, Giardia could be controlled effectively with moderate doses of chlorine. In
vivo and cell culture infectivity assays are being used in current studies to more accurately determine UV
doses necessary for inactivation (Linden 2000); these doses are several times smaller than those in earlier
studies. Using medium pressure UV doses of less than 25 mJ/cm2, Finch and Belosevic (1999) achieved
2 to 3 log inactivation of G. muris measured via mouse infectivity; inactivation did not increase with
increased dose, even at 75 mJ/cm2. This inactivation was similar to that achieved for Cryptosporidium.
Shin et al. (2000a) achieved much better results with G. lamblia and low pressure UV. They attained >4
logs inactivation of Giardia with a dose of only 2 mJ/cm2. Campbell (2000) reached an inactivation of 3
logs. He stated that this was a conservative estimate because shielding and variability in resistance could
decrease inactivation.
2.3 Viruses
Waterborne pathogenic enteric viruses are among the organisms that continue to be regulated
under the Long Term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR). Enteric viruses, which
are shed via the intestinal tract and may be ingested through drinking contaminated water, have been
shown to cause a variety of diseases, including poliomyelitis, heart disease, encephalitis, aseptic
meningitis, hepatitis, and gastroenteritis. Little information is available about viruses' association with
waterborne disease because viruses are not detected frequently in drinking water during waterborne
disease outbreaks (a list of outbreaks is shown in Appendix A). The viral pathogens of concern include
adenoviruses, astroviruses, caliciviruses, hepatoviruses, enteroviruses, and rotaviruses; each is discussed
in section 2.3.2 below. Because viruses are so diverse, each type is affected differently by disinfection
and effectiveness of disinfection will not be described in as much detail as in the previous sections.
EPA's Office of Science and Technology (OST) recently prepared final drafts of two drinking
water criteria documents for viruses: Drinking Water Criteria Document for Viruses: An Addendum
(USEPA 1999b) and Drinking Water Criteria Document for Enteroviruses and Hepatitis A: An
Addendum (USEPA 1999c). They update EPA's draft drinking water criteria document for enteric
viruses, which was compiled in 1985 (USEPA 1985). These two documents present available
information regarding virus properties, occurrence, health effects, detection, and water treatment
technologies.
2.3.1 Health Effects
Humans and a variety of animals throughout the world may be infected by adenoviruses.
Adenoviruses are pathogenic only within the species in which they originate (USEPA 1999b). Only a few
adenoviruses are enteric. Adenoviruses, astroviruses, and caliciviruses all can cause gastroenteritis.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-23
December 2005
-------
Enteroviruses are associated with a variety of diseases, including poliomyelitis, aseptic
meningitis, heart disease, respiratory disease, gastroenteritis, and diabetes mellitus (USEPA 1999c).
Enteroviruses also account for approximately 10 to 20 percent of encephalitis cases with proven viral
etiology. Infants and young children experience the highest rates of serious enteroviral disease, and males
are infected at a 50 percent higher rate than females (Modlin 1997; USEPA 1999c).
Hepatoviruses cause hepatitis; of which types A and E can be transmitted through drinking water.
Rotaviruses are the most significant cause of severe gastroenteritis in young children and infants.
Although these viruses cause gastroenteritis in young children, non-Group A rotaviruses also can cause
outbreaks in older children and adults. Rotaviruses consistently outrank other known etiologic agents of
severe diarrhea in scores of illnesses and hospitalizations reported annually (USEPA 1999b). Fecal-oral
transmission is the primary mode of transmission for rotaviruses.
2.3.2 Viral Pathogens
2.3.2.1 Adenoviruses
Some enteric adenoviruses are fastidious, or difficult to grow in cell cultures; they are thus more
difficult to detect, and their role as an agent of gastroenteritis may be underestimated. Adenoviruses most
commonly cause respiratory illness through routes other than drinking water. These viruses can also
cause conjunctivitis, cystitis, and rashes. However, Ad40 and Ad41, the two serotypes of "enteric"
adenoviruses, occur in large numbers in stool samples of individuals suffering from gastroenteritis
(USEPA I999b, CDC 200 la). Common symptoms of gastroenteritis include watery diarrhea, vomiting,
headaches, fever, and abdominal cramps (CDC 1998b). Enteric adenoviruses are thought to be a
significant cause of childhood diarrhea (USEPA 1999b). Currently, however, no direct evidence
associating "enteric" adenoviruses with transmission of disease via drinking water exists, because no
studies of adenoviruses in drinking water have been conducted (USEPA 1999b).
2.3,2.2
Astroviruses
Seven serotypes of human astrovirus have been identified. Extensive seroepidemiological studies
in the United Kingdom between 1975 and 1987 revealed that serotype 1 astrovirus accounted for 65 to 72
percent of the cases of astrovirus-induced gastroenteritis, while each of serotypes 2 through 5 accounted
for 6 to 8 percent of the cases, respectively (Kurtz and Lee 1987; USEPA 1999b). This single-stranded
RNA virus can cause gastroenteritis, predominantly in children under the age of 5. Transmission occurs
mainly through fecal-oral contact. Water has been suggested as another possible route of transmission,
but there has been no strong evidence to support this assertion (AWWARF 1997).
Occurrence and Exposure Assessment
for the LT2ESWTR
2-24
December 2005
-------
2.3.2.3 Caliciviruses
Norwalk viruses are classified in the family Caliciviridae and the genus Calicivirus (1CTV 1995).
The Norwalk virus is the prototype strain of a group of fastidious 26- to 35-nm non-enveloped, single-
stranded RNA viruses associated with outbreaks of gastroenteritis. The Norwalk-like viruses were named
after the location of the 1968 Norwalk, Ohio, outbreak from which they were first isolated (Kapikian et
al. 1996; USEPA 1999b).
Since the original report of Norwalk virus as the cause of an acute gastrointestinal illness, several
morphologically similar agents have been detected and shown to be associated with gastroenteritis. These
viruses, named after the locations where they were first found, include the Hawaii, Snow Mountain, and
Taunton agents, and have been designated as small round structured viruses (SRSV).
Human caliciviruses may fall into a range of antigenic types, which have been associated with
outbreaks in all age groups in North America, Australia, Asia, Africa, and Europe (Matson et al. 1989).
The more prevalent antigenic types appear to be those that infect primarily infants and young children,
such as HCV (Sapporo), although strains that produce symptomatic infections in adults have been
identified (Matson et al. 1989; USEPA 1999b). Studies on the environmental occurrence and
susceptibility to treatment of human caliciviruses have been hindered by the inability to grow them in cell
culture.
2.3.2.4 Hepatoviruses
Hepatitis A virus causes liver disease (CDC 2000b). Symptoms include jaundice, fatigue,
abdominal pain, loss of appetite, intermittent nausea, and diarrhea. These symptoms can last up to 6
months, but regularly last less than 2 months. The virus is transmitted orally, through food or water
outbreaks, and through feces. Those at risk are international travelers, persons living in regions of
endemic hepatitis A, such as American Indian reservations or Alaska Native villages, and people in
frequent contact with an infected person, either sexually or through another form of contact. During
outbreaks, the most common groups at risk are people who frequent day care centers, homosexually
active men, and intravenous drug users (CDC 2000a). Cases of hepatitis E, whose symptoms include
abdominal pain, dark urine, fever, enlargement of the liver, jaundice, and vomiting (CDC 2001b), are
usually transmitted through fecally contaminated food and water (CDC 200 Ic). Hepatitis E cases have
been identified in the United States among people with no known risk factors, and serology data suggest
that hepatitis E is endemic to the United States (Karetnyi et al. 1999).
2.3.2.5
Enteroviruses
Enteroviruses include the polioviruses, coxsackieviruses, and echoviruses; they are the second
most common viral pathogen known to infect humans. Although adults are less likely to become
infected, everyone is at risk. Infection can occur through contact with a contaminated surface, secretions
from an infected person, or through contact with feces. Most of the time, infected individuals are
asymptomatic, but those who become ill express symptoms of a common cold or flu, such as mild upper
respiratory illness. In rare cases, an infected individual can develop aseptic or viral meningitis, or even
more serious illnesses that affect the heart, brain, or organs. Infections in the United States are most
likely to occur during the summer and fall (CDC 1998a).
Occurrence and Exposure Assessment
for the LT2ESWTR 2-25 December 2005
-------
2.3.2.6
Rotaviruses
Rotavirus is a double-stranded RNA virus, common among children and infants, that can cause
severe diarrhea and can result in hospitalization or possibly death due to dehydration. Symptoms include
watery diarrhea and vomiting for 3 to 8 days with frequent abdominal pains and fever. The incubation
period for the disease is about 2 days (CDC 200Id). Transmission occurs by contact with contaminated
surfaces, ingestion of contaminated food or water, or fecal-oral contact, that is, oral contact with
something that has been contaminated by stool. In temperate climates, the disease is most common in the
winter and early spring (CDC 200Id).
2.3.3 Persistence
Human pathogenic viruses of concern that are common in wastewater include hepatitis A virus
(HAV), hepatitis E virus (HEV), rotavirus, astrovirus, caliciviruses, enteric adenoviruses, and
enteroviruses such as poliovirus, echovirus, and coxsackievirus. These viruses can contaminate source
waters, along with recreational waters and waters used for growing shellfish. Limited research has been
performed recently on virus persistence in the environment. Much of the research on persistence of
viruses in the environment is based on studies of marine waters because of the potential for beach and
shellfish contamination. However, some data on viruses in fresh water and ground water are available,
and some research has been conducted on the ability of viruses to survive wastewater treatment. Data on
persistence of viruses in the environment are presented below.
Sewage contains culturable virus concentrations between 5,000 and 28,000 plaque-forming units
per liter (PFU/L), and plant effluents contain about 50 PFU/L (Metcalf et al. 1995). Sediments typically
contain concentrations 2 to 4 logs higher than those in water. Because cell culture methods are not yet
available for many enteric viruses, the actual number of viral pathogens present in a sample is probably
much higher than these numbers reflect. The implications for source waters that receive wastewater
treatment plant effluent are significant.
Alvarez et al. (2000) found that, in ground water environments, MS2 coliphages (a virus that
infects coliform bacteria) kept at 27°C were inactivated in 8 days. When the temperature was reduced to
4°C, the viruses were reactivated. The same experiment performed with poliovirus resulted in no
reactivation. Dahling and Safferman (1979) studied survival of enteric viruses in an Alaskan river and
found that they survived for 7 days (34 percent) in an ice-covered river. In their review of
hydrogeological characteristics that protect against ground water contamination, Robertson and Edberg
(1997) report that viruses can travel up to 250 to 408 meters in glacial silt and sand aquifers and up to
1600 meters in karst aquifers. A study of animal viruses in nonaerated animal wastes found that virus
inactivation could take anywhere from 1 week to 6 months depending on the virus type, temperature, pH,
and other conditions (Pesaro et al. 1995).
Marzouk et al. (1979) examined ground water samples for the presence of enteroviruses and
indicators following land application of sewage. Indicators included heterotrophic plate count bacteria,
fecal coliforms, and fecal streptococci. Enteroviruses were found in 20 percent of ground water samples,
including 12 samples that contained no detectable fecal organisms.
Survival of viruses and bacteria in ground water was examined by Keswick et al. (1982).
Coxsackievirus was the most stable, followed by poliovirus, fecal streptococci, echovirus, E. coli,
Occurrence and Exposure Assessment
for the LT2ESWTR
2-26
December 2005
-------
rotavirus, and F2 coliphage. Enteroviruses survive longer than 24 days in ground water. Fecal
streptococci proved to be a better indicator of viruses in ground water than E. coli.
Alhajjar et al. (1988) used poliovirus, total coliforms, fecal coliforms, and fecal streptococci to
model the infiltration of effluent from a septic system. Poliovirus (vaccine strain) was tracked from a
single inoculum introduced through a toilet. No tracer bacteria reached the ground water; however,
poliovirus was recovered at a concentration of 62 PFU/100 mL from an inoculum of 108 PFU of virus.
Vaughn et al. (1979) discussed virus response to wastewater treatment and lack of correlation
between bacterial fecal indicators and viruses in the environment. Viruses survived in chlorinated
wastewater effluents treated with combined chlorine doses sufficient to effect a 5 log reduction in
bacterial populations (Berg et al. 1978).
Aulicino et al. (1996) reported reduction of indicator bacteria and viruses between influent and
effluent of a sewage treatment plant. Enteric viruses in raw sewage were reduced 2 to 3 logs by
treatment. Fecal coliforms, fecal streptococci and total coliforms were reduced less than 2 logs.
Land application of sewage sludge has been shown to introduce enteroviruses into the
environment (Sagik et al. 1978). Enteroviruses survive for 28 days (Wellings et al. 1975) and poliovirus
survive 11 days in summer and 96 days in winter in effluent-irrigated soil (Tierney et al. 1977). Viruses
attach to soil particles and are transported through the environment by water (Hurst et al. 1980a). Virus
concentrations in upper layers of soil are higher than in subsurface layers by at least 1 log. Viral
adsorption to particles is variable between viruses (Goyal and Gerba 1979). Echovirus moved through
the soil at a slower rate than poliovirus, and did not survive as long. Drying soil greatly affected virus
survival. The maximum depth to which viruses migrated in soil was 60 cm. Temperature, moisture
content, presence of aerobic microorganisms, adsorption affinity for a particular virus, pH, and mineral
content were the chief influences affecting virus survival in soil (Hurst et al. 1980b).
2.3.4 Response To Disinfection
While some viruses are known to be resistant to disinfection, the general assumption is that
viruses are less resistant than Cryptosporidium. For this reason, treatment processes designed to control
for Cryptosporidium are thought to be adequate for the inactivation of viruses.
Enteroviruses are thermolabile and rapidly destroyed when exposed to a temperature greater than
50°C (Melnick 1996; USEPA 1999c). Some viruses have been shown to be resistant to chlorine
disinfection. EPA (USEPA 1999b) has available a table summarizing inactivation of viruses in water by
free chlorine, based on information adapted from Sobsey (1989). The table indicates levels of
disinfectant concentration and contact time in minutes required for a given level of inactivation.
According to Sobsey, a chlorine residual of 0.5 mg/L or less was sufficient for 2 to 4 log inactivation of
viruses in buffered demand-free water.
In general, enteric viruses are more resistant to free chlorine than are enteric bacteria (SDWC
1980). Recent data for rotaviruses, hepatitis A virus, and virus indicators such as MS2 coliphages support
previous results for other enteric viruses, showing that they can be substantially inactivated by free
chlorine. Norwalk virus is relatively resistant to chlorination, however, compared with other enteric
viruses, such as poliovirus type 1, human rotavirus (Wa), and simian rotavirus (SA-11) (Keswick et al.
Occurrence and Exposure Assessment
for the LT2ESWTR 2-27 December 2005
-------
1985; USEPA 1999b). Some human volunteers became ill from Norwalk virus after drinking water with
a chlorine dose of 3.75 mg/L, a dosage similar to that in most municipal water treatment systems. The
same dose of chlorine was effective against other viruses. The authors concluded that chlorine treatment
alone cannot be relied on to inactivate Norwalk viruses and that doses up to 10 mg/L might be required to
inactivate the virus.
Peterson et al. (1983) evaluated the effect of chlorine treatment on the infectivity of hepatitis A
virus in marmoset monkeys and concluded that the virus was more resistant to chlorine than were other
enteroviruses (USEPA 1999c). Sobsey et al. (1991) studied the chlorine and monochloramine
inactivation kinetics of cell-associated hepatitis A virus (viruses inside cells) and found that cell-
associated hepatitis A was always inactivated more slowly than dispersed hepatitis A virus (viruses
unattached to solids or cells). The authors concluded that disinfection criteria for inactivation of hepatitis
A and other enteric viruses should be based on viruses associated with particulates, because they are
better models for viruses found in water than are free-floating viruses (Sobsey et al. 1991). Payment and
Armon (1989) reported on a virus sampling program in seven drinking water treatment plants; all plants
delivered finished water in which the average cumulative reduction of viruses was 95.15 percent with
disinfection using chlorine or ozone (USEPA 1999c). Melnick (1996) reported that a free residual
chlorine treatment of 0.3 to 0.5 mg/L chlorine can cause rapid inactivation of enteroviruses, but that the
viruses can be protected from such inactivation by organic substances (USEPA 1999c).
Ma et al. (1994) compared chlorine inactivation of poliovirus using cell culture and PCR
methods. They found that with cell culture, 1 minute of exposure to 0.5 mg/L of chlorine was necessary
for inactivation. With PCR, 6 minutes were needed for the same concentration of chlorine. The authors
suggested that PCR was detecting inactivated viruses with undamaged nucleic acids. More recently,
Blackmer et al. (2000) determined that researchers had previously underestimated the contact time
necessary for chlorine to inactivate viruses. Many previous experimenters had observed only one cycle of
cell culture. Blackmer et al. noted that if seemingly inactive viral material from one culture was applied
to a second cell culture, virus activity appeared (infected cultured cells changed morphology). However,
it took several weeks for the results of the cultures to become apparent. The authors also performed PCR
early during the first cell culture, detecting viral nucleic acids within a few days and eliminating the need
to wait for changed cell morphology. With both methods, they found that with a concentration of 0.5
mg/L of chlorine, 10 minutes of exposure was necessary for complete inactivation.
Through a monitoring program, Dee and Fogleman (1992) examined the ability of
monochloramine to inactivate coliphage viruses at a full-scale plant. They determined that
monochloramine alone was insufficient for even 2 log removal of coliphages. Herman et al. (1992)
compared MS2 coliphage inactivation by monochloramine prepared three different ways. They found
that inactivation was most effective when ammonia was applied after chlorine. Inactivation using
preformed monochloramine was ineffective, resulting in only a 1 log reduction after 4 hours.
Little information is available on the mode of action by which chlorine dioxide inactivates
viruses. Viruses react rapidly to chlorine dioxide; when chlorine and chlorine dioxide are combined, the
inactivation appears to be synergistic. Inactivation of poliovirus type 3 by chlorine dioxide has been
documented at rates that increase as pH increases from 5.6 and 8.5 (SDWC 1980). Chen and Vaughn
(1990) found that chlorine dioxide doses of 0.05 to 0.5 mg/L effectively inactivated human and simian
rotaviruses within 20 to 120 seconds, with more efficient inactivation occurring with increasing pH.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-28
December 2005
-------
White (1986) concluded that ozone was a superior virucide. Payment and Armon (1989)
maintain that viruses not eliminated by prechlorination-coagulation-sedimentation-filtration were
relatively insensitive to the ozonation process. Finch and Fairbairn (1991) found that MS2 coliphage
experienced a 4 log reduction when exposed to a residual of 40 ug/L for 20 seconds. However, poliovirus
type 3 was inactivated to a lesser extent (2.4 logs), suggesting that MS2 may not be an effective
enterovirus indicator. As with other oxidizing agents, the efficiency of ozonation is reduced when
dissolved organic matter and reduced inorganic constituents are present (USEPA 1999b).
Viruses in nonturbid water and on exposed surfaces can be inactivated with UV light (Oliver
1997). Viruses as a group appear to be the most UV-resistant organisms and will likely be the limiting
organism in determining a reactor design dose for systems requiring virus inactivation. Currently,
California Title 22 standards for design of UV systems for water reclamation require a dose of 140
mJ/cm2 based on a 4 log inactivation of poliovirus (including a safety factor).
Maier et al. (1995) and Meng and Gerba (1996) reported that a dose range of 20 to 30 mJ/cm2
was required for a 4 log inactivation of poliovirus type 1. Batttgelli et al. (1993) reported that
coxsackievirus required a dose of 29 mJ/cm2 for a 4 log reduction, while hepatitis A virus required a dose
of 16 mJ/cm2 for a 4 log reduction. Wiedenmann et al. (1993) reported similar results, indicating that the
hepatitis A virus required a UV dose of 22 mJ/cm2 for a 4 log reduction. Rotaviruses are more resistant
than hepatitis. Battigelli et al. reported that rotavirus strain SA 11 required a dose of 42 mJ/cm2 for a 4
log inactivation (Battigelli et al. 1993). The most resistant virus type appears to be adenoviruses. Meng
and Gerba (1996) reported that adenovirus required doses of up to 120 mJ/cm2 for a 4 log inactivation,
even more than the 62 mJ/cm2 required to inactivated MS2 coliphage. MS2's resistance to UV radiation
(Malley et al. found that MS2 required doses of 64 to 93 mJ/cm2- depending on water quality) has led
some researchers to consider it a good indicator of the extent of inactivation (Meng and Gerba 1996;
Malley etal. 1999).
Zhang et al. (1991) examined the levels of indicator bacteria and viruses in source water and
treated drinking water. Indicator bacteria (HPC, total coliforms and fecal coliforms), coliphage, and
enteroviruses were reduced but not eliminated by complete treatment (prechlorination,
coagulation-sedimentation, sand filtration and final chlorination). Reductions were 97.95-99.99 percent
for standard bacterial indicators, 90.63 percent for coliphage, and 53.18 percent for enteroviruses.
Complete water treatment was more effective at reducing turbidity, coliforms, and coliphage than
enteroviruses.
Differences in virus susceptibility may be related to the morphology of the virus, such as the type
of nucleic acid present (single- vs. double-stranded RNA or DNA) or the lipid/protein envelope. Studies
are currently underway to systematically investigate the effects of virus morphology on disinfection
effectiveness. Because viruses exhibit the most resistance to UV of the pathogens considered, numerous
studies are underway to determine the fundamental basis for the relative resistance of viruses compared to
other types of microorganisms.
2.4 Waterborne Disease Outbreaks (1997 - Present)
In 1997 and 1998, 10 drinking water-related waterborne disease outbreaks occurred in public
water systems in the United States. One was a Cryptosporidium outbreak, three were Giardia outbreaks,
two were caused by bacteria, and four were characterized by acute gastrointestinal illness of unknown
Occurrence and Exposure Assessment
for the LT2ESWTR 2-29 December 2005
-------
cause (Barwick et al. 2000). Of the 10 outbreaks, two occurred in systems with surface water sources
(including a giardiasis outbreak at an unfiltered system). Three outbreaks occurred in systems served by
springs, and the remaining five occurred in ground water systems.
Several other outbreaks of different types occurred throughout the United States at private water
systems, and 18 outbreaks associated with recreational water occurred (nine of which were caused by
Cryptosporidium) (Barwick et al. 2000).
The largest outbreak associated with drinking water during this period occurred in 1998 in
Williamson County, Texas, where 160,000 gallons of sewage spilled into Brushy Creek. The sewage
infiltrated a karst aquifer and contaminated four wells, one-quarter mile from the creek, with
Cryptosporidium. An estimated 1,400 people were infected. Under normal conditions, the wells, which
were 100 feet deep and encased in cement, would not be influenced by surface water. Extreme drought
and high water demand, however, had lowered the water table, allowing the sewage to be drawn down
into the aquifer (Bergmire-Sweat et al. 1998).
In 1999 and 2000,20 drinking water-related outbreaks occurred in PWSs, where the outbreaks
were caused by microbes or were manifested by acute gastrointestinal illness of unknown cause (AG1)
(Lee et al. 2002). AGI could be caused by chemical as well as biological causes, although symptoms
might differ for chemical agents. One AGI outbreak known to be caused by a chemical but where the
chemical was not identified is excluded from the group of 20 outbreaks. During 1999-2000,
Cryptosporidium caused only one outbreak, and Giardia caused two. Norwalk-like viruses, E. coli
O157:H7, and other bacteria were identified as causes of some of the other outbreaks. AGI was listed as
the source for 8 of the 20 outbreaks.
The largest of the outbreaks during this period was an E. coli outbreak. In 1999, an E. coli
O157:H7 outbreak occurred at the Washington County Fair in New York. More than 900 people were
affected, including 2 who died from kidney failure caused by toxins produced by the O157:H7 strain.
The contaminated water came from a shallow unchlorinated well (CDC 1999).
A list of drinking water-associated outbreaks in PWSs from 1991-2000, based on CDC data, is
shown in Appendix A. This list excludes outbreaks associated with chemical ingestion. It is possible that
additional outbreaks associated with microbiological causes occurred in PWSs during this time, but they
were not reported to CDC.
2.5 Indicators of Fecal Contamination
Monitoring for Cryptosporidium and Giardia is difficult in part due to their low concentrations in
source water. Inaccuracies of current detection methods, discussed in section 3.3.1, also contribute to this
difficulty. Therefore, other organisms are commonly used to indicate the possible presence of pathogens
in drinking water and source water. This section discusses the use of total coliforms, fecal coliforms, and
Escherichia coli (E. coli) as indicators of Cryptosporidium and Giardia.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-30
December 2005
-------
2.5.1 Total Coliforms
Most coliform bacteria are harmless. Total coliform bacteria are common inhabitants of soil and
vegetative material, and their presence in drinking water suggests that treatment was incomplete or a
breach exists in the well or distribution system. For this reason, they are used as indicators. Traditionally,
total coliform bacteria were identified by characteristics common to many species of bacteria from
different genera. For example, coliform bacteria are rod-shaped, do not form spores, are gram-negative,
and break down lactose under certain conditions. More recently, tests for the presence of certain enzymes
have been added as criteria for determining whether bacteria are coliforms (Toranzos and McFeters
1997). All total coliforms belong to the family Enterobacteriaceae, and include members of the genera
Enterobacter, Klebsiella, Citrobacter, and Escherichia. Although total coliform bacteria can indicate
fecal contamination, their presence may be due to other causes. In distribution systems, for instance,
coliform bacteria grow in the organic matter that sometimes accumulates on the interior walls of
distribution pipes (biofilm). Biofllm growth can and does occur when no fecal contamination exists.
2.5.2 Fecal Coliforms
Fecal coliforms compose a subgroup of total coliform bacteria commonly found in the feces of
warm-blooded animals. They grow at a higher temperature than most total coliforms, which explains
their ability to colonize and survive in mammalian intestines (Toranzos and McFeters 1997). Fecal
coliforms typically do not cause disease, but their presence in the environment correlates with that of
several waterborne pathogens. In some cases, however, fecal coliforms, such as some Klebsiella bacteria,
have been found in industrial effluent or other waters with high carbohydrate or plant content and no
apparent fecal contamination. Some fecal coliforms, including E. coli, also have been found in
distribution systems growing in biofilm, and in water from pristine sources (Toranzos and McFeters
1997).
2.5.3 Escherichia coli
E. coli is the primary species of fecal coliform bacteria that normally inhabits the gastrointestinal
tract. Most strains do not cause disease, but a few strains can cause gastroenteritis, urinary tract
infections, neonatal meningitis, and kidney failure. For example, E. coli O157:H7 has been responsible
for several waterborne disease outbreaks in recent years. Considered a better indicator of fecal
contamination than other coliforms (Edberg et al. 2000), E. coli can be differentiated from other fecal
coliform by biochemical tests.
2.5.4 Persistence
The ability of indicator bacteria to survive in the environment can affect their usefulness as
indicators of fecal contamination and associated pathogens. If indicator bacteria die off before pathogens
do, and if no pathogens are monitored, contamination can go undetected. If indicators outlast the
pathogens, the indicators' presence is a false alarm. The survival of bacteria in aqueous environments is
dependent upon a variety of factors, including aggregation, adsorption to particles, sedimentation,
coagulation, flocculation, solar radiation, availability and competition for nutrients, predation by other
microorganisms, lysis by bacteriophage, presence of algal and bacterial toxins, chemical toxicity, and
physicochemical effects such as pH, temperature, and salinity (Evison and Tosti 1980).
Occurrence and Exposure Assessment
for the LT2ESWTR 2-31 December 2005
-------
Coliform bacteria demonstrate a significant growth potential when held in darkness (Grigsby and
Calkins 1980). The effect of visible light on E. coli in natural waters was examined by Barcina et al.
(1989), who showed that biosynthetic processes in I4C labeled glucose uptake experiments were inhibited
by visible light.
Sunlight also affects the survival of fecal coliforms and fecal streptococci in seawater (Fujioka et
al. 1981). Both fecal coliforms and fecal streptococci suspended in seawater were reduced by 1 to 2 logs
within 1 to 4 hr. at 24° C. Fecal streptococci were slightly more stable in seawater than fecal coliforms.
Sunlight can penetrate up to 3.3 m beneath the surface of seawater with bactericidal effects. In a similar
study, when sewage samples were exposed to sunlight, 90 percent of fecal coliforms were inactivated
within 28 to 38 minutes, whereas 90 percent of fecal streptococci were not inactivated after a 2-hour
exposure to sunlight. Ninety to 99 percent of fecal coliforms and fecal streptococci retained upon
membranes were inactivated when the membranes were exposed to sunlight for 10 to 15 minutes (Fujioka
andNarikawa 1982).
Survival of total coliforms, fecal coliforms and fecal streptococci were measured at 0° C for 7
days (Davenport et al. 1976). The percentage of survivors was 8.4 percent for total coliforms, 15.7
percent for fecal coliforms, and 32.8 percent for fecal streptococci. The fecal coliform/fecal streptococci
ratio was > 5 at all locations, suggesting that the ratio may not reliably indicate source of contamination at
low water temperature. The minimum growth temperature of £. coli is 7.5 to 7.8° C (Shaw et al. 1971).
Fecal coliforms, fecal streptococci, and C. perfringens were tested in freshwater and seawater for
survival using diffusion chambers (Davis et al. 1995). A 1 log reduction of fecal coliforms and fecal
streptococci occurred within 85 days in freshwater and marine sediments. E. coli in sediments remained
culturable throughout a period of 68 days, suggesting that sediments provide both protection and
nutrition. Use of diffusion chambers simulates natural conditions, and results in markedly longer survival
times than use of free cell suspensions of E. coli, where reported survival times are considerably shorter.
Bacteria in natural environments are subjected to lethal and sublethai stresses, which impact the
ability to culture them. Injured cells may make up as much as 90 percent of bacterial populations (Domek
et al. 1984). Injured bacteria may be sensitive to agents in culture media such as bile salts, surface tension
reducing agents, dyes, etc., all of which may reduce their recovery. Bacteria exposed to disinfectants
frequently require resuscitation or enrichment before they can be recovered on culture media. When
mEndo and mT7 agar were compared, mT7 consistently recovered more coliforms than mEndo (Du Preez
et al. 1995). Selection of media and methods affects recovery of bacteria from the environment and the
interpretation of results.
Enrichments have been used to resuscitate injured bacteria prior to enumeration from
environmental samples (Davies et al. 1995). There is ample evidence from acridine orange or
immunofluorescence stains that the number of intact bacterial cells far exceeds the numbers cultured by
various media. Culturable but non-viable bacteria continue to respire at low levels and their presence
may be detected by use of electron transport indicators or the activity of inducible enzymes such as
beta-D-galactosidase.
Payment (1999) studied the inactivation of E. coli, other bacteria, and several types of viruses by
chlorine. He concluded that chlorine in water rapidly inactivates E. coli and thermotolerant coliforms, but
the most resistant pathogens can be unaffected for hours.
Occurrence and Exposure Assessment
for the LT2ESHTK
2-12
December 2005
-------
2.6 Summary
Although some Cryptosporidium and Giardia species can survive in different hosts, and some
nonhuman hosts (e.g., cattle) contribute to the widespread occurrence of these protozoa, the primary
species affecting humans are C. parvum and G. lamblia. On occasion, C. felis and C. meleagridis have
infected immunocompromised humans. Transmission is typically from person-to-person, although
outbreaks caused by contaminated drinking water do occur. Several types of viruses can cause illness in
humans in the United States, including adenoviruses, enteroviruses, and hepatoviruses. All of these
viruses can cause gastrointestinal illnesses and may be transmitted from person to person as well as
through drinking water. It is difficult to estimate viruses' contribution to waterborne disease, as they are
often not detected in drinking water during outbreaks.
Extensive research has been performed on the ability of protozoa to withstand environmental
conditions and disinfection; the data obtained are difficult to compare, due to questionable reliability of
methods used to assess viability and infectivity. In general, however, Cryptosporidium oocysts are
resistant to extreme temperatures, particularly freezing, and halogen-based disinfectants have been shown
to be ineffective against Cryptosporidium in practical doses, due to the impermeability of oocyst walls.
Ozone and UV disinfection have been shown to be effective against Cryptosporidium. Giardia are more
susceptible than Cryptosporidium to environmental stress and to most chemical disinfectants. While only
one Cryptosporidium and one Giardia species typically affect humans, numerous types of viruses causing
different illnesses are found in drinking water. Viruses are assumed to be inactivated by the disinfection
required under the Surface Water Treatment Rule; however, specific disinfection needs for each type of
virus vary due to the varied resistance of different viruses to chemical disinfection. Data on UV
disinfection of viruses are limited; however, viruses appear to be more resistant to UV radiation than
Cryptosporidium and Giardia.
Occurrence and Exposure Assessment
for the LT2ESWTR
2-31
December 2005
-------
3. Methods for Characterizing the Occurrence of Pathogens
In an effort to determine the occurrence of pathogens in source water, Cryptosporidium and
Giardia in particular, EPA conducted two surveys, the Information Collection Rule (ICR) and ICR
Supplemental Surveys (ICRSSs). This chapter describes the data sources, laboratory analytical methods
used in the surveys, and data analyses. The data analyses include statistical analyses of observed data and
a Bayesian model that incorporates both observed data and known relationships between data parameters.
A model was desired to account for limitations in laboratory analytical methods, among other difficulties
in accurately estimating the occurrence of pathogens, Cryptosporidium in particular, in source water.
3.1 Data Sources
To assess microbial occurrence and human exposure via drinking water, data were obtained from
various sources. Before the ICR, data were obtained from various small scale studies; these data were
used in the development of the IESWTR and the LT1ESWTR and are characterized in the regulatory
support documents for those rules.
Microbial occurrence data were collected through the ICR and ICRSS monitoring programs for
use in the development of the LT2ESWTR. The ICR was promulgated in 1996, 2 years before the
IESWTR.
3.1.1 Pre-ICR Occurrence Data
Before the ICR data collection effort, no federal monitoring programs required routine
monitoring of PWS source waters to determine the occurrence of protozoa. Previously, evidence of
occurrence of Cryptosporidium and Giardia in drinking water supplies was gathered only through
epidemiological surveillance reports and a limited number of surveys and individual monitoring studies.
Cryptosporidium and Giardia have been shown to exist in drinking water supplies, as indicated by
various studies of waterborne disease outbreaks, but these studies are site-specific and may not be
representative of nationwide occurrence. In addition, comparing results from the various studies is
difficult because the studies have different recovery rates, different sample volumes ("detection limits"),
and they use different detection methods. Some studies also have small sample sizes, which can increase
error. For summaries of outbreak incidences, the reader is referred to the Occurrence Assessment for the
Interim Enhanced Surface Water Treatment Rule (USEPA 1998c) and the Occurrence Assessment for the
Long Term 1 Enhanced Surface Water Treatment and Filter Backwash Rule (USEPA 2000d).
3.1.2 ICR Monitoring Program
The ICR obtained plant-level data sets that link water quality and treatment from source to tap.
Additionally, the ICR described the relationships between treatment conditions and actual population
served (USEPA 2000e). Systems included in the ICR monitoring program are surface and ground water
PWSs serving populations of at least 100,000 people. However, only surface water and GWUDI systems
serving populations of at least 100,000 were required to conduct microbial monitoring. These large plants
sampled source water for Cryptosporidium, Giardia, viruses, and coliforms on a monthly basis for 18
months. These systems also were required, with some exceptions permitted, to monitor their finished
water if 10 or more oocysts or cysts per liter were detected in their raw water during any of the first 12
Occurrence and Exposure Assessment
for the LT2ESWTR
3-1
December 2005
-------
months of monitoring. More detailed information, such as sampling locations and frequencies, can be
found in the ICR Data Analysis Plan (USEPA ZOOOe).
The data were reported and tracked through the ICR Data Management System (DMS), which
contains information on treatment processes used, water source type, and sample data from the PWSs
participating in the ICR monitoring program. The ICR DMS consists of three data systems: the ICR
Water Utility Database System used by PWSs to report data; the ICR Laboratory Quality Control
Database System used by independent laboratories to analyze and report water utility sample quality
control information; and the ICR Federal Database System, which can upload and maintain data from
utilities and laboratories to a central database. For use in the data analysis, several auxiliary databases
were created using the data from the ICR Federal Database System after quality assurance and quality
control (QA/QC) measures were performed. The Auxiliary 1 (AUX1) database was the primary source of
data for the analyses that are discussed in this document. The information in the other auxiliary databases
was generally not needed for this document.
3.1.3 ICR Supplemental Surveys
The ICRSSs were conducted to complement the data collected through the ICR. Additional data
were needed to characterize the distribution of protozoa in source waters because (1) the ICR collected
microbial data only from systems serving at least 100,000 people and (2) the ICR Method has low and
variable recovery. The systems included in the ICRSSs are large (47 systems with more than 100,000
people served), medium (40 systems with 10,000-100,000 people served), and small (40 systems with
fewer than 10,000 people served). Protozoa data were not collected from small systems.
A stratified random sample of 40 plants for medium and large system size categories was
monitored in the ICRSS, for a total stratified random sample of 80 plants. The selection of the stratified
random sample of 40 plants for each system size was conducted in two steps. The plants were first
stratified into two source water categories: plants served primarily by flowing stream-type surface waters
and plants served primarily by reservoir/lake-type surface waters. A random sample of plants was then
selected from each of these strata and recruited to participate in the surveys; the number of plants selected
from each stratum reflected the proportion of plants served by each source water type across the United
States for each plant size.
The surveys also collected protozoa data from a "certainty sample" of seven of the largest surface
water plants along with 40 large randomly sampled plants. The water quality of these 7 plants could be
very different or very similar to the 40 large randomly sampled plants. Because these 7 plants were not
randomly selected, the results from these plants are not included in the national occurrence modeling
discussed in this document.
The ICRSSs were designed to meet four objectives, two primary and two secondary. The
primary objectives are shown below:
1. Characterize the national distributions of protozoa concentration (Cryptosporidium and
Giardid) for plants at large and medium water systems. Also, accurately characterize
protozoa concentrations at individual plants (e.g., characterize the mean, median, and 90th
percentile). These characterizations will support development of national estimates of the
impacts (costs and benefits) of various regulatory options.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-2
December 2005
-------
2. Compare the national distributions of protozoa concentrations for plants in large systems to
the national distribution for plants in medium systems. This comparison will help determine
whether national impacts of required treatment estimated for large systems might be expected
for medium systems and possibly small systems.
The secondary objectives are below:
1. Characterize the distribution of protozoa by source water body type (e.g., reservoirs/lakes vs.
flowing streams) and watershed attributes (e.g., coliform density). This will support a
potential classification of systems into categories of relative risk.
2. Collect data on water quality parameters and disinfection byproduct (DBF) precursors in
source water that will support regulatory impact analyses for medium and small systems.
For medium and large plants, sampling occurred twice per month at each plant for 12 consecutive
months, beginning in March 1999. The samples were analyzed for concentrations of Cryptosporidium
using Method 1622 for the first 4 months; then Method 1623 was implemented for the completion of the
ICRSS so that Giardia concentrations could also be determined. For the large plants, pH, temperature,
turbidity, and coliform measurements were taken with every sample; total organic carbon (TOC)
measurements were taken with every other sample (once a month); and additional water quality
measurements were taken with every matrix spike sample. In addition to these parameters, medium
plants collected additional water quality parameters monthly, including alkalinity, ammonia, bromide, and
UV-254 absorbance. Also, split samples were collected at each plant on five of the sampling dates for
spiking and analysis to estimate the recovery rate of Method 1622/1623.
Small plants were included in the ICRSSs but did not test for protozoa. Twice a month, samples
were analyzed for coliform, pH, temperature, and turbidity. Monthly samples were tested for TOC,
alkalinity, bromide, ammonia, and UV-254.
A comparison of the monitoring programs for the ICR and ICRSS is shown below in Exhibit 3.1.
The data set for medium plants participating in the ICRSS is called the ICRSSM set, the set for large
systems in the ICRSS is called the ICRSSL set.
Occurrence and Exposure Assessment
far the LT2ESWTR
3-3
December 2005
-------
Exhibit 3.1 ICR and ICRSS Comparison
System size of plants participating
(population served)
Number of plants participating
Method for selecting participating
plants
Sample frequency
Sampling period
Total number of samples per plant
Laboratory method
Required sample volume for
Cryptosporidium
Median sample volume analyzed
for Cryptosporidium
Sample concentration process
Microscopic examination process
Average recovery rates for lab
method for Cryptosporidium
ICR
> 100.000
350
All plants in size
category
Monthly
July 1997-
December 1998
18
ICR Method
100 L
3.2 L
Percoll-sucrose
density gradient
centrifugation
Immunofluorescent
assay, differential
interference contrast
(DIG)
12%
ICRSSM
10,000-99,999
40
Stratified random
selection
Semi-monthly
March 1999-
February 2000
24
Method 1622/23
10L
10L
Immunomagnetic
separation
Immunofluorescent
assay, DAPI
staining, and DIG
ICRSSL
>1 00,000
40
Stratified random
selection
Semi-monthly
March 1999-
February 2000
24
Method 1622/23
10L
10L
Immunomagnetic
separation
Immunofluorescent
assay, DAPI
staining, and DIG
43%
Notes: DAPI stands for 4',6-diamidino-2-phenylindole.
The recovery rates for the ICR Method and Method 1622/23 were calculated in spiking studies described in Sections
3.3.2.1 and 3.3.2.3. The recovery rates for plants using Method 1622/23 were not calculated separately for each plant
size category.
3.1.4 Representativeness of ICR and ICRSS Plants
Exhibit 3.2 and 3.3 show locations within each watershed of the 80 ICRSS randomly selected
plants and the 350 ICR plants broken down by filtered or unfiltered status (there are no unfiltered medium
ICRSS plants). All systems serving 100,000 or more people were included in the ICR. The seven large
ICRSS plants that were not randomly selected (four filtered and three unfiltered) are not included in the
exhibit because these plants were not used in the data analysis presented in this document. Note, the
plants not included in the survey are assumed not to be fundamentally different from those included in the
survey. In both maps, there appear to be fewer than the actual number of plants because those plants that
share zip codes are superimposed on one another.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-4
December 2005
-------
Based on the maps, both the medium and large ICRSS plants appear to be geographically
representative of the whole set of ICR plants. This is to be expected, since the ICRSS plants were
randomly selected.
Exhibit 3.2 Distributions of ICRSS Plants
ICRSS Plants
ICRSS Plants
• LG-RANDOM-FILT
a LO-RANOOM-UNF
. Med-FILT
I I Watershed Boundaries
Exhibit 3.3 Distributions of ICR Plants
ICR Plants
ICR Plants
« FILT
o UNF
rn Watershed Boundary
Occurrence and Exposure Assessment
for the LT2ESWTR
3-5
December 2005
-------
3.2 Analytical Methods
Sections 3.2.1 and 3.2.2 review three analytical methods used to determine Cryptosporidium and
Giardia concentrations. Section 3.2.3 describes the method used to detect viruses.
3.2.1 ICR Method
The ICR specified a method for sampling and analysis of protozoa referred to as the ICR Method
(USEPA 1996b), which provided for the quantitative measurement of Cryptosporidium oocysts and
Giardia cysts in surface water. Four basic steps constituted the method: sample collection,
concentration, purification, and assay. Each step offers ample opportunity to introduce error and reduce
the efficiency of recovery.
The protocol calls for samples to be collected by passing water through a polypropylene yarn-
wound filter or Filterite cartridge with a nominal pore size of I micrometer (nm) at a rate of 1 gallon per
minute and pressure of 30 pounds per square inch. The method specifies volumes of 100 liters (L) (26.4
gallons (gal)) of raw water or 1,000 L (264 gal) of finished water to be passed through the filters. These
filters retain oocysts, cysts, and particulate matter. After the appropriate volume of raw or finished water
has been passed through the filter, the filter is delivered to a qualified ICR laboratory for analysis. At the
laboratory, sample preparation begins with removal of the material retained by the filter fibers by washing
the filter with an eluting solution. The material removed (the eluate) is then concentrated by centrifuging.
Next, the Cryptosporidium oocysts and Giardia cysts are separated from other debris by Percoll-sucrose
density-gradient centrifugation. The oocysts and cysts are placed on a cellulose acetate membrane filter
and, using a technique called immunofluorescence assay (IFA), are stained with fluorescent antibodies
specific to Cryptosporidium oocysts and Giardia cysts. The stained oocysts and cysts are then viewed
under a microscope and counted.
To ensure that accurate and valid data were collected for the ICR, the United States
Environmental Protection Agency (EPA) evaluated participating chemical and microbiological
laboratories to ensure they were qualified to perform analyses required for the ICR. To be accepted for
the ICR, a laboratory had to submit an application for EPA's approval, satisfactorily analyze unknown
performance evaluation (PE) samples (described in section 3.3.2.2), and pass an on-site evaluation. In
addition, laboratory personnel assigned to perform the protozoa analyses had to be approved individually.
Any of the several steps of the ICR Method—from collection to analyses—can contribute to
variability in analytical results. Oocysts and cysts can be lost during filtration, elution, concentration of
the eluate, purification, or staining of the concentrate. Optimal recovery of oocysts and cysts through
these steps requires the technician to be well trained and follow the method carefully. Exceeding the
specified flow rate may reduce recovery by forcing cysts and oocysts through the filter. Exceeding the
specified sample volume can result in an excessive amount of debris and reduced recovery during the
elution, concentration, and other steps of the method. Interferences can occur during staining and
counting of the mounted specimens. Identification of the immunofluorescent oocysts and cysts requires a
skilled microscopist and expensive equipment. Even for skilled microscopists, field samples are more
difficult to read than spikes because of the different kinds and amounts of artifacts in the samples. Many
samples contained large amounts of autofluorescent debris, such as algae, diatoms, and other protozoa,
which could obscure Giardia and Cryptosporidium. The assay is complex; each assay requires several
hours of sample preparation. In addition, method specifications requires sample elution to be completed
Occurrence and Exposure Assessment
for the LT2ESWTR
3-6
December 2005
-------
within 96 hours of sample collection to minimize potential changes of the physical and biological
characteristics of the oocysts and cysts in the sample.
The ICR Method specifies the volume of sample to be collected; however, the amount of the
sample analyzed is not specified. As a result, laboratories often analyzed varying volumes when the ICR
Method was applied for analysis of ICR samples. The volume analyzed depended on the sample and
pellet volume and was based on the volume needed to meet the desired detection limit. It also depended
on the workload at the laboratory, the traditional volumes analyzed by a laboratory for a particular
customer, the discretion of the laboratory analyst, and the amount of interfering turbidity or debris in the
sample.
3.2.2 Method 1622 and Method 1623
EPA developed and validated improved analytical protozoan methods, Methods 1622 and 1623,
to detect and enumerate Cryptosporidium oocysts and Giardia cysts in water (USEPA 1999d). The
primary difference between the predecessor method, Method 1622 (USEPA 1999e), and Method 1623 is
that the Method 1622 immunomagnetic separation (IMS) kit includes only reagents for Cryptosporidium
purification, whereas Method 1623 uses the Giardia/Cryptosporidium-combination kit, which includes
reagents for both Cryptosporidium and Giardia purification. The methods were developed to collect
source water protozoan occurrence data during the ICRSSs. The strength of Methods 1622 and 1623 is
that they are performance-based and can be adapted to changes in technology. The principal components
of Methods 1622 and 1623 are the same as the ICR Method; however, there are significant changes in
sample size, filtration medium, filter elution buffer, concentration technique, and the staining technique.
The procedure for analyzing water samples using Method 1623 consists of several components:
sample collection, concentration, purification, and assay. The method involves filtration of a 10-L water
sample through an absolute porosity capsule filter. Cryptosporidium oocysts and Giardia cysts are eluted
from the filter with aqueous buffered salt and detergent solution, and immunomagnetic separation is used
to separate the target organisms from extraneous materials in the water sample. In immunomagnetic
separation, magnetic beads coated with pathogen-specific antibodies react with the pathogens and the
sample is exposed to a magnetic field to separate the beads and attached pathogens from the sample
debris. The magnetic beads used in the immunomagnetic separation procedure are then treated to release
the oocysts and cysts for assay. The assay procedure for the method is performed by staining the oocysts
and cysts with immunofluorescent antibodies and a mixture of 4, 6-diamidino-2-phenylindole (DAPI).
Oocysts and cysts are identified by fluorescence and differential interference contrast (DIC)
microscopically and are classified on the basis of fluorescence of the oocysts and cysts (apple-green
fluorescence) and uptake of DAPI (blue nuclei or blue internal staining). Qualitative and quantitative
analyses are performed by viewing the microscope slide to determine if the microorganisms meet the size,
shape, and fluorescence characteristics of Cryptosporidium oocysts and Giardia cysts and by counting the
total number of microorganisms on the slide confirmed to be oocysts or cysts, respectively.
Method 1623 helps analysts identify and enumerate Cryptosporidium oocysts and Giardia cysts
and minimizes false positives. The method also has a higher recovery (an average of 43 percent of
Cryptosporidium and 53 percent of Giardia in matrix spike water) than the ICR Method (12 percent of
Cryptosporidium and 26 percent of Giardia in matrix spike water). It is postulated that Method 1623 has
higher recovery than the ICR Method due to enhanced features such as optimized filtration procedures
and immunomagnetic separation. The ability to detect Cryptosporidium oocysts and Giardia cysts at low
concentrations depends on the concentration of interfering particles. Interferences from extraneous
Occurrence and Exposure Assessment
for the LT2ESWTR
3-7
December 2005
-------
particles should be less than in the ICR Method because immunomagnetic separation removes most non-
protozoan particles from the sample prior to microscopic observation.
Although Method 1623 has several advantages over the ICR Method, neither method
distinguishes among species of the Cryptosporidium oocysts and Giardia cysts, nor do they determine
viability and infectivity. The method still requires a skilled microscopist and expensive microscopic
equipment for the identification of immunofluorescent oocysts and cysts. Furthermore, each assay
requires several hours of sample preparation.
3.2.3 ICR Method for Viruses
The sample volume that must be collected for viruses is 200 L, and the volume that must be
assayed is 100 L. Samples are collected by passing water through a positively charged filter designed to
trap viruses. Beef extract solution is then poured through the filter under pressure to wash off the filtered
material. The solution containing the filtered virus material, the eluate, is mixed with hydrochloric acid,
which causes a precipitate to form to entrap the virus particles. The precipitate is centrifuged, and the
supernatant is discarded. The pellet is redissolved and centrifuged again. At this stage, the supernatant is
retained; this concentrated sample is divided and used to inoculate cultured cells derived from monkey
kidney (also called BGM) cells. The cells are observed under a microscope for cytopathic effects, such as
cell disintegration and changes in cell morphology (but not cell death). If actual cell death occurs, a
portion of the sample is diluted, and the inoculation is repeated in a new culture until only cytopathic
effects occur. Each flask of cultured cells exhibiting cytopathic effects is counted, and the total is entered
into EPA software that determines the most probable number (MPN) of culturable viruses per milliliter.
This number is later converted to MPN per liter.
3.3 Data Analysis Procedures
This section describes the procedures used to analyze the ICR and ICRSS data. The AUX1
database contains the ICR data for 18 monthly samples from approximately 300 large systems (surface
water and GWUDI systems serving a population of over 100,000 people), representing approximately 500
plants. The ICRSS database contains semi-monthly data collected over 12 months from 47 large systems,
40 medium systems, and 40 small systems (coliform data only). The data analysis and statistical
calculations used data from the AUX1 and ICRSS databases on Cryptosporidium, Giardia, viruses (ICR
data only), and coliforms.
EPA developed a hierarchical model using Bayesian parameter estimation techniques to describe
the uncertainty of individual assays, the variability of occurrence over space and time, and the
contributions of source water type, turbidity, and month. This model used observed data to better
characterize the national distribution of protozoa occurrence in source water. "Occurrence" is the term
used to describe the nationwide distribution of concentrations of organisms. The model is described in
section 3.3.3.2 and detailed in Appendix B.
3.3.1 Challenges in Analyzing Microbial Occurrence Data
Analysis of the ICR microbial data is complex due to multiple factors. This section discusses the
challenges in analyzing microbial occurrence data and indicates reasons for data discrepancies.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-8
December 2005
-------
3.3.1.1 Low Occurrence and Concentration of Microbes
One way to increase the likelihood that a sampling program's findings will be statistically
representative of a system with rare, low-concentration events is to obtain a large number of samples and
large sample volumes. Although the ICR analyses used data from about 5,600 samples, the ability to
predict future sample events of low concentrations at individual plants depends on having sufficient data
available for each plant. The ICR goal was to collect 18 samples of 100 L at every plant and analyze at
least 10 L from each sample. However, the median volume analyzed was 3 L and only 44 percent of
plants had 18 usable observations of Cryptosporidium (after QA/QC controls).
Small numbers of samples and small sample volumes analyzed make accurate predictions of low
concentration sampling events difficult. Consider the following example that uses a Poisson distribution
to estimate the likelihood that samples with zero counts (non-detects) contain Cryptosporidium: in 18
samples of 3.33 L (similar to the median volume analyzed in the ICR analyses), 18 non-detects would still
represent more than a 50-percent chance of having at least 0.01 oocysts/L in the source water and a 30-
percent chance of having 0.02 oocysts/L. Thus, although no Cryptosporidium oocysts were reported
through laboratory analysis in more than 90 percent of the samples in the ICR, it cannot be concluded that
Cryptosporidium oocysts were not present in the source waters. These factors are discussed in more
detail in section 3.3.1.3. To better estimate a more likely true distribution, data were modeled using a
hierarchical model discussed in section 3.3.3.2.
3.3.1.2 Sample Variability in Laboratory Technique
The detection of the true number of Cryptosporidium oocysts and Giardia cysts in a laboratory
sample is complex and can lead to variable results. Average recovery in spiking studies in which a
known amount of oocysts and cysts were added to a sample ranged from 12 to 43 percent for
Cryptosporidium and 26 to 53 percent for Giardia, depending on the condition of the cysts and oocysts,
the spiking method, and the detection method. A detailed discussion can be found in section 3.3.2.
Recovery is calculated as the number of oocysts or cysts detected in the matrix sample less the detects
from the unspiked field sample, divided by the amount in the original spike. Cryptosporidium and
Giardia spiking studies used in the JCRSSs indicate that recovery varies from laboratory to laboratory
and from analyst to analyst.
Some spiking studies resulted in recoveries of more than 100 percent. Sample error may be the
strongest contributing factor to these results; that is, the analyzed portion of the sample may have actually
contained a concentration that exceeded the total sample concentration.
Algae or other constituents in the water could have been counted as protozoa during laboratory
analysis, leading to "false positives" or higher reported concentrations in some of the ICR and ICRSS
observed data. Analyses of blanks in the spiking studies sometimes produced non-zero results, which
supported the existence of false positives in the observed data. Inaccuracy also occurred in counting
spike stocks—oocysts or cysts may clump, and aggregates of 10 or more oocysts may appear as a single
oocyst.
Occurrence and Exposure Assessment
for the LT2ESWTR 3-9 December 2005
-------
3.3.2 ICR and ICRSS Recovery Studies
To determine the percentages of the Cryptosporidium oocysts and Giardia cysts captured in
laboratory analyses, spiking studies were performed. The results of these studies are shown in
Exhibit 3.4.
3.3.2.1 ICR Lab Spiking Program
The purpose of the ICR Lab Spiking Program (LSP) was to assess the recovery of oocysts and
cysts from field samples analyzed with the ICR Method (described in section 3.1). At the time of the ICR
sample collection, a duplicate 100 L sample was collected on two sampling dates from 70 plants during
the last 8 months of ICR monitoring. The duplicate samples were spiked with a known quantity of
Giardia cysts and Cryptosporidium oocysts (based on the use of hemacytometer counting methods) and
filtered. Spiking doses targeted 5,000 oocysts/cysts per 100 L (low-spiking period) and 10,000
oocysts/cysts per 100 L (high-spiking period). Mean recovery using the chamber-counting procedure was
calculated to be 12 percent for Cryptosporidium and 26 percent for Giardia (see Exhibit 3.4 below).
3.3.2.2 ICR Performance Evaluation Study
EPA implemented the ICR PE Study to ensure the ICR laboratories' ability to perform the ICR
Method throughout the duration of the ICR. To become an approved protozoa principal analyst for the
ICR, an analyst was initially required to correctly analyze a set of eight PE samples (seven spiked, one
blank). Ongoing approval required the analyst to correctly analyze two samples per month throughout the
ICR monitoring period. Filters were spiked with mixtures of Cryptosporidium oocysts and Giardia cysts
enumerated using a hemacytometer chamber-counting procedure and analyzed along with unspiked
filters. Mean recovery using this method was 41 percent for Cryptosporidium and 48 percent for Giardia
(see Exhibit 3.4 below). Results were higher for the PE study than the LSP because the filter, not the
water sample, was spiked, and organism losses through the ICR filter are reported to be a significant
source of the overall losses of organisms through the entire method.
3.3.23 ICRSS Matrix Spiking Program
As with the LSP, the purpose of the ICRSS Matrix Spike Program was to assess the recovery of
oocysts and cysts from field samples using Method 1622/1623 (described in section 3.1.2). Method 1622
was used for Cryptosporidium analysis only. During five sampling events during the year-long program,
each plant collected an additional 10 L sample at the same time as the routine sample and shipped both
samples to its assigned protozoa laboratory. The laboratory spiked the additional sample with a known
quantity of Cryptosporidium oocysts and Giardia cysts (the quantity was unknown to the laboratory
performing the analysis) and filtered and analyzed both samples using Method 1622/1623. The spikes
were enumerated using a flow cytometer procedure. Recovery using this method averaged 43 percent for
Cryptosporidium and 53 percent for Giardia (see Exhibit 3.4).
Occurrence and Exposure Assessment
for the LT2ESWTR
3-10
December 2005
-------
Exhibit 3.4 Spiking Study Results
Program Period
Method
Spiking
Enumeration
Technique
Spiking Method
Volume Analyzed
Spiking Doses
Cryptosporidium
Mean Recovery
Cryptosporidium
Spiking Relative
Standard Error*
Giardia
Mean Recovery
Giardia Spiking
Relative Standard
Error
Study / Program
ICR Lab Spiking
Survey
May 1998 -Dec. 1998
ICR Method
Hemacytometer
(Region 1 0)
Bulk, 100 L water
sample spiked in line by
central spiking
laboratory, then filtered
8L
5,000-10,000
12%
Low Spike Period
10% -22%
High Spike Period
11% -24%
26%
Low Spike Period
14% -22%
High Spike Period
11% -23%
ICRPE
Study
Jan. 1997 -Dec. 1998
ICR Method
Hemacytometer
(Contractor)
Spiked directly onto
filter (did not account
for losses due to
filtration)
47 L
200-6,100
41%
7% - 32%
48%
7% - 25%
ICRSS Matrix Spiking
Program
Mar. 1999 - Feb. 2000
Method 1622/23
Flow cytometry (Wisconsin
State Lab)
Bulk, 10 L water sample
spiked by analytical
laboratory, then filtered
10L
80 - 200
43%
1.0%
53%
1%-2%
* The Cryptosporidium Spiking Relative Standard Error accounts for the uncertainty in the amount spiked.
3.3.3 Microbial Occurrence Data Analysis Techniques
This section describes the techniques used for analyzing observed data and for conducting the
Bayesian analysis, which models occurrence using ICR and ICRSS data. These data were then used to
estimate the exposure associated with Cryptosporidium and Giardia in the source water.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-11
December 2005
-------
3.3.3.1 Observed Data Analysis
This section describes how the protozoans were counted when observed under a microscope,
describes any special considerations for analyzing viruses and indicators, and describes how source water
was taken into consideration.
Categorization of Cryptosporidium and Giardia
The protozoa data are divided into four categories for Cryptosporidium (total, empty, amorphous,
and with internal structures) and five categories for Giardia (total, empty, amorphous, with one or more
internal structures, and greater than one internal structure). These categories are described for
Cryptosporidium and Giardia in this section.
Cryptosporidium
Cryptosporidium concentration data are divided into four differential interference contrast
microscopy categories in the AUX1 ICR database and ICRSS database:
• Total: Empty oocysts, amorphous oocysts, and oocysts with internal structures combined.
These are combined in data analysis as an indicator of viable or infectious oocysts to some
degree.
• Empty: A wall is present; appears to be an oocyst. While these oocysts are empty and
presumed to be unlikely to be viable or infectious (at the time of the laboratory analysis), they
still are indicators of the presence of Cryptosporidium.
• Amorphous: Oocysts with internal material structures, but whose structures may be masked
(e.g., because of their position on a slide).
With Internal Structures: Oocysts that have a wall and recognizable internal structures
consistent with Cryptosporidium.
For ICR data, the value for the total nearly always equals the sum of the other categories for each
type of protozoa, but does not in a few cases, due to rounding and possibly to reporting errors. For
instance, in a few cases, a plant appears to have reported a total number of oocysts, but not the numbers of
oocysts in each category.
Because expert opinions vary regarding the data categories that best describe occurrence, three
Cryptosporidium categories were used for analysis and graphical presentations in this document:
• Total
• Non-empty (oocysts with either amorphous or internal structures)
• With Internal Structures
Occurrence and Exposure Assessment
for the LT2ESWTR 3-12 December 2005
-------
Giardia
Giardia concentrations are divided into five categories in both the ICR and the ICRSS: total
cysts, empty cysts, amorphous cysts, cysts with one or more internal structures, and cysts with more than
one internal structure. The four categories used in data analysis are total cysts, non-empty cysts
(amorphous plus cysts with one or more internal structures), cysts with one or more internal structures,
and cysts with more than one structure.
Viruses
Virus data were analyzed as described earlier for ICR data only. Virus occurrence data were not
collected as part of the ICRSS.
Indicators (Coliforms)
Three categories of coliforms were analyzed during these studies: total coliforms, fecal
coliforms, and E. coli. These bacteria were monitored to determine whether any subsets of coliforms
correlate with the presence of pathogens (see Chapter 2 for more on coliforms as indicators of
contamination). Each is analyzed as described earlier. All plants monitored total coliform for the ICR
and ICRSS. Each plant also monitored either fecal coliform or E. coli. A few monitored both fecal
coliform and E. coli.
Analysis of Source Water Type
All microbial data were categorized according to source water type, which was reported by the
treatment plant. All ICR plants sampling for protozoa and viruses were large systems (serving
populations of 100,000 or more), so data are not categorized by system size.
• All-Includes all filtered source water types—surface, GWUDI, and mixed (plant uses both
surface and groundwater sources)
• Flowing Stream (FS)
• Reservoir/Lake (RL)
• Unfiltered
Analysis of ICRSS Observed Data
For the ICRSS, microbial data are categorized according to source water type and system size.
As in the ICR, the source water type for each plant was identified by the treatment system and includes
the following categories:
• All-Includes all water source types (a few unfiltered sources were included in the ICRSS)
' Flowing Stream (FS)
Occurrence and Exposure Assessment
for the LT2ESWTR
3-13
December 2005
-------
• Reservoir/Lake (RL)
In the ICRSS, two system sizes were monitored for protozoa occurrence:
• Large systems (serving populations of 100,000 or more)
Medium systems (serving populations of 10,000 - 99,999)
Observed data are graphically presented in Chapter 4 and in the appendices using cumulative
distribution of plant means (for 80 plants).
3.3.3.2 Modeled Distributions
The ICR and ICRSS data collection efforts faced practical limitations including sample volume,
the number of samples analyzed, and microbial measurement techniques. To account for these limitations
and other sources of variability in the data, model-based occurrence estimates were chosen over observed
data to predict national estimates of occurrence.
Using model-based estimates rather than observed data can also properly account for:
• Variability in the data, based on location, sampling technique, and other contextual variables.
• Low and variable recovery of Cryptosporidium by the analytical method.
• The volume assayed and its adequacy in representing the much larger volume of source water
at the time of sampling.
• The small number of samples assayed at each location and their ability to represent the
average concentration for a given source water during the periods of the survey.
Two hierarchical models (one for filtered plants and one for unfiltered plants) with Bayesian
parameter estimation techniques were developed to help analyze and correct for the limitations of data
collection and variability inherent in the data. Good statistical answers require models that are consistent
with observed data, but flexibility within the model is also required to allow specification of realistic
underlying features not adequately addressed by observed data. Markov Chain Monte Carlo models
provide such flexibility with the following features:
An ability to accommodate many parameters.
« Hierarchical structuring, which is the essential tooi for achieving partial pooling of estimates
and compromising in a scientific way between alternative sources of information.
• An emphasis on inference in the form of distributions or at least interval estimates, rather than
simple point estimates.
• Parameter estimates that are robust over a wide range of model assumptions.
Occurrence and Exposure Assessment
for the L T2ESWTR 3-14 December 2005
-------
• Special statistical issues can be easily incorporated into the modeling framework; these
include the probability of false-positive readings, differential recovery rates, and the ability to
include as an input observed values of zero for the number of microbes and in the model the
probability that a finding of zero microbes reflects the true number of microbes in the source
water.
Model for Source Water Occurrence
In the process of developing the models for characterizing microbial source water occurrence,
EPA developed a comprehensive model, also referred to here as the "full model", for describing filtered
systems, which comprise the majority of surface water systems in the U.S. EPA also developed a
reduced-form model for the unfiltered systems to better accommodate the limited data available on those
systems. The majority of the information presented in this occurrence document describe the inputs,
assumptions, and methods used for the full model. Both the full model and the reduced-form model are
discussed in detail in Appendix B. For consistency sake, EPA used reduced-form models for both filtered
and unfiltered systems to support the economic impact analyses performed te evaluate regualtory
alternatives for the LT2 rule.
The statistical model employed for this analysis is called "hierarchical," because it includes
parameters at three tiers or levels. In the highest tier, three precision parameters describe how random
effects vary around mean zero for locations, months, source water type. A fourth precision parameter
describes additional variability. A number of global fixed effects (an overall median, a coefficient for
turbidity, a false-positive rate, and a percentage of waters with zero occurrence) complete the highest tier.
In the second tier are the effects for individual locations, months, and source water types. At the third,
and lowest tier, are the observed counts, volumes assayed, and turbidity as well as the unobserved
concentrations and measurement recoveries associated with those observations.
The oocyst or cyst count for a particular assay is modeled as a Poisson random variable, with
expected value equal to the product of the concentration, the volume assayed, and the measurement
method's recovery for that assay. A zero count is highly likely when this product is small. Because of
false positives, nonzero counts are possible, even when the actual concentration is zero. Through the
likelihood function (probability of observing what was observed, as a function of parameter values), the
survey data inform us about the model parameters. Appendix B, Modeling Microbial Source Water
Occurrence, provides greater detail about how this is accomplished.
While the ICRSS data sets included oocyst counts, the ICR database contains only concentrations
(concentration per 100L). Counts were derived from the reported concentrations using two methods: (1)
dividing the concentration by the "detection limit" and (2) dividing the concentration by 100 times the
volume assayed. Both methods will produce the same count value when data are reported to the same
degree of accuracy, because "detection limit" is defined as 100 divided by volume assayed. The product
of "detection limit" and volume assayed should be 100, but this is not always the case. Most
discrepancies were explained by rounding (the volume assayed was reported to the nearest liter, while the
"detection limit" had as many as five significant digits). The remaining discrepancies were resolved by
reviewing other microbial data for the same sampling event and determining the appropriate volume
assayed.
Graphs
Using the Bayesian model with MCMC sampling, 500 draws of the complete set of uncertain
variables were produced to evaluate the occurrence distributions. For each of these draws, the model
Occurrence and Exposure Assessment
for the LT2ESWTR
3-15
December 2005
-------
calculates a plant-mean for each plant in the data set. The resulting set of plant-means provides a possible
true distribution that is not inconsistent with the observed data. In the aggregate, the 500 occurrence
distributions describe a range of true distributions that could have reasonably given rise to the 1CR or
ICRSS data set. From these 500 distributions, a central estimate of occurrence was generated with a 90-
percent "credible interval" for the estimate. Exhibit 3.5 is an example of the cumulative distribution of
Cryptosporidium concentrations from this approach. The solid curve represents the central estimate of
the predicted occurrence of microbial pathogens in influent water based on plant-month data. The dotted
curves on either side of the solid curve are the bounds of the 90-percent "credible interval."
Exhibit 3.5 Bayesian Cumulative Distribution of Source Water Occurrence
1.0-
O.fr
t 0.7
to
O.fr
e
rx
g O.S-
1 0.4r
I
o.r
Empirical CDF. 6 per. data, adj tor 11% recovery
Empirical CDF, 12 per data, adj for 11 % recovery
Empirical CDF, 16 per. data, adj for 11 % recovery
<- 196 fof 3501 raw rates of zero
[35!
i mill in n ii
: II ill I III Ml III Mil I SIM (If I sH ill I
1e-005 0.0001
0.001 0.01 0.1 1
Concentration (oocysts/L)
10
100
Note: The plant-means of the 18-month observed ICR data are depicted by the bottom row of hash marks in the lower
right-hand comer of the graph. The top row of hash marks represents the plant-means of the 18-month modeled
data.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-16
December 2005
-------
The dashed lines in Exhibit 3.5 represent the cumulative distributions of the empirical, or
observed (adjusted for recovery), data. They represent the plant-means of source water Cryptosporidium
concentrations, calculated from the 6, 12, and 18 months of ICR data. These three empirical distributions
show that as more data were collected, the closer the observed occurrence distribution came to the
modeled distribution. Comparing the 18-month empirical CDF to the modeled data at the upper ends
{>0.1 oocysts/L) indicates the model fits the observed data. At the lower end (<0.1 oocysts/L) the
empirical CDF is driven by the large number of non-detects for which the model estimates low
occurrence levels. Overall, the comparison supports the model's capability to predict source water
occurrence.
The plant-means of the 18-month observed ICR data are depicted by the bottom row of hash
marks in the lower right-hand corner of the graph. The top row of hash marks represents the plant-means
of the 18-month modeled data. These give another visual perspective of the plant-mean concentrations
that are represented by the distribution curves. The number of plants with non-detects for all months of
sampling and the total number of samples reported in the ICR data set are given in the bottom left-hand
corner of the graph.
Cryptosporidium
The Bayesian models contain parameters that are assigned values, allowing the data to be viewed
in many ways. For Cryptosporidium analysis of filtered plants, the following parameters were assigned
the following values:
Data Subset Total, Non-Empty, and With Internal Structure
• Source Water All, Flowing Stream (FS), and Reservoir/Lake (RL)
• False Positives =1 percent
True Zero =various small percentages (with little effect on results)
• Recovery Rate r^ =beta distribution based on spiking studies.
The distributions used in the Economic Analysis are based on total oocysts. The distribution of
Cryptosporidium occurrence in unfiltered plants is derived using a slightly different model and is
described in Appendix B.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-17
December 2005
-------
Giardia
Giardia data were also run through the Bayesian models with ICR data. The analysis was
performed using the following parameter values:
• Data Subset Total, Non-Empty, With One Internal Structure, and Greater than One
Internal Structure
• Source Water All, Flowing Stream (FS), and Reservoir/Lake (RL)
• False Positives = 0 percent
• True Zero = 0 percent
3.4 Conclusion
The data analysis procedures outlined in this chapter and Appendix B were used to describe the
ICR and ICRSS data pertaining to pathogen occurrence in source water, which are presented in Chapter 4.
Analyzing samples for protozoa using the ICR Method and Methods 1622 and 1623 is a complex task that
may lead to varied results. This variation is evident in the mean and range of recoveries between the two
methods determined from spiked source water sample studies performed under the ICR and the ICRSS.
The ICR Method was characterized by mean Cryptosporidium recoveries of 12 percent, with a range of 0
to 60 percent, while Methods 1622 and 1623 were characterized by mean Cryptosporidium recoveries of
43 percent, with a range of 0 to 100 percent. Similarly, the ICR Method was characterized by mean
Giardia recoveries of 26 percent, with a range of 0 to 83 percent, while Methods 1622 and 1623 were
characterized by mean Giardia recoveries of 53 percent, with a range of 0 to 105 percent. To account for
variability in recovery and other variable factors among plants (e.g., geographical and temporal),
microbial occurrence can be examined more appropriately with Bayesian analyses. This modeling
approach was used to produce the modeled occurrence distributions from the ICR and ICRSS data.
Occurrence and Exposure Assessment
for the LT2ESWTR
3-18
December 2005
-------
4. Occurrence of Pathogens in Source Water
This chapter summarizes the source water microbial occurrence data used in the development of
the LT2ESWTR rule. The primary sources of data are the observed data from the ICR Monitoring
Program and ICRSS, along with modeled results using those data. Sections 4.1 through 4.4 present and
briefly discuss the Cryptosporidium, Giardia, and virus data, as well as potential indicator data for these
microbes. Section 4.5 presents data analyses pertaining to microbial co-occurrence. Section 4.6 discusses
temporal effects and the possibility of seasonal bias in annual estimates of Cryptosporidium occurrence.
For each microbial contaminant (i.e., Cryptosporidium, Giardia, virus, and coliforms), summary
tables contain statistics of observed and modeled ICR and ICRSS data. The statistics presented are
means, medians, and 90th percentiles for each source water type. These statistics are of plant-mean data,
that is, for each plant all monthly data are averaged together and statistics are calculated from those plant-
means. The total number of plants and the number of plants with positive samples are also included in the
summary tables. A discussion of results follows each summary table.
Additional information on the observed occurrence of pathogens in the ICR and ICRSS data are
provided in the appendices. Appendix C contains statistics of all ICR observations by month for each
source water type (as compared to the plant-means in this chapter). Appendix D contains bar charts of
ICRSS data by month for each source water type.
4.1 Cryptosporidium
Cryptosporidium monitoring results are presented in this section and in Appendices C and D.
Observed results are presented in sections 4.1.1.2, and 4.1.1.3 for the ICR and ICRSS, respectively. Note,
due to the different mean Cryptosporidium recovery rates of the two monitoring programs - 12% for ICR,
and 43% for ICRSS - raw data from the two monitoring programs are not directly comparable. Appendix
B provides details of adjustments made in the model for differences in mean recovery rates between the
ICR and ICRSS. Cryptosporidium was not detected in most of the samples (93 percent of ICR samples
and 86 percent of ICRSS samples). Cryptosporidium, however, may have been present in many more
source waters but simply not captured in samples as a result of small sample volumes and low
concentrations. Moreover, even if captured in a sample, Cryptosporidium oocysts were recovered in the
laboratory at an estimated rate of just 12 percent for ICR data (see section 3.3.2).
Another byproduct of small sample volumes is high "detection limits". The median volume
analyzed in ICR samples was only 3 L. If a 3-L volume is analyzed, it cannot have a measured
concentration below 0.33 oocysts/L, the concentration that would be reported if a single oocyst was
detected in that sample (1 oocyst +• 3 L = 0.33 oocysts/L). Section 4.1.2 presents modeled occurrence
estimates that are adjusted to compensate for these sampling and testing conditions.
4.1.1 Observed Results
As described in Chapter 3, oocyst observations were reported as falling into one of three
categories: oocysts containing defined internal structures characteristic of Cryptosporidium ("oocysts
with internal structures"); probable oocysts with characteristic walls and with internal material, but that
material does not appear characteristic of a Cryptosporidium oocyst ("oocysts with amorphous
structure"); and oocyst-type walls not presently containing internal material ("empty oocysts"). The
Occurrence and Exposure Assessment
for the LT2ESWTR
4-1
December 2005
-------
presence of internal structure increases the confidence that the observation is indeed a Cryptosporidiwn
oocyst and not some other item or organism with similar gross morphology. On the other hand, an empty
oocyst, if actually an oocyst, may have contained infectious Cryptosporidium at some time prior to
analysis. For the following analyses, amorphous oocysts and oocysts with internal structures were
combined into a category called "non-empty oocysts." Non-empty and empty oocysts counts were also
combined for a count of "total oocysts."
4.1.1.1 ICR Monitoring Program Results
Exhibits 4.1 and 4.2 summarize the observed ICR Cryptosporidium data for filtered plants for the
18-month monitoring period in terms of selected oocyst categories and source water characteristics.
Exhibit 4.1 Summary of ICR Cryptosporidium Data for Filtered Plants
Source
Total
Number
of Plants
Number and
Percentage of
Plants with at Least
One Positive
Sample
Observed Plant-Mean Data (oocysts/L)
Mean
Median
90* Percentile
Total Cryptosporidium Oocysts
All
FS
RL
338
128
206
147 (43%)
81 (63%)
62 (30%)
0.068
0.137
0.033
0
0.022
0
0.194
0.407
0.058
Non-Empty Oocysts
All
FS
RL
338
128
206
118 (35%)
61 (48%)
52 (25%)
0.038
0.067
0.026
0
0
0
0.121
0.253
0.031
Oocysts with Internal Structures
All
FS
RL
338
128
206
34 (10%)
19 (15%)
15 (7%)
0.011
0.024
0.005
0
0
0
0.001
0.049
0
All = FS + RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake
The percentage of positive samples (7 percent of all total Cryptosporidium observations) and the
percentage of plant positives (43 percent of all plants had one or more positive sample) demonstrate that
Cryptosporidium is relatively rare in many source waters. The observed data, however, do not account
for the ICR Method's low recovery efficiencies and the small sample volumes analyzed. Adjusting for
these factors increases the estimated occurrence several-fold (see section 4.1.2).
As shown in Exhibit 4.1, the mean concentration of total Cryptosporidium plant-means for all
filtered source waters is 0.068 oocysts/L, but the mean of non-empty (observations of amorphous and
internal structures) is 0.038 oocysts/L, 56 percent of total Cryptosporidium levels. The mean plant-mean
Occurrence and Exposure Assessment
for the LT2ESWTR
4-2
December 2005
-------
concentration of oocysts with internal structures is 0.011 oocysts/L, 16 percent of total Cryptosporidium
detections. Exhibit 4.2 displays the cumulative distributions of selected oocyst categories for all filtered
source water types.
As is clear from Exhibit 4.1, the occurrence of Cryptosporidium in flowing stream sources was
greater and more variable than the occurrence in reservoir/lake waters. A comparison of total
Cryptosporidium plant-mean concentrations by source water category shows a substantial difference
between plants with sources of flowing streams and those with reservoir/lake sources. The percentage of
positive samples is also higher for flowing streams than for reservoir/lakes. Exhibit 4.3 displays the
cumulative distributions of plant-mean oocyst concentrations for each filtered source water type for
selected oocyst categories.
Exhibit 4.4 summarizes Cryptosporidium concentrations for unfiltered plants. Unfiltered plants
had a much lower concentrations of total oocysts than filtered plants (0.002 oocysts/L as opposed to 0.068
oocysts/L). This difference is expected, since unfiltered plants must meet high source water quality
standards to avoid the filtration requirements of the Surface Water Treatment Rule. However, the
percentage of plants with one or more detections is higher for unfiltered plants (58 percent vs. 43
percent). This result may be influenced by sampling noise, since the number of unfiltered plants
participating in the ICR is small (12 unfiltered vs. 338 filtered plants).
Exhibit 4.2 Cumulative Distribution of Plant-Mean
Cryptosporidium Concentrations for Filtered Plants for All Source
Water Types—ICR Observed Results
80%
<5 CfW.
Q_ °u% H
C
O 40%
20%
0% -
—-' —
,.^~*^
-&•***
'^f""°*""~"
•~~~— «— «. T o t a I
Non-empty
Intrnl ^tn tr
0.001 0.010 0.100 1.000 10.000
Mean Cryptosporidium (oocysts/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-3
December 2005
-------
Exhibit 4.3 Cumulative Distribution of Plant-
Mean Cryptosporidium Concentrations by
Source Water Type—ICR Observed Results
Ci
c
_ro
Q.
"E
£
60% -
20% -
o.<
' ——.^.^——p^....,. . ,_
* **" f'"air f*1 F«I =
' ^'"*~*~ S RL :
^^ |
)l 0.10 1.00 10.00
Mean Total Cryptosporidium (oocysts/L)
c
Q.
Q.
80%"
0
.^-•^flff^1^ 1
fjf - * -^a**"*1" __sS re
.-*-*" ..^.*"-^^ ' !
\^^^^ —* ;
—^
JO 001 010 1 00 10 00
Mean Non-Empty Cryptosporidium
(oocysts/L)
w »u/.
m
«
0
rf— — ' _ ^,
'-
K\
i
)l 010 00
Mean Cryptosporidium With Internal Structures
(oocysts/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-4
December 2005
-------
Exhibit 4.4 Summary of ICR Cryptosporidium Data for Unfiltered Plants
Total Number of
Plants
Number and
Percentage of
Plants with at Least
One Positive
Sample
Observed Plant-Mean Data (oocysts/L)
Mean
Median
90th Percentile
Total Cryptosporidium Oocysts
12
7 (58%)
0.002
0.001
0.005
Non-Empty Oocysts
12
6 (50%)
0.002
0.001
0.005
Oocysts with Internal Structures
12
2 (17%)
0.0002
0
0.001
4.1.1.2 ICRSS Results
Exhibit 4.5 summarizes the ICRSS Cryptosporidium data from large and medium filtered plants
over the entire 12-month monitoring period, by oocyst structure and source water type. Exhibit 4.6 plots
cumulative distributions for the same data without the breakout by system size and source water type.
Exhibit 4.5 Summary of ICRSS Cryptosporidium Data
Source
Total
Number
of Plants
Number and
Percentage of
Plants with at
Least One
Positive Sample
Observed Plant-Mean Data (oocysts/L)
Mean
Median
90th Percentile
Total Cryptosporidium Oocysts
All
Large
Medium
FS
RL
80
40
40
33
41
68 (85%)
34 (85%)
34 (85%)
32 (97%)
32 (78%)
0.06
0.04
0.08
0.09
0.04
0.02
0.02
0.02
0.04
0.01
0.10
0.10
0.11
0.11
0.06
Occurrence and Exposure Assessment
for the LT2ESWTR
4-5
December 2005
-------
Source
Total
Number
of Plants
Number and
Percentage of
Plants with at
Least One
Positive Sample
Observed Plant-Mean Data (oocysts/L)
Mean
Median
90* Percentile
Non-Em pty Oocysts
All
Large
Medium
FS
RL
80
40
40
33
41
66 (83%)
34 (85%)
32 (80%)
32 (97%)
31 (76%)
0.05
0.03
0.06
0.07
0.03
0.01
0.01
0.01
0.03
0.01
0.08
0.08
0.09
0.08
0.06
Oocysts with Internal Structures
All
Large
Medium
FS
RL
80
40
40
33
41
41 (51%)
20 (50%)
21 (53%)
22 (67%)
17 (42%)
0.02
0.01
0.03
0.03
0.02
0.004
0.002
0.006
0.04
0
0.03
0.03
0.06
0.04
0.03
All = FS + RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake
Cryptosporidium total oocysts were detected in 14 percent of samples overall. Cryptosporidium
results were highly variable, and plant-mean concentrations for total oocysts ranged from a low value of 0
oocysts/L to a high value of 12.1 oocysts/L, with a median concentration of 0.02 oocysts/L. Eighty-five
percent of 1CRSS plants detected Cryptosporidium in at least one sample, well beyond the equivalent 43
percent of plants detecting oocysts in the ICR. Approximately half of the plants detected
Cryptosporidium oocysts with internal structures in at least one sample. Given the larger average sample
volume analyzed in the ICRSS and the higher recovery rates of laboratory Method 1622/1623, one would
expect a higher proportion of plants to detect Cryptosporidium in the ICRSS than in the ICR. Although
more plants did detect Cryptosporidium in the ICRSS, it is also possible that the source waters sampled
during the ICRSS contained higher concentrations of Cryptosporidium, leading to higher detection rates.
Exhibit 4.5 also shows that the mean of the plant-mean concentrations was approximately 0.06
oocysts/L for total Cryptosporidium oocysts and was only slightly lower at 0.05 oocysts/L for non-empty
oocysts. This implies that most of the Cryptosporidium oocysts identified in the ICRSS were not empty
shells. The mean of the plant-mean concentration of oocysts with internal structures was 0.02 oocysts/L.
Plant-mean concentrations in the upper 10th percent! le of concentrations for oocysts with internal
structures ranged from 0.03 to 0.06 oocysts/L.
In general, as shown in Exhibits 4.5 and 4.7, the occurrence of Cryptosporidium oocysts in the
ICRSS was greater in flowing stream plants than in reservoir/lake plants: 97 percent of flowing stream
plants had at least one sample in which Cryptosporidium was detected, while Cryptosporidium was
detected at 78 percent of reservoir/lake plants. The plant-mean concentration of 0.09 oocysts/L for
flowing stream plants is more than double the plant-mean of 0.04 oocysts/L for the reservoir/lake plants.
Occurrence and Exposure Assessment
for the LT2ESWTR 4-6 December 2005
-------
The 90th percentile to the maximum plant-mean of total Cryptosporidium concentrations ranged from 0.11
to 1.18 oocysts/L for flowing stream plants and from 0.06 to 0.40 oocysts/L for reservoir/lake plants.
Similar patterns were evident for non-empty oocysts and oocysts with internal structures. The 90th
percentile to the maximum of plant-mean concentrations of oocysts with internal structures ranged from
0.04 to 0.38 oocysts/L for flowing stream plants and 0.03 to 0.26 oocysts/L for reservoir/lake plants.
Although Cryptosporidium oocysts were detected by a much larger percentage of plants in the
ICRSS, the mean plant-mean concentrations and the plant-mean concentrations on the high end of the
distributions are substantially lower in the ICRSS than in the ICR. For example, the 90th percentile for
total Cryptosporidium for all sources in the ICRSS is 0.10 oocysts/L versus 0.194 oocysts/L for the same
category in the ICR. It is not possible to determine whether this difference was driven by smaller sample
size (only 80 plants participated in the ICRSS), more precise sampling and testing, or true concentration
differences in the source waters at the time of sampling.
The distributions of plant-mean total Cryptosporidium concentrations were similar between large
and medium plants, as shown in Exhibit 4.8, except at the upper end of the distribution. This difference
was attributable to several high measurements at two medium plants, which also contributed to the
difference in the means of plant-mean concentrations—the mean of 0.08 oocysts/L for medium plants is
double that of 0.04 oocysts/L for large plants. The maximum plant mean concentrations (not shown in
Exhibit 4.5) are approximately one order of magnitude greater than the 90th percentiles.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-7
December 2005
-------
Exhibit 4.6 Cumulative Distributions of Plant-Mean Cryptosporidium
Concentrations for Total, Non-Empty, and Internal Oocysts for All
Sources—ICRSS Observed Data
100%
80%
1
re
0-60% •
•5
O
I
0 40% •
P
«
0.
20% •
0% -
^^^z*^*--"
f ^-jf^
J~~ {/
S**""^ /* &
1 ~ " ' 1
•' f
_) • jf
;'•/
. ---' l
—-'r>
100%
80%
60%
40%
20%
. fVA.
0001 0.01 0.1 1 10
Mean Cryptosporidium concentration (oocyste/L)
(Mean of 24 results/plant 80 plants)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-8
December 2005
-------
Exhibit 4.7 Cumulative Distribution of Plant-
Mean Cryptosporidium Concentrations for
Total, Non-Empty, and Internal Oocysts by
Source Water Type—ICRSS Observed Data
100%
1
a am,
1
g p<•-;'''
,.r" -.'''
__ ,/
r
_-/ • .-'"'
- — — "~ "" ' ."
; ' " '
101 001 0.1
Mean Cryptosporidium concentration (oocytttrt.)
(Mean of 24 rwutts/plant)
--- -Ftovsing stream (J3ptort») - - RM*rtAirikk* (41 ptorts) Al(80p**)
v*h (nterrwl Sfructurw
100%-! -.*--•---
«>». t _r/r
« -- f- '"
t .•
40% •
3D*
OCD1 001 01
Mean Cryptosporidium concentration (oocy*ts/L)
(M*inof24resuts/ptant)
Bowirv s»eam{33 ptwti) RMenarfelce (41 ptarti) M{«0 pUrts)
tco%
00%
•0%
40%
20%
0%
D
100%
BOTfr
BOW
4*
am
- -
ion
KHt
49%
20*
Occurrence and Exposure Assessment
for the LT2ESWTR
4-9
December 2005
-------
Exhibit 4.8 Cumulative Distribution of Plant-
Mean Cryptosporidium Concentrations for
Total, Non-Empty, and Internal Oocysts by
Plant Size-ICRSS Observed Data
100%
•i ».
i
a «,
*
& «~
g m
0
itag* by plant (Ize
j $ j
5 w*
09
|
!
£
f
I
Total Cryptcaporiatum Oocysta
j-S*
/r
/-•
^r1
-./"
^;-"
Ml O.OI 01 1 1
Mean CfyfKDtporiaium concentration (oocyaWL)
(Mean of 24 raauH*/plant|
La^e (40 plants) Medlun (40 plants) AKaOptenb)
Non-emjUy OyptotporMiurn Oocytt*
,/•
v~*"
J1 DOT 0.1
Mean Oyfttxaporidfum coowntraflon (oocyBtBflj
(Ma«noT24 rwutte/plint)
C-T73*ti^>oridhjfnCtocv»l»¥^lri^^
vw ___ _ — •
•» f~~~~ i
mm .--'
V*
M*
0001 OOi 0.1 i
(MeanofUrwjitiVpUn*)
Large (40 pto*} Medun (« ptaffcj AI(Mp.ii«)
torn
m
«.
3OV
H%
00%
«K
10k
i
00*
urn
.»
«M
!0%
Occurrence and Exposure Assessment
for the LT2ESWTR
4-10
December 2005
-------
4.1.2 Modeled Results
A hierarchical model with Bayesian parameter estimation techniques was used to model the
source water occurrence of Cryptosporidium and Giardia in an attempt to account for some of the
variability and limitations in the data collection (previously discussed in Chapter 3). A summary of this
model with a description of the graphical display is presented in Section 3.3.3.2. The model is described
in further detail in Appendix B. As noted previously, EPA developed both a full and a reduced-form of
the model. The majority of the description that follows here addresses the full form of the model for
filtered systems and the reduced form for unfiltered systems.
For the modeling effort, each of three separate data sets (ICR, ICRSS Large Systems, or ICRSS
Medium Systems) were used to estimate occurrence of either Cryptosporidium or Giardia in drinking
water sources. Each analysis produced a large sample of occurrence distributions. Each single
occurrence distribution from that large set describes variability in the average concentration among the
nation's drinking water sources. The large "sample" of such distributions, taken as a whole, reveals
uncertainty that is due to limited data and limitations in the analysis methods. This section summarizes
the modeled results, reporting summary statistics and graphics that depict both variability and uncertainty.
Note, as with many modeling efforts of this type, the scope of the uncertainty analysis is constrained by
the specific distributional assumptions adopted in performing the hierarchical modeling, and therefore
results obtained from the analysis represent a lower bound on the overall uncertainty.
4.1.2.1 Modeled Results for ICR Data
Exhibit 4.9 displays the estimated cumulative distribution of total Cryptosporidium for all source
waters (including unfiltered plants). Appendix E contains the graphs of modeled occurrence for all
source water and microbial categories.
The lower part of the curve is built on the model's assumptions, since all observations below 0.01
oocysts/L were zero counts. The model is not limited by the observations of zero below that level and
predicts a positive concentration for nearly all the observed zero counts. The observed data consisted of
196 plant-means (out of 350) with zero concentration. For Cryptosporidium, the model assigned a very
small percentage of true zeroes for the zero count observations, causing the 196 plant-means to have a
positive mean concentration. This also slightly increased the modeled plant-means for some of the low
observed plant-mean concentrations that had zero count observations in the 18-month period.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-11
December 2005
-------
Exhibit 4.9 Cumulative Distributions of Total Cryptosporidium in All
Source Water Types—ICR Modeled Data
Empirical CDF. 18 per data, adj for 11% recovery
1e-005 0.0001
0.001 0.01 0.1 1
Concentration (oocysts/L)
10
100
At the upper portion of the curve the empirical cumulative distribution function converges with
the predicted "true" distribution. The median of reported "detection limits" was 0.2 oocysts/L, which is
approximately where the empirical distribution begins to follow the modeled distribution. Above 0.2
oocysts/L there are slightly more high concentration source waters of observed data than the predicted
data.
The modeled results shown in Exhibit B.9 reflect the median (solid line) and 90% confidence
bounds (dotted lines) of the model output. These were obtained by ordering the results from the 1,000
iterations at each concentration value and using the median of those for the solid line and the 5lh and 95lh
percentile values for the lower and upper bounds.
From these distributions of probable source water occurrence, statistics were calculated for all
source water types and microbial categories. Exhibits 4.10 and 4.11 summarize the results of the ICR
modeled data; the discussion following the exhibit refers to those results.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-12
December 2005
-------
Exhibit 4.10 Summary of ICR Cryptosporidium Modeled Data for All Plants
Source
Modeled Plant-Mean Data (oocysts/L)
Mean
Median
90th Percent! le
Total Cryptosporidium
All
FS
RL
0.57
1.2
0.25
0.048
0.20
0.021
1.3
3.2
0.46
Non-Empty Oocysts
All
FS
RL
0.31
0.60
0.16
0.024
0.081
0.014
0.63
1.4
0.28
Oocysts with Internal Structures
All
FS
RL
0.10
0.23
0.025
0.00062
0.0025
0.00068
0.069
0.27
0.023
All = FS + RL + other categories. FS = Flowing Stream, RL = Reservoir/Lake
Note: Modeled values in Exhibit 4.10 account for low recovery rate and should not be compared directly to similar
statistics presented elsewhere for observed results, which were produced using observations without adjustments for
recovery.
Exhibit 4.11 Summary of ICR Cryptosporidium Modeled Data for
Filtered and Unfiltered Plants
Source
Modeled Plant-Mean Data (oocysts/L)
Mean
Median
90th Percentile
Total Cryptosporidium
Filtered
Unfiltered
0.59
0.014
0.052
0.0079
1.4
0.033
For all plants the median and 90lh percentile values for total Cryptosporidium were 0.048 and
1.3 oocysts/L, respectively. The mean concentration was 0.57 oocysts/L. (This value is not marked on
the graph.) The mean was considerably greater than the median concentration, as expected, because of the
predicted large number of low concentrations and the presence of a few very high data points.
The median concentrations of non-empty Cryptosporidium oocysts (amorphous plus oocysts with
internal structures) (Exhibits E.3, E.6, and E.I 1) were 50, 40, and 67 percent, respectively, of the median
concentration of total Cryptosporidium oocysts for the three source water categories (all, flowing stream,
Occurrence and Exposure Assessment
for the LT2ESWTR
4-13
December 2005
-------
and reservoir/lakes). The difference between the total and non-empty categories was greatest for flowing
streams and least for reservoir/lake plants (i.e., a higher percentage of empty oocysts occurred in flowing
stream plants). The median concentrations of oocysts with internal structures (Exhibits E.4, E.7, and
E.10) were about 1 to 3 percent of the total Cryptosporidium concentration for flowing streams and
reservoirs/lakes, respectively.
Cryptosporidium concentrations in flowing stream plants were much higher than in reservoir/lake
plants. The median concentration of total Cryptosporidium was 0.2 oocysts/L for flowing streams,
compared to 0.021 oocysts/L for reservoir/lake plants. The 90th percentile values were 3.16 and 0.46
oocysts/L, respectively, for flowing stream and reservoir/lake plants.
Because unfiltered sources have higher water quality than most reservoir/lake and flowing stream
sources, occurrence was modeled for the subsets of filtered sources and unfiltered sources (see
Exhibit 4.11). The occurrence of Cryptosporidium was estimated to be substantially lower for unfiltered
sources. Unfiltered systems must have watershed protection programs and source waters that meet very
low turbidity (5 NTU) and low coliform limits.
4.1.2.2 Modeled Results for ICRSS Data
Exhibits 4.12 and 4.13 display the predicted "true" cumulative distribution of total
Cryptosporidium for ICRSS large and medium plants. Appendix E contains the cumulative distributions
of each source water category and Cryptosporidium category. The gap between the empirical ICRSS data
and modeled JCRSS data (seen by comparing Exhibits 4.5 and 4.14) is less than that for the ICR because
the ICRSS observed data had fewer plant-means with zero concentrations.
Exhibit 4.12 Cumulative Distribution of Modeled Data-—ICRSS
Large Plants
1.0
0.9-
0.8-
0.7 j
o.e-
0.5-
0.4-
0.3-
0.2-
0.1-
o.o-1
Etrpinc«l CDF. 24 per <- 6 iof 4Q1 raw rates of zero
1e-005 0.0001 0.001 0.01 0.1 1
Concentration (oocysts/L)
10
100
Occurrence and Exposure Assessment
for the LT2ESWTR
4-14
December 2005
-------
Exhibit 4.13 Cumulative Distribution of Modeled Data—ICRSS
Medium Plants
0.9
O.fr
I' 0.7
| 0.8
i
a> 0.5
ts
^5 Q.^
d 0.3
o.r
0.0
Empirical CDF. 24 per. dlU, »rjj (or 40% recovery
if 40\
raw ratfla of zero
1e-005 0.0001
0.001 0.01 0.1 1
Concentration (oocysts/L)
A comparison of the large and medium plant distributions in Exhibits 4.12 and 4.13 indicates the
medium plants have a slightly wider distribution. However, the medium central estimate curve falls
within the 90 percent credible limits of the large plants, except in the approximate 90-95 percent
cumulative probability range, where the medium plants have higher concentrations.
Exhibit 4.14 summarizes the results of the Bayesian analysis using ICRSS data, categorized by
plant size and source type.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-15
December 2005
-------
Exhibit 4.14 Summary of ICRSS Cryptosporidium Modeled Data
Source
Plant Size
Modeled Plant-Mean Data (oocysts/L)
Mean
Median
90th Percentile
Total Cryptosporidium
AH
All
FS
FS
RL
RL
Large
Medium
Large
Medium
Large
Medium
0.094
0.19
0.10
0.32
0.087
0.083
0.045
0.050
0.077
0.095
0.030
0.023
024
0.33
0.22
0.88
0.19
0.16
Non-Em pty Oocysts
All
All
FS
FS
RL
RL
Large
Medium
Large
Medium
Large
Medium
0.077
0.16
0.071
0.25
0.084
0.073
0.036
0.031
0.055
0.060
0.026
0.016
0.18
031
0.15
0.72
0.18
0.14
Oocysts with Internal Structures
All
All
FS
FS
RL
RL
Large
Medium
Large
Medium
Large
Medium
0.030
0.072
0.018
0.11
0.041
0.038
0.0079
0.0072
0.0099
0.019
0.0070
0.0048
0.067
0.18
0.042
0.43
0.11
0.049
All = FS + RL + other categories. FS = Flowing Stream, RL = Reservoir/Lake
The median and 90th percentiles for total Cryptosporidium were 0.045 and 0.24 oocysts/L for
targe plants and 0.05 and 0.33 oocysts/L for medium plants. Although medium and large plants had
similar median values, the mean value for medium plants is much higher than that for large plants,
because the values at the upper end of the medium plant distribution are higher. The medium plants had
greater concentrations than the large plants for flowing streams, with medians of 0.095 and 0.077
oocysts/L and 90th percentiles of 0.88 and 0.22 oocysts/L, for medium and large plants, respectively. This
pattern was not evident for reservoir/lake—the median and 90th percentiles were slightly greater with
large plants.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-16
December 2005
-------
The median concentrations of non-empty oocysts in all source water types were approximately 80
and 60 percent of the total Cryptosporidium median for large and medium plants, respectively. Oocysts
with internal structures were a small portion of the non-empty oocysts, 18 and 14 percent for large and
medium plants calculated with the medians of all source water types.
4.2 Giardia
4.2.1 Observed Results
Observed Giardia monitoring results are presented below for plant-mean data. In general,
Giardia cysts occurred in almost three times as many samples and in higher concentrations than
Cryptosporidium oocysts did. Unlike the data for Cryptosporidium, Giardia concentrations between 0
and 0.01 cysts/L were reported.
The ICR Giardia data analyses in Appendix C describe statistics for all monthly observations (as
compared to plant-mean data) for each source water category.
4.2.1.1 ICR Monitoring Program Results
Exhibits 4.15 and 4.16 summarize the ICR Giardia results for the entire monitoring period,
categorized by-cyst structure, source water type, and treatment.
Giardia was found in 18 percent of all laboratory samples and in 64 percent of all plants (filtered
and unfiltered). Detected Giardia concentrations were highly dependent on the volume assayed in the
laboratory, with sample concentrations ranging from 0.01 to more than 25 cysts/L. Laboratory tests
showed that mean recovery of Giardia cysts was approximately 25 percent and varied from sample to
sample. The recovery factor also caused Giardia concentrations and detects to be under-reported in the
ICR survey.
Mean concentration of total Giardia (which includes empty, amorphous, cysts with internal
structures, and cysts with more than one internal structure) is 0.28 cysts/L for filtered source waters. The
mean concentrations of non-empty cysts are 50 percent of the mean total Giardia. Cysts detected with
internal structures are 11 percent of the total Giardia, and cysts with more than one internal structure are
7 percent of total Giardia, calculated from the means. Exhibit 4.17 displays the cumulative distributions
of each cyst category for all filtered source water types.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-17
December 2005
-------
Exhibit 4.15 Summary of ICR Giardia Data for Filtered Plants
Source
Number of
Plants
Number of Plants
with Positive
Samples (Percent)
Observed Data (cysts/L)
Mean
Median
90th Percentile
Total Giardia
All
FS
RL
338
128
206
218 (64%)
110 (86%)
105 (51%)
0.282
0.625
0.054
0.024
0.328
0.002
0.815
1.633
0.144
Non-Empty Cysts
All
FS
RL
338
128
206
179 (53%)
97 (76%)
79 (38%)
0.141
0.305
0.034
0.004
0.085
0
0.423
0.638
0.064
Cysts with Internal Structures
All
FS
RL
338
128
207
58 (17%)
36 (28%)
19 (9%)
0.031
0.070
0.006
0
0
0
0.023
0.118
0
Cysts with More Than One Internal Structure
All
FS
RL
338
128
206
43 (13%)
27 (21%)
14 (7%)
0.020
0.046
0.004
0
0
0
0.007
0.087
0
All = FS + RL •+• other categories, FS = Flowing Stream, RL = Reservoir/Lake
Occurrence and Exposure Assessment
for the LT2ESWTR
4-18
December 2005
-------
Exhibit 4.16 Summary of ICR Giardia Data for Unfiltered Plants
Number of Plants
Number of Plants
with Positive
Samples (Percent)
Observed Data (cysts/L)
Mean
Median
90th Percentile
Total Giardia
12
10 (83%)
0.026
0.014
0.054
Non-Empty Cysts
12
10 (83%)
0,019
0.012
0.034
Cysts with Internal Structures
12
5 (42%)
0.002
0
0.005
Cysts with More Than One Internal Structure
12
3 (25%)
0.001
0
0.002
In general, Giardia was detected more often and in larger concentrations in flowing stream plants
than in reservoir/lake plants. Based on total Giardia results, 86 percent of flowing stream plants detected
Giardia compared to 51 percent of reservoir/lake plants. The concentrations were also greater for flowing
stream plants, as plant mean concentrations were more than 10 times as high (mean of 0.625 as compared
to 0.054 cysts/L). The 50 highest reported Giardia concentrations were all from flowing stream plants,
with a maximum value greater than 25 cysts/L. The highest reported concentration for a reservoir/lake
plant was approximately 5.4 cysts/L (Exhibit C.6). Exhibit 4.18 displays the cumulative distributions of
each cyst category for each source water type.
Exhibit 4.17 Cumulative Distribution of Plant-Mean Giardia
Concentrations for All Source Water Types—ICR Observed
Results
8 0 */o -
CO
^— «
10 60% -,
Q.
*±
03
20%-
0% -
0.0
±_^_^C^=^=-= "*
-XC'"
»_ «•** .--'v
__ -.-^
01 0010 0.100
Mean Giardia
— ; ^-^
-"••".,••••-'
j- '
- -Total
one p y
. ->] Im| S true
1000 1C
(cysts/L)
).
300
Occurrence and Exposure Assessment
for the LT2ESWTR
4-19
December 2005
-------
Exhibit 4.18 Cumulative Distribution of Plant-Mean Giardia Concentrations by
Source Water Types—ICR Observed Results
-a 80% •
8 40% -I
**• 20% -
0% -
0
,S~ .**' ,***' '•
x-' / I
^ / ^ \
_,-- ^ / RL i
wjmw'W*" S ™~_.., -„-. i 11 1
^f^ f •""•"•"*•*'"•"••"• A.U |
wr-""1 s~* :
^_^ :
0 01 10 10.0
Mean Total Giardia (cysts/L)
Percent Plants
40% -
s ••""Jsp?*""
~* ^ X
x-'" X" /
..---^ JT /
...--•• ^' /
...«••-" ""^ x-^
t>s
— •All
_^-— ^
000 001 010 1.00 1000
Mean Non-Empty Giardia (cysts/L)
to 80% -\
n
?T 60% •
Q- 20%
0
-- * - L^^'-^^J^ — ^-^^Z. __.
L&fftyfffw""*" — — ** *" ~*
AU
DO 001 010 100 1000
Mean Giardia With One or More Internal
Structures (cysts/L)
100% -1
JA R0% •
c
-------
Exhibit 4.19 Summary of ICRSS Giardia Data
Source
Number of Plants
Number of Plants with
Positive Samples (Percent)
Observed Data (cysts/L)
Mean
Median
90th Percentile
Total Giardia
All
Large
Medium
FS
RL
80
40
40
33
41
66 (83%)
33 (83%)
33 (83%)
31 (94%)
30 (73%)
0.27
0.30
0.24
0.55
0.07
0.06
0,07
0.06
0.29
0.02
0.74
0.72
0.90
1.30
0.13
Non-Empty Cysts
All
Large
Medium
FS
RL
80
40
40
33
41
65 (81%)
32 (80%)
33 (83%)
31 (94%)
29 (71%)
0.21
0.23
0.18
0.43
0.05
0.05
0.05
0.04
0.21
0.01
0.49
0.42
0.74
0.98
0.09
Cysts with One or More Internal Structures
All
Large
Medium
FS
RL
80
40
40
33
41
55 (69%)
28 (70%)
27 (68%)
31 (94%)
20 (49%)
0.08
0.08
0.09
0.18
0.02
0.02
0.02
0.02
0.06
0
0.24
0.20
0.37
0.47
0.03
Cysts with More Than One Internal Structure
All
Large
Medium
FS
RL
80
40
40
33
41
45 (56%)
22 (55%)
23 (58%)
29 (88%)
14 (34%)
0.03
0.03
0.04
0.07
0.01
0.01
0.01
0.01
0,02
0
0.08
0.05
0.12
0.24
0.01
All = FS + RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake
Exhibit 4.20 shows the distribution of plant-mean Giardia concentrations for total cysts, non-
empty cysts, cysts with one or more internal structures (henceforth referred to as "with internal
structures"), and cysts with more than one internal structure (henceforth referred to as "with internal
structures >!")• Giardia cysts were detected in 33 percent of all samples. In addition, 83 percent of the
Occurrence and Exposure Assessment
for the LT2ESWTR
4-21
December 2005
-------
utilities that participated in the surveys detected Giardia in at least one sample, and approximately two-
thirds detected Giardia cysts with internal structures in at least one sample.
The mean of the plant-mean concentrations of total Giardia cysts was approximately 0.27
cysts/L. The mean of the plant-mean concentrations of non-empty cysts was only slightly lower at 0.21
cysts/L. This means that most of the Giardia cysts identified in the ICRSS were not empty shells. The
mean of the plant-mean concentrations of cysts with one or more internal structures was 0.08 cysts/L.
The upper 10lh percentile plant-mean concentrations of cysts with internal structures ranged from 0.24 to
0.98 cysts/L.
Exhibit 4.20 Cumulative Distribution of Plant-Mean Giardia
Concentrations for Total, Non-Empty, Internal, and Internal >1
Cysts—ICRSS Data
100%
80%
a
1
a
•8 "»•
&
C 40% •
8
£
20%-
0% -
• _:/7::^~-'
... -'"x . /j^
•' . - ''ft*"
. • ,'••' 1 f^~
-''' ' *r^
^—S r,Jf*
'/ ''s*^
J --Jr~
/-"
-100%
90%
60%
40%
-20%
Me,
0001 001 0.1 1 10
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
Total (80 plants) --- • Non-empty (80 plants) ttemal > 1 (80 plants) Internal (BO plants)
In general, as seen in Exhibits 4.19,4.21, and 4.22, the occurrence of Giardia cysts was greater in
flowing streams than in reservoir/lake plants; cysts were detected in at least one sample for 94 percent of
flowing-stream plants, compared to 73 percent for reservoir/lake plants. The mean of the plant-mean
concentrations for flowing stream plants was almost an order of magnitude greater than that of
reservoir/lake plants, with means of 0.55 and 0.07 cysts/L, respectively. The upper 10th percentile plant-
mean concentrations of total cysts ranged from 1.30 to 1.69 cysts/L for flowing stream plants and from
0.13 to 0.48 cysts/L for reservoir/lake plants. Similar patterns were observed for plant-mean
concentrations of non-empty cysts, internal cysts, and cysts with internal structures >1. For example, the
upper 10th percentile plant-mean concentrations of internal cysts ranged from 0.47 to 0.99 cysts/L for
flowing stream plants and 0.03 to 0.20 cysts/L for reservoir/lake plants.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-22
December 2005
-------
Exhibit 4.21 Cumulative Distribution of Plant-Mean Giardia Concentrations for
Total and Non-Empty Cysts by Source Water Type—ICRSS Observed Data
8. 100% •
tj 80% •
Sj 60% •
O
in
£ 40% -
o
a
•g 20% -
al 0% •
0.0
Jota\ Giardia Cysts
J _s~-^f
s^^?
^Xx" ....•-''
r~^^
...--"-••"*""""""
01 0.01 0.1 1 1
. 100%
80%
. 60%
- 40%
- 20%
• 0%
0
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
O)
5
o>
e
»
Q.
Nonempty Giardia Cysts
100% -
0.001
0.01 0.1 1
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
- Flowing stream (33 plants) Reservoir/lake (41 plants) All (80 plants)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-23
December 2005
-------
Exhibit 4.22 Cumulative Distribution of Plant-Mean Giardia Concentrations
for Internal and >1 Internal Structure Cysts by Source Water Type—ICRSS
Observed Data
Internal Giardia Cysts
100% -
&
$ 80% •
1
8 60% •
"3
o
in
£ 40% •
e
20% -
i
Q»
Q.
0% •
_/~^""~""~~ ^._-/ f
_/ — " _..'''"
— ~/X~ t^ •" " " w
___/"" ^-"^
— — — "™" „.- — f • * *
^ — ^ t
/
— _ -
. / "
f
0 001 0 01 0 1
- 100%
- 80%
60%
40%
20%
0%
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
~ ™ • • F 10 wing s trc aiti (33 plants) "~ ~ ™* Re s c rvo irfa kc (41 plants) All (80 punts)
1 10a% •
3 80% -
§ 60% -
|
>, 40%
e
S 20% •
* °d
Giardia Cysts with Greater than One Internal Structure
- -"~ " ^'^^r^^'
1 ^X
— - ^ ~
1
01 0.01 0.1 1
- 100%
- 80%
-80%
- 40%
- 20%
^ 0%
Mean Giardia concentration (cysts/L)
-
(Mean of 15 results /plant)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-24
December 2005
-------
Very little difference was observed between the distribution of plant-mean Giardia
concentrations for large and medium plants, as shown in Exhibits 4.19,4.23, and 4.24. The means of the
plant-mean total cyst concentrations for large and medium plants were 0.30 and 0.24 cysts/L,
respectively. The range of the 90lh percent! le for large plants was wider than the range of values from
medium plants, ranging from 0.72 to 3.03 cysts/L for large plants and from 0.90 to 1.34 cysts/L for
medium plants. Distribution trends of Giardia by plant size were similar for non-empty cysts and cysts
with internal structures. For instance, the 90th percentile of plant-mean concentrations of cysts with
internal structures for large and medium plants ranged from 0.20 to 0.98 cysts/L and 0.37 to 0.55 cysts/L,
respectively.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-25
December 2005
-------
Exhibit 4.23 Cumulative Distribution of Plant-Mean Giardia Concentrations
for Total and Non-Empty Cysts by Plant Size—ICRSS Observed Data
Total G/ardiaCysts
80%
60%
§>
$
40%
20%
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
— — Large (40 plants)
'Medium (40 plants)
All (SO plants)
Non-empty G/ard/a Cysts
S "°* -
*M
10
•5. eo* •
20% -=_SV=
0 01
0.1
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
80%
10
— — Large (40 plants)
-Medium (40 plants)
-All(SO plants)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-26
December 2005
-------
Exhibit 4.24 Cumulative Distribution of Plant-Mean Giardia Concentrations
for Internal and >1 Internal Structure Cysts by Plant Size—ICRSS Observed
Data
Internal Giardia Cysts
100%
d>
.B
10
«*
f
ge
001 0.1
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
60%
20%
Large (40 plants)
- -Medium (40 plants)
-All (80 plants)
Cysts with Greater than One Internal Structure
100% •
£
in ao% .
£
IV
°- 60% •
£
O) 40% -
g 20% -
S.
^^^^~ '
• 'f*~^
- -^-^^^
p?~~ ^~"^
___/
100%
ao%
r 60%
40%
20%
0001 001 0.1 1
Mean Giardia concentration (cysts/L)
(Mean of 15 results/plant)
— — Large (40 plants) - - - -Medium (40 plants) All(80 plants)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-27
December 2005
-------
4.2.2 Modeled Results for ICR Data
Exhibit 4.25 displays the predicted "true" cumulative distribution of total Giardia for all source
waters.
Exhibit 4.25 Cumulative Distribution of Total Giardia in All
Source Water Types—ICR Modeled Data
10-
0.9-
0.8-
•g 0.6-
£
g 05-
1 0.4-
E
O 0.3-
0.2
0.1-
o.o-
1Bp»f. .^n
16-005 0.0001 0.001 0.01 0.1
Concentration (cysts/L)
10
100
Exhibit 4.26 summarizes the results of the ICR Giardia modeled data; the following discussion
refers to these results. Graphics of the modeled data are presented in Appendix E. Note that data
modeled for Giardia have only one category for internal structures, whereas the ICR data collection also
included the greater than one internal structure category.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-28
December 2005
-------
Exhibit 4.26 Summary of ICR Modeled Giardia Data
Source
Modeled Results (cysts/L)
Mean
Median
90* Percentile
Total Giardia
All
FS
RL
1.2
2.5
0.35
0.11
0.72
0.039
3.0
6.5
0.65
Non-Empty Cysts
All
FS
RL
0.59
1.1
0.21
0.045
0.22
0.015
1.41
2.8
0.35
Cysts with Internal Structures
All
FS
RL
0.13
0.24
0.044
0.00090
0.0045
0.00038
0.096
0.32
0.024
All = FS + RL+ other categories, FS = Flowing Stream, RL = Reservoir/Lake
Exhibit 4.27 Summary of ICR Giardia Modeled Data for Filtered and Unfiltered
Plants
Source
Modeled Plant-Mean Data (cysts/L)
Mean
Median
90th Percentile
Total Giardia
Filtered
Unfiltered
1.2
0.14
0.11
0.06
3.1
0.40
The median and 90* percentile values for total Giardia were 0.11 and 3.0 oocysts/L, respectively.
The mean concentration was 1.2 cysts/L. The mean was considerably greater than the median
concentration, as expected, because of the predicted large number of low concentrations and the presence
of a few very high data points.
The median concentrations of non-empty Giardia cysts (amorphous plus cysts with internal
structures) (Exhibits E.I 2, E.I 5, and E. 18) were 41, 31, and 38 percent, respectively, of the median
concentration of total Giardia cysts for the three source water categories (all, flowing stream, and
Occurrence and Exposure Assessment
for the LT2ESWTR
4-29
December 2005
-------
reservoir/lakes). The difference between the total and non-empty categories is greater for flowing streams
than for reservoir/lake plants (i.e., a higher percentage of empty oocysts occurs in flowing stream plants).
Giardia concentrations in flowing stream sources were much higher than in reservoir/lake
sources. The median concentration of total Giardia was 0.72 cysts/L for flowing streams, compared to
0.039 cysts/L for reservoir/lake plants. The 90th percentile values were 6.5 and 0.65 cysts/L, respectively,
for flowing stream and reservoir/lake plants. This same pattern was evident for the non-empty and with
internal structure Giardia categories.
Exhibit 4.27 shows the difference between predicted Giardia occurrence in filtered and unfiltered
sources.
4.3 Viruses
Observed results for virus monitoring are shown in Exhibit 4.28 and 4.29. Fewer samples were
taken for viruses than for Cryptosporidium and Giardia, because many plants were eligible for a
monitoring waiver if they could demonstrate that their total coliform density was consistently less than
100 per 100 ml or that their fecal coliform/£. coli density was consistently less than 20 per ml during 6
consecutive months of monitoring. Although viruses were found more frequently than total
Cryptosporidium and Giardia (occurring in 24 percent of samples rather than 7 and 19 percent,
respectively), their concentrations were relatively low compared to concentrations of total oocysts and
cysts. Total Cryptosporidium and Giardia concentrations, however, included empty oocysts and cysts,
which are unlikely to be infectious. In contrast, the virus concentrations represented only those viruses
capable of forming plaques (i.e., those that are infectious).
Exhibits 4.28 and 4.29 summarize the ICR virus detection data for the entire monitoring period,
by source water type and treatment. Virus concentrations are expressed as most probable numbers of
plaque-forming units per liter (MPN/L) (see Section 3.1.3). Appendix C contains results of all monthly
observations (as compared to plant-mean data).
Exhibit 4.28 Summary of ICR Virus Results, Filtered Plants
Source
All
FS
RL
Number of
Plants
215
113
94
Number of Plants
with Positive
Samples
(Percent)
181 (84%)
104 (92%)
70 (74%)
Observed Plant-Mean Results (MPN/L)
Mean
0.036
0.048
0.013
Median
0.004
0.011
0.0031
90* Percentile
0.047
0.12
0.016
All = FS + RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake
Occurrence and Exposure Assessment
for the LT2ESWTR
4-30
December 2005
-------
Exhibit 4.29 Summary of ICR Virus Results, Unfiltered Plants
Number of Plants
2
Number of Plants with
Positive Samples
(Percent)
2 (100%)
Observed Plant-Mean Results
(MPN/L)
Observation #1
0.0009
Observation #2
0.0012
Approximately 24 percent of all influent samples contained viruses, and 84 percent of filtered
plants had virus detections (Exhibit C.10). The median of the plant-mean virus data at filtered plants was
0.0047 MPN/L for all source water types.
Flowing stream filtered plants had greater virus occurrence and higher virus concentrations than
reservoir/lake plants. About 92 percent of flowing stream plants had positive plant-means, as opposed to
about 74 percent in reservoir/lake plants. The mean for flowing stream plants was 0.048 MPN/L, more
than three times higher than the mean for reservoir/lakes. Exhibit 4.30 displays the filtered plant-mean
cumulative distributions for each source water type. Only two unfiltered plants reported virus sampling
data and they were considerably lower than filtered plants.
Exhibit 4.30 Cumulative Distribution of Plant-Mean Virus
Concentrations for All Source Water Types—ICR Observed Results
80% -
Q. °
o 40%
<5
Q.
20% -
0% -
/ Sr '•
/ r"*" /
f / f - -All :
f /V - - - -RL ;
J / / FR i
r* *-•'• V
J J !
r - / / I
r *•'"»* V .... __ _ . _|
'•'"* "° r/ s- \
i
0.0001 0.001 0.01 0.1 1 10
Mean Virus Concentration (MPN/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-31
December 2005
-------
4.4 Indicators
4.4.1 Observed Results
Because of their low concentrations in source water, Cryptosporidium and Giardia are difficult to
monitor. Technical limitations in current detection methods also contribute to the difficulties in assessing
the occurrence of Cryptosporidium and Giardia. Therefore, organisms such as coliforms are commonly
used as indicators of fecal contamination in drinking water. Most coliforms are relatively harmless;
however, coliforms are often found in water that contains disease-producing organisms (see Section 2.5).
Feca! coliforms are a subset of total coliform bacteria commonly found in the feces of warm-blooded
animals. E. coli is considered a better indicator of fecal contamination and can be differentiated from
other fecal coliforms by the production of certain enzymes, specifically beta-galactosidase and
glucuronidase.
4.4.1.1 ICR Monitoring Program Results
Exhibits 4.31 and 4.32 summarize the [CR data collected for each type of indicator, by water
source and treatment. Statistics are calculated from plant-mean data.
Exhibit 4.31 Summary of ICR Coliform Data for Filtered Plants
Source
Number
of Plants
Number of Plants
with Positive
Samples
(Percent)
Observed Plant-Mean Results (CFU/100mL)
Mean
Median
90th Percentile
Total Coliforms
All
FS
RL
340
125
199
339 (100%)
125 (100%)
199 (100%)
2633
4480
1599
276
1174
83
6738
7140
4540
Fecal Coliforms
All
FS
RL
241
87
141
216 (90%)
83 (95%)
125 (89%)
276
605
118
14
101
8
417
789
87
E CO//
All
FS
RL
233
83
141
209 (90%)
75 (90%)
127 (90%)
256
428
215
10
138
4
353
761
44
All = FS + RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake, CPU = colony-forming unit
Occurrence and Exposure Assessment
for the LT2ESWTR
4-32
December 2005
-------
Exhibit 4.32 Summary of ICR Coliform Data for Unfiltered Plants
Number of
Plants
Number of Plants with
Positive Samples
(Percent)
Observed Plant-Mean Results (CFU/IOOmL)
Mean
Median
90th Percentile
Total Coliforms
12
12 (100%)
137
36
482
Fecal Coliforms
12
12 (100%)
2
2
4
£. co/i
5
4 (80%)
1
0.33
2
Nearly all plants had a positive sample for total coliform, fecal coliform, and E. coll during the
data collection period. Total coliform bacteria were detected in 88 percent of all filtered plant samples
(Exhibit C.I la), with fecal coliforms and E. colt detected in approximately 66 and 63 percent of the total
samples (Exhibits C. 12a and C. 13a).
Coliform bacteria generally were detected more often and at higher concentrations in flowing
stream plants than in reservoir/lakes. Filtered plant-mean values for total coliforms in flowing streams
were 180 percent higher than reservoir/lake plants (based on means). The difference between flowing
streams and reservoir/lake plants was greatest with fecal coliform concentrations, which were 413 percent
higher for flowing stream plants than for reservoir/lake plants. The source water concentration of
coliforms from unfiltered plants was substantially lower than filtered plants.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-33
December 2005
-------
4.4.1.2 ICRSS Results
Exhibit 4.33 summarizes the ICRSS data collected for each type of indicator and source water
type.
Exhibit 4.33 Summary of ICRSS Coliform Data
Source
Number
of Plants
Number of Plants
with a Positive
Sample
(Percent)
Mean
(CFUHOOmL)
Median
(CFU/100mL)
90th Percentile
(CFUHOOmL)
Total Coliform
All
Large
Medium
FS
RL
80
40
40
33
41
79 (99%)
40 (100%)
39 (98%)
33 (100%)
40 (98%)
1897
1840
1955
4051
366
417
530
331
1751
77
5848
5848
4491
11,006
718
Fecal Coliform
All
Large
Medium
FS
RL
42
19
23
19
22
40 (95%)
19 (100%)
21 (91%)
19 (100%)
20 (91%)
432
315
528
937
15
18
9
40
146
3
795
611
1074
2408
46
£. coli
All
Large
Medium
FS
RL
57
30
27
22
30
52 (91%)
29 (97%)
23 (85%)
22 (100%)
26 (87%)
118
95
144
278
14
9
18
7
83
3
294
227
369
740
25
All = FS -»- RL + other categories, FS = Flowing Stream, RL = Reservoir/Lake, CFU = colony-forming unit
Exhibits 4.34 and 4.35 plot the distributions of plant-mean concentrations for total coliforms,
fecal coliforms, E. coli, and the percentage of plants with a given concentration. Total coliforms were
detected in 91 percent of all samples analyzed. Approximately two-thirds of the samples analyzed for
fecal coliforms and E. coli were positive. For all three coliform categories, more than 90 percent of the
plants had at least one positive sample. Coliform results in general are highly variable and, for the
Occurrence and Exposure Assessment
for the LT2ESWTR
4-34
December 2005
-------
ICRSS, ranged several orders of magnitude. Often, a few very high results had a strong impact on the
mean values.
Coliform bacteria generally were detected more often and at higher concentrations in flowing
stream plants than in reservoir/lake plants. The means of the total coliform plant-mean concentrations
were 4,051 colony forming units (CFU)/100 mL for flowing stream plants and 366 CFU/100 mL for
reservoir/lake plants. The means of the fecal coliform plant-mean concentrations for flowing stream and
reservoir/lake plants were 937 CFU/100 mL and 15 CFU/100 mL, respectively. The means of the E. coli
plant-mean concentrations for flowing stream and reservoir/lake plants were 278 and 14 CFU/100 mL,
respectively.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-35
December 2005
-------
Exhibit 4.34 Cumulative Distribution of Plant-Mean
Coliform Concentrations by Plant Size—ICRSS
Observed Data
Total Coliform
It-0*» •
1
£ (-0% •
& "•• •
J 101.
0
^.
;if
-^
\ I 10 i ofl I (KHJ I ooeio louooo
MCJID Total Coliform (Tolil Colifarm/100 mL)
Uroe (40 plants) --• -Medium (40 plants) All («0 plants)
IOOS
IDS,
i *
t «-
i
&, 36*. •
0
F«c«l Coliform
^
,^-':>"*
<' /
-.-<'.''--""
-' ..-•
1 I 1Q 101] IUIIO IVtlOO
MHO Ftcil Conform (Ft c .1 Ci liform/1 00 mL)
Larje (1 B plants) --- -Medium (23 plants) Al(42planl9)
E. toll
•' f
Mon £. coN (E. COU/IOO m
Ulgt (30 pfcnts) -- .-Medium (27 pliKli)
AKMpbno)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-36
December 2005
-------
Exhibit 4.35 Cumulative Distribution of Plant-Mean
Coliform Concentrations by Source Water
Type—ICRSS Observed Data
8 ao%
§ ta%.
8
_g» 40%
S »»
a
Total Coliform
/"""/
,.-••' /
f
.---'"': ^-'r
1 1 14 100 1000 10000 100000
Mean Total Coliform (Total Coliform/100 mL)
Flowing stream f)3 plants) ReserwirJlake(«l plans) Al (80 plants)
Fecal Coliform
I "^
S HK
y eo%
Percentage tay so
„> $ -
^^'
^^^
/'
...•'' J
r
01 0.1 1 10 100 HMD 10000
Mean Fecal Coliform (Fecal Coltform/100 mL)
Flowing stream (19 plank) • •- -ResenoiKteke (22 plants) All (42 plants)
E. coll
1 10CT'
•S eo».
|
>. 40*
2 20* •
§ on .
.-. -
.''"'"' ^"
/ /
'' \^-<"~
1 1 ID 100 1000 10000
Mean E. colt (E. co/i/100 mL}
Flowing stream (22 plants) Resets intake (30 plants) All (57 plants)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-37
December 2005
-------
The distributions of large and medium plant-mean concentrations were very similar for total
coliforms, fecal coliforms, and E. coli, with the exception of a small difference around the medians for
fecal coliforms and E. coli. The mean of the total coliform plant-mean concentrations for large plants was
1,840 CFU/100 mL; the mean of the medium plant-mean concentrations was 1,955 CFU/100 rnL. Fecal
coliform plant-means for large and medium plants were 315 and 528 CFU/100 mL, respectively. For E.
coli, the mean of the plant-mean concentrations was 95 CFU/100 mL for large plants and 144 CFU/100
mL for medium plants.
EPA asked utilities participating in the ICRSS to analyze samples for total coliforms and E. coli,
but permitted them to analyze samples for fecal coliforms rather than E, coli if they did not have E. coli
analysis capability. (Several utilities submitted results for both parameters.) Because not all plants
provided data on E. coli or fecal coliforms, caution should be used when analyzing data for these
parameters, because differences between E. coli, fecal coliform, and total coliform concentrations could
be the result of utility-specific effects.
4.5 Co-Occurrence Data Analyses
Identifying relationships between source water pathogen concentrations and other source water
characteristics could be helpful for controlling pathogens in the water treatment processes. Because
Cryptosporidium and Giardia occur in low concentrations and are difficult to detect, it might be
beneficial to use coliform or turbidity as indicators for pathogen presence if a correlation can be shown
between pathogen and indicator levels. Sections 4.5.1 and 4.5.2 present the ICR and ICRSS data analyses
of Cryptosporidium, Giardia, E. coli, coliforms, and viruses relative to indicators. To further assess the
use of indicators, a "microbial index" was developed from the ICR and ICRSS data and is described in
Section 4.5.3.
4.5.1 Turbidity Related to Occurrence of Cryptosporidium, Giardia, E, coli, and Viruses
Turbidity measurements indicate the clarity of water. Material suspended in water increases
turbidity. Periodic or continuous monitoring of turbidity is commonly used in process control for water
treatment. Because protozoa are particulates and often float in water, and because they often adsorb to
suspended material in the water, many have hoped that turbidity can be used to indicate the presence of
protozoa. Exhibits 4.36 and 4.37 contain the plots of source water turbidity compared to observed
Cryptosporidium (total), Giardia (total), E. coli, and viruses (Exhibit 4.36 only) using ICR and ICRSS
data on a log-log scale for individual water samples. No strong correlations are evident for any of the
pathogens. However, Cryptosporidium, and to a lesser extent Giardia concentrations, show a weak
increasing trend with increased turbidity. Because these graphs are plotted on logarithmic scales, values
of zero cannot be shown. Therefore, where protozoans were not present in a sample, their concentrations
were assigned a value of 0.001 oocysts or cysts/L (for Exhibit 4.36) or 0.01 oocysts or cysts/L (for
Exhibit 4.37). Zero values for viruses and E. coli concentrations were also assigned low values so they
could be shown on the graphs.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-38
December 2005
-------
Exhibit 4.36 ICR Total Cryptosporidium, Total Giardia, Viruses, and E. coli vs.
Turbidity
8
o
100 i
10
0.001
Total Cryptosporidium vs. Turbidity
0.01 0.1 1 10 100
Influent Turbidity (NTU)
1000 10000
100 i
10 -
H
o 0.1 •
0.01 -
0.001
0.01
Total Giardia vs. Turbidity
0.1
1 10 100
Influent Turbidity (NTUs)
1000 10000
Occurrence and Exposure Assessment
for the LT2ESWTR
4-39
December 2005
-------
Exhibit 4.36 ICR Total Cryptosporidium, Total Giardia, Viruses, and £. cotf vs.
Turbidity (continued)
100 i
10
1 -
0.1 •
5 0.01 -
0.001
0.0001
Viruses vs. Turbidity
• * «N>»
0.01 0.1 1 10 100 1000 10000
Turbidity (NTU)
1000000
100000
E 10000
§
2- 1000
9. 100
1
111
10
1 -
0.1
£ co/i vs. Turbidity
• * ^ »*,
0.01 0.1
1 10 100
Influent Turbidity (NTUs)
1000 10000
Occurrence and Exposure Assessment
for the LT2ESWTR
4-40
December 2005
-------
Exhibit 4.37 ICRSS Total Cryptosporidium and Total Giardia vs. Turbidity
Total Cryptosporidium vs. Turbidity
10-
1
o
0 1-
S
|
I °'1 "
O
1
0.01 -
0
*
»* * *
* $ «•* * *
• » * *
* *** *» **
* •* + *H^»\* W«»H«P^^« ^» * ** * •
1 1 10 100 1000
Influent Turbidity (NTUs)
Total Giardia vs. Turbidity
100-
(Cysts/L)
o
ital Giardia
H 0.1 -
0.01 -
0
4 » V * •
4u 9^^ + ^ • A -^
. ^^^fr ^r ^UJ^^^^ff ^ • A.
1 1 10 100 1000
Influent Turbidity (NTUs)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-41
December 2005
-------
4.5.2 Indicators Related to Occurrence of Cryptosporidium, Giardia, and Viruses
E. coli is commonly used to indicate the presence of other waterborne pathogens and is thought to
be the best indicator of the three types of coliforms used for this purpose. Exhibits 4.38 and 4.39 display
E. coli concentrations compared to Cryptosporidium, Giardia, and virus (Exhibit 4,38 only)
concentrations using ICR and ICRSS data from individual water samples on a log-log scale. No
correlations with E. coli are visually evident for Cryptosporidium, Giardia, or viruses using these data.
Because these graphs are plotted on logarithmic scales, values of zero cannot be shown. Therefore, where
protozoans were not present in a sample, their concentrations were assigned a value of 0.001 oocysts or
cysts/L (for Exhibit 4.38) or 0.01 oocysts or cysts/L (for Exhibit 4.39). Zero values for viruses and £.
coli concentrations were also assigned low values so they could be shown on the graphs. The same was
done in Exhibits 4.40 through 4.43.
Further analysis of correlation between Cryptosporidium and E. coli, based on comparison of
annual average concentrations of these organisms for each plant, is presented in section 4.5.3.
Exhibits 4.40 and 4.41 show ICR and ICRSS Cryptosporidium, Giardia, and virus (Exhibit 4.40
only) concentrations vs. fecal coliforrn concentrations for individual water samples, and Exhibits 4.42 and
4.43 compare ICR and ICRSS protozoa and virus (Exhibit 4.42 only) levels to total coliform
concentrations for individual samples. No correlation is apparent in these exhibits.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-42
December 2005
-------
Exhibit 4.38 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. E. coli
Total Cryptosporidium vs. E. coli
100
01
0.001
X - » »» *
tir^
N= 1159 of 3374
10 100 1000 10000 100000 1000000
E. colt (Den*ttyf1MmL)
Total Giardia vs. E coli
100
10
I 1
5 0.1
I
0.01
0.001
N=1071 Of 3374
10 100 1000 10000 100000 1000000
£. coli (OeiwIty/KJOmL)
Viruses vs. E. coli
10-
1 •
01
001
0001
t $»««
« » * ***«
* » mm\
N= 415 Of 2080
10 100 1000 10000 100000 1000000
E. coll (Dm«ttynOOmL)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-43
December 2005
-------
Exhibit 4.39 ICRSS Total Cryptosporidium and Total Giardia vs. £ coli
(Oocysts/L)
1
SB
i <
& 0.1 I
o
I
0.01 <
0
Total Cryptosporidium vs. E. co//
»
•
»
• *
** * *
* * * *
• • • 4 X* *+**
I $ •««««»^««N^ ••H* *V*
yN=230of786
1 1 10 100 1000 10000
£.co// (Density/1 OOmL)
Total Giardia vs. E. coli
100 -
j 10-
1
>,
"
i 'i
io,;
0.01 <
0
•
* *
* * *****
* * **
• *v^V^/::«*
• * * *• ^»* • **
N=223 of 786
1 1 10 100 1000 10000
E. co// (Donsity/100mL)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-44
December 2005
-------
Exhibit 4.40 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. Fecal
Coliform
Total Cryptosporidium vs. Fecal Conforms
100
10
01 •
001 '
0001
N=611 of 5614
0.1 1
10 ioo iooo 10000 100000 loooooo
Fecal Conform* (Deniity'lOOmL)
100
10
0.1
0.001
Total Giardia vs. Fecal Collforms
* •»*» '
N= 974 of 3441
0.1 1
10 100 1000 10000 100000
Focal Collform* •«•» •• « * »
N= 298 of 1806
01
10 100 1000 10000 100000
Fecal Coliform* (D*n*Ky/100mL)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-45
December 2005
-------
Exhibit 4.41 tCRSS Total Cryptosporidium and Total Giardia vs. Fecal Coliform
Total Cryptosporidium vs. Fecal Conforms
10 -
o
2. 1-
g ,
1 i
o
fl
,2
0.01 *
0
*
4 *
• •
* *
* »* * * • * *
N=195of512
1 1 10 100 1000 10000 100000
Fecal Conforms (Density/1 OOmL)
Total Giardia vs. Fecal Conforms
100-
r°
M
g
I I'
3
3
o
1- 0.1 «
0.01 J
*f * *
* A
* * ***«*^{ /*** * *
> » • • *• J*
' • •«••««» «•»«*» • •• ••
N=202of512
' * * • ••••••^^•••••^•'••sisiiejsi !•••• •» •** '
0.1 1 10 100 1000 10000 100000
Fecal Conforms (Density/1 OOmL)
Occurrence and Exposure Assessment
for the LT2ESWTR
4-46
December 2005
-------
Exhibit 4.42 ICR Total Cryptosporidium, Total Giardia, and Viruses vs. Total
Coliform
100
f 10
8 ° 1
|001
0001 <
0
Total Cryptosportdlum vs. Total Conforms
» *_« «*»**• *
• *•• • -V •*•«& *»*
* • * •?&*•. V*"** •
• %»«V^V** •• »
••*'. * . •* »
• • * *
N= 612 0*5614
1 1 10 100 1000 10000 10OOOO 1000000
Total Colironra (D«i*Ry/100inL)
100
_ 10 '
t ,
§ 01
!
001
Total Giardia vs. Total Conforms
• « * » *
f !£&!iBJi9|f9Mi»'* -
• » « «*•«« «
N= 576 01 561 1
/
0.1 1 10 100 1000 10000 100000 1000000
Total Colltormt (D*niKy/100mL)
100
10
I •
I °1
5
001
0001
0
Viruses vs. Total Conforms
»
» »
* / *** »>
,v>y *i. «•
M= 183 Of 3367
/
1 1 10 100 1000 10000 100000 1000000
Total Conform* (DemltyflOOmL)
Occurrence and Exposure Assessment
for the LT2ESUTR
4-47
December 2005
-------
Exhibit 4.43 Total Cryptosporidium and Total Giardia vs. Total Coliform
10-
?
e
5,
u
o
2. 1-
|
1 '
& ° 1 (
$
t-
f\ f\4 t
0.01 <
0
Total Cryptosporidium vs. Total Coliforms
*
4 *
* * *»
• «•• « *
^ * * » • * *
^ * ***!.***
N=75of1100
/
1 1 10 100 1000 10000 100000
Total Coliforms (Density/IOOmL)
100
10
0.1 -
0.01
Total Giardia vs. Total Coliforms
*:* *
N=85of 1100
0.1
10 100 1000
Total Coliforms (Density/IOOmL)
10000 100000
Occurrence and Exposure Assessment
for the LT2ESWTR
4-48
December 2005
-------
4.5.3 Microbial Index
An alternate approach to evaluating the relationships between Cryptosporidium concentrations
and potential indicator concentrations in individual samples is to evaluate the average or maximum
Cryptosporidium concentrations and potential indicator concentrations from multiple samples at a plant
over time. This analysis tests the ability of an indicator to identify plants susceptible to Cryptosporidium
levels of concern. That is, in contrast to considering only co-occurrence of pathogens and indicators in
each individual sample presented in the previous sections, this approach considers whether the occurrence
of an indicator in one or more samples from a given location over some period of time is predictive of
occurrence of pathogens at that location regardless of whether the indicator and the pathogen occur
together in any individual samples. Such an approach was applied to ICR and ICRSS data to develop a
"microbial index," which could be used as a screening tool to identify plants that are more likely to be
susceptible to Cryptosporidium levels of concern in their source waters.
In developing the index, the ICR and ICRSS data were evaluated to determine parameters that
could be used to assess a water source's susceptibility to high Cryptosporidium occurrence. Initially,
fecal coliforms, total coliforms, E. coli, viruses, and turbidity were compared to concentrations of total
Cryptosporidium oocysts and oocysts with internal structures. Although preliminary analyses indicated
that total coliform, fecal coliform, and E. coli might be useful screening tools, analyses placed greater
emphasis on E. coli and fecal coliforms because of the direct relationship between these parameters and
fecal contamination. The microbial index approach is based on the assumption that waters exposed to
greater levels of fecal contamination have a greater probability of Cryptosporidium occurrence. After the
microbial index screening protocol was developed using E. coli, turbidity was evaluated in an effort to
improve the screening tool. However, turbidity was found not to improve the screening tool.
4.5.3.1 Expressing Plant Summary Data
Several approaches for presenting plant data were evaluated, including a maximum rolling annual
average, a maximum rolling 6-month average, a plant maximum, and a simple mean. Plant data were not
summarized using plant maxima because of concerns that a utility might be unfairly classified as a result
of analytical error in a single sample. This concern stemmed from the results of the Cryptosporidium
occurrence data collected during the ICR, which may have included overestimates of Cryptosporidium
concentrations in some samples as a result of false positives and extrapolation errors from the analysis of
small sample volumes. As discussed below, overestimates in the ICRSS data set were not considered to
be a significant issue, as a result of steps taken during that study to mitigate false positives and to
standardize the volumes analyzed for each sample.
For the ICR plant data, Cryptosporidium concentrations were summarized as maximum rolling
annual averages and E. coli concentrations were summarized as simple means. The maximum rolling
annual average for Cryptosporidium ICR plant data was used in an effort to characterize the maximum
numbers of Cryptosporidium in a specific source without overestimating Cryptosporidium concentrations,
as mentioned above. The Cryptosporidium maximum rolling annual averages were calculated for each
plant as follows:
1) The average concentration of 12 monthly samples was calculated for each of the seven, 12-
month periods within the ICR sample collection period (i.e., for July 1997-June 1998,
August 1997-July 1998, September 1997-August 1998, etc.). The analysis used data from
plants that had at least nine Cryptosporidium records and nine E. coli records.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-49
December 2005
-------
2) The maximum of the seven 12-month-period averages was considered the maximum rolling
annual average.
Because all ICRSS samples were collected semimonthly within a single year, simple plant means
were calculated for both Cryptosporidium and E. coli for ICRSS data. A minimum of six E. coli results
was required. ICRSSM and ICRSSL results were combined for the microbial index analysis.
4.5.3.2 Plant Source Water Designations
As the microbial index was being developed, it became apparent that it would be necessary to
evaluate plants based on source water type, due to significantly lower E. coli concentrations in
reservoirs/lakes as compared to flowing streams. Source water categories were somewhat ambiguous in
the ICR database because of the use of multiple water sources and ambiguity in how the source water
categories were defined. The data sets used in the index analyses were constructed based on the best
available data from multiple tables in the ICR AUX 1 database. Only data that could be clearly
associated with a source water category were used in these analyses. Only samples from one source type
were used when calculating plant means. This classification approach could result in difficulty in
duplicating these results for the ICR data. For the ICRSS, each plant's source water type was designated
as "flowing stream," "reservoir/lake" or "both." Plants with a source water designation of "both," used
both flowing streams and reservoir/lakes during the ICRSS. Plants using both sources were not included
in the index development.
4.5.33 Microbial Index Design
The microbial index is an approach for categorizing source water data in a way that allows the
effectiveness of the indicator to be evaluated. The format is based on two values:
• The Cryptosporidium concentration that is considered to be the level of concern.
• The potential indicator "trigger" level that would be used to identify potentially high-
Cryptosporidium source waters in the absence of direct Cryptosporidium measurements.
These values were used to sort plants into four categories presented in Exhibit 4.44, based on
Cryptosporidium and indicator concentrations.
Plants with high Cryptosporidium concentrations that exceeded the indicator trigger level
(Exhibit 4.44, box D).
• Plants with high Cryptosporidium concentrations that did not exceed the indicator trigger
level (Exhibit 4.44, box C).
• Plants with low Cryptosporidium concentrations that exceeded the indicator trigger level
(Exhibit 4.44, box B).
• Plants with low Cryptosporidium concentrations that did not exceed the indicator trigger level
(Exhibit 4.44, box A).
E. coli average trigger values presented in this document range from 5 E. coli per 100 mL to 100
E. coli per 100 mL. The Cryptosporidium "level of concern" was 0.075 oocysts/L, the average
concentration above which additional water treatment might be necessary under the LT2ESWTR.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-50
December 2005
-------
Cryptosporidium concentrations below this level were considered low; concentrations above this level
were considered high.
Exhibit 4.44 Summary of Microbial Index Approach
CryptasporWum Indicator Index Table Template
Breakdown of Plant* Into Cryptosporidium "bin"
"Low"
"High"
Breakdown of Plants U»lng "Trigger"
Did not exceed trigger value Exceed trigger value
A
c
T" v: -.= •'. -B • - -;.V:;:'
D
Calculations Used to Asses* Index Performance
What percent of plants with "high" Cryptospondiiim exceed the E. coli trigger?
(Sensitivity of indicator)
What percent of plants wo Ud be required to morutof?
O/(C+D(
(B+D(/(A+B+OO)
4.5.3.4 Censored Coliform Data
The ICR and ICRSS data sets contain censored E. coli data ("greater than" or "less than" results
for individual samples). The Long Term 2 Enhanced Surface Water Treatment Rule's Microbial
Occurrence Subgroup felt that retaining the maximum amount of data on E. coli concentrations in each
plant's source water was important for developing the microbial index. The Subgroup also wanted to
ensure that censored data that might inaccurately characterize E. coli for a particular plant was not used in
development of the microbial index. The following approach to censored E. coli data was agreed upon by
the Subgroup as the most appropriate approach to censored data and was applied to E. coli data during
development of the microbial index:
Low-censored data
* Results below the "detection limit" were replaced with zero if it was <, 10. (For example, <1
E. coli /100 mL would be replaced by zero).
After reviewing the ICR and ICRSS data, the Microbial Occurrence Subgroup noted that
many of the censored E. coli data points are reflective of £. coli levels below the "detection
limits" (i.e., E. coli < 1/100 mL) and agreed that it was appropriate to include this data in
plant source water characterization. In the absence of a more accurate measure of the actual
E. coli concentration, and in recognition that the concentration was very low, relative to other
samples, a value of zero E. coli/\QQ mL was considered most appropriate.
• Results below the "detection limit" were not used if it was > 10. (For example, < 100 E. coli
/100 mL was not used).
The Microbial Occurrence Subgroup wanted to ensure that a plant's source water was not
inaccurately characterized as having high E. coli concentrations in cases where no E. coli
Occurrence and Exposure Assessment
for the LT2ESWTR
4-5 i
December 2005
-------
were detected but the "detection limit" was very high. For example, some utilities, in an
effort to quantify high total coliform concentrations, diluted samples, resulting in E. coli
results with very high "detection limits" (i.e. E. coli <1,000/100 mL). After reviewing the
ICR and ICRSS data, the Subgroup agreed that samples with high "detection limits" should
not be included in the development of the microbial index.
High-censored data
Censored high results were replaced by the censor limit if the limit was >500 (for example,
> 1,600 E. coli /100 mL is replaced by 1,600 E. coli /100 mL).
Some plants did not consistently analyze additional dilutions, as is appropriate to quantify £.
coli levels, despite numerous requests for the analysis of additional dilutions. After
reviewing the ICR and ICRSS data available for each plant, the Microbial Occurrence
Subgroup agreed that retaining high censored data, when the censor limit was high, was the
most appropriate approach to characterizing a plant's source water.
Censored high results were not used if the censor limit was s500 (for example, >350
E. coli/100 mL was not used).
After reviewing the ICR and ICRSS data available for each plant, the Microbial Occurrence
Subgroup agreed that using high censored data when the censor limit was low would result in
inaccurate plant source water characterization, and as a result, deemed the use of such data
inappropriate.
4.5.3.5 Assessment of Microbial Index Performance
The usefulness of the microbial index was evaluated based on sensitivity and the percentage of
plants required to monitor for Cryptosporidium (see Exhibit 4.45). Sensitivity was defined as the
percentage of plants with "high" Cryptosporidium concentrations that would be triggered into monitoring
based on a specific E. coli trigger level. Sensitivity was considered critical because it evaluated the
ability of the index to require Cryptosporidium monitoring when a plant had a source water with "high"
Cryptosporidium concentrations. A high sensitivity value is preferred and can be achieved by lowering
the E. coli trigger value. However, if £. coli trigger values are lowered, more plants would be required to
monitor. A balance between sensitivity and the number of plants required to monitor was considered an
optimum result.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-52
December 2005
-------
1
§
E
o
1
§
UJ
e
o
o
ui
o
5
o
Ui
»
V
a;
0
.
i
I
£
8
e
z
|
£
uj & •
t Si
§
8*
m
Sf
N
t
$
1
-s
O
o
o
s:
!
ui
§
«*
-
£
o z
UJ u.
§ —
->_l
HI DC
o y
e «
= w
§
I
u
UJ ^
I
1
1
i
UJ
i
g
-------
4.5.3.6 Flowing Stream Index Results
The results of the E. coli index for flowing stream plants are shown in Exhibit 4.45 and Exhibit
4.46. A trigger of 50 E. coli per 100 mL is favorable with regard to sensitivity, producing 100-percent
sensitivity for both data sets but resulting in 73 percent oflCR plants, and 62 percent of ICRSS plants,
being required to monitor for Cryptosporidium.
Exhibit 4.46 Effect of Different Trigger Levels on Microbial Index Sensitivity
(Flowing Streams)
I
Q.
•5
S,
s
S.
ICR Sensitivity
ICR Monitor
ICR SS Sensitivity
ICR SS Monitor
10 50
Trigger level (E. coli 1100ml_)
4.5.3.7 Reservoir/Lake Index Results
The results of the E. coli index for reservoir/lake plants are shown in Exhibit 4.45 and Exhibit
4.47. For the ICR, a trigger of 10 E. coli per 100 mL results in a sensitivity of 80-percent and would
require 28-percent of plants to monitor. Lowering the trigger to 5 E. coli per 100 mL for the ICR data
increased the sensitivity to 90 percent but the number of plants required to monitor increased to 43
percent. A greater difference in sensitivities between the triggers is observed for the ICRSS data set, with
sensitivities increasing from 33 percent to 67 percent using triggers of 10 and 5 E. coli per 100 mL,
respectively. The percentage of plants required to monitor based on the ICRSS data increased from 21
percent to 45 percent using triggers of 10 and 5 E. coli per 100 mL, respectively. Although an E. coli
trigger of 5 per 100 mL would be more conservative for identifying source waters with high
concentrations of Cryptosporidium, it would increase the number of reservoir/lake plants required to
monitor for Cryptosporidium from approximately one-quarter to almost one-half.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-54
December 2005
-------
4.5.3.8 Conclusions of Microbial Index Analysis
The results of the analyses indicate that the microbial index could potentially be used to screen
drinking water plant sources for susceptibility to high occurrences ofCryptosporidium. However,
because of ICR and ICRSS data set limitations—including analytical methodology, limited E. coli data,
variations in E. coli method, number of plants, sample volumes analyzed, data reporting anomalies, and
false positive results—verification of the microbial index results with additional, larger Cryptosporidium
and E. coli data sets is imperative.
Exhibit 4.47 Effect of Different Trigger Levels on Microbial Index Sensitivity
(Reservoir/Lake)
I
I
41
a
S.
I ICR Sensitivity
ICR Monitor
ICR SS Sensitivity
ICR SS Monitor
10 50
Trigger level (£. coli 1100mL)
100
Occurrence and Exposure Assessment
for the LT2ESWTR
4-55
December 2005
-------
4.6 Cryptosporidium and Giardia Seasonal Patterns
Temporal events such as spring rains, which increase runoff from farming activities, are thought
to affect the occurrence of protozoa. Analysis of observed data did not allow for geographical
considerations and other factors that substantially affect temporal changes. Analysis of modeled data
allowed comparison of more non-zero concentrations. The modeled total Cryptosporidium concentrations
at all plants were averaged together for each month to provide a monthly mean concentration, as shown in
Exhibit 4.48. Mean concentrations during November and December were lower than in other months,
and the months of June, July (1997 only), August, and September had the highest values. Values during
other months do not show a clear seasonal trend. At the same time, the existence of such a trend cannot
be ruled out.
Exhibit 4.48 ICR Monthly Mean Cryptosporidium
Concentrations—Modeled Data
*&&p-- \ laan^T'^S^xglBgia^^^^H-'J'^srjT^iiT
Month u\A Yur
4.7 Source Water Occurrence—Summary
The source water occurrence of Cryptosporidium and Giardia, based on ICR and ICRSS data,
was low but highly variable, due to difficulties with sampling for low occurrence events and the sample
analysis methods. To account for these factors and other uncertainties, a hierarchical model with
Bayesian parameter estimating techniques was used with the ICR and ICRSS data to provide a more
probable national distribution of Cryptosporidium and Giardia occurrence in source water. The
Cryptosporidium modeled distributions for source water, described in this chapter, were used to estimate
a Pre-LT2ESWTR finished water occurrence (described in Chapter 5) and were used to determine the
extent to which systems would be affected by the LT2ESWTR and the benefits consumers would gain as
a result of the rule (described in the Economic Analysis for the LT2ESWTR).
Occurrence and Exposure Assessment
for the LT2ESWTR
4-56
December 2005
-------
Based on the observed ICR and ICRSS data, flowing stream sources proved to be more
susceptible to Cryptosporidium and Giardia occurrence than reservoir or lake sources. The ICRSS data
showed little to no difference between occurrence in large and medium plants.
The percentages of unfiltered ICR plants with samples positive for Cryptosporidium, Giardia,
viruses, and coliforms were higher than those for filtered ICR plants, but the overall number of unfiltered
ICR plants was very small. For viruses and coliforms, 80 to 100 percent of unfiltered plants had positive
samples. However, as expected, the unfiltered sources had very low concentrations of pathogens and
coliforms, due to the high source water quality standards unfiltered systems must meet.
The modeled data show most plants having low Cryptosporidium occurrence; 50 percent of plants
had plant-mean levels below 0.048 oocysts/L based on ICR data and 90 percent of plant-means fell below
0.24 oocysts/L and 0.33 oocysts/L, based on ICRSS large system and medium system data, respectively.
No direct correlation was found between Cryptosporidium, Giardia, and virus occurrence and
coliform or turbidity levels in the observed data at the individual sample level. However, a microbial
index developed to find a connection between selected pathogen and indicator levels suggests that higher
E. coli levels can indicate an increased likelihood that pathogens are present.
Occurrence and Exposure Assessment
for the LT2ESWTR
4-57
December 2005
-------
5. Treatment by Physical Removal
Treatment systems remove Cryptosporidium oocysts, Giardia cysts, other pathogens, and
particles to protect consumers from drinking contaminated water. Because Cryptosporidium oocysts are
resistant to inactivation by chlorine (the most commonly used disinfectant in the United States) and
because of the health concerns associated with high concentrations of disinfection byproducts, significant
attention has been focused on the efficacy of filtration for removing pathogens. This chapter summarizes
information on the removal of Cryptosporidium, Giardia, and viruses using filtration technologies,
including bench-scale and pilot-scale studies of removal efficiencies. Treatment via disinfection is
discussed in Chapter 2.
5.1 Removal of Cryptosporidium and Giardia
Many studies have been conducted to evaluate the level of Cryptosporidium and Giardia removal
achieved by filtration and the conditions under which optimal removal occurs. Aspects of
Cryptosporidium and Giardia removal and the results of those studies are discussed in this section and
summarized in Exhibit 5.1. Results from studies using conventional and direct filtration treatment trains
are discussed in sections 5.1.1 to 5.1.3.
Exhibit 5.1 Cryptosporidium and Giardia Removal Efficiencies
Type of
Treatment Plant
Conventional
filtration plants
Log Removal
Cryptosporidium 0.2->5.3
Cryptosporidium 4.2-5.2
Giardia 4.1-5.1
Cryptosporidium 1 . 9-4.0
Giardia 2.2-3.9
Cryptosporidium 1 .9-2.8
Giardia 2.8-3.7
Cryptosporidium 2-2.5
Giardia 2-2.5
Cryptosporidium 2.7-3.1
Giardia 2.2-2.8
Cryptosporidium 3-5
Giardia 3-5
Experimental
Design
Pilot plant
Pilot plants
Pilot plants
Full-scale plants
Full-scale plants
Full-scale plants
Pilot plant
Reference
Dugan et al. 2001
Pataniaetal. 1995
Nieminski and Ongerth
1995
Nieminski and Ongerth
1995
LeChevaltier et al.
1991
LeChevallier and
Norton 1992
McTigue et al. 1998
Occurrence and Exposure Assessment
for the LT2ESWTR
5-1
December 2005
-------
Type of
Treatment Plant
Direct-filtration
plants
Slow sand
Diatomaceous
earth
Microfiltration and
Ultrafiltration
membranes
Nanofiltration
membrane
Cartridge filtration
Cartridge filtration
Log Removal
Cryptosporidium 2.7-5.9
Giardia 3.4-5.0
Cryptosporidium 2.7-3.1
Giardia 3.05-3.6
Cryptosporidium 1 .9-3.8
Giardia 2.9-4.0
Cryptosporidium 3.9-5.4
Cryptosporidium 2-3
Cryptosporidium 3.7-4.5
Cryptosporidium 0.25-5.6
Cryptosporidium 1.3-5.8
Cryptosporidium 3.3-4.2
Cryptosporidium 3.9-6.1
Giardia 2.8->6.5
Cryptosporidium >3.6
Cryptosporidium 0.1-0.5
Giardia 0.9-1 .4
Cryptosporidium 2.8->6.4
Cryptosporidium 4.6->6.4
Giardia 4.3->6.7
Cryptosporidium 5.0-6.6
Cryptosporidium >6.1->7.0
Giardia >6.4->7.0
Cryptosporidium 5-6
Giardia 5-6
Cryptosporidium 1 . 1 -> 3 . 5
Cryptosporidium 3.2-3.6
Experimental
Design
Pilot plants
Pilot plants
Pilot plants
Pilot plant
Pilot plant
Pilot plant
Pilot plants
Pilot plant
Bench-scale
Pilot plant
Pilot plant
Full-scale plant at 0.5 to
1.0°C
Pilot plant
Pilot plant
Pilot plant
Pilot-scale plant
Pilot plant
Bench-scale
Full-scale filter
Reference
Pataniaetal. 1995
Ongerth and Pecoraro
1995
Nieminski and Ongerth
1995
Pataniaetal. 1999
Westetal. 1994
Yatesetal. 1997
Huck et al. 2000
Emelko et al. 2000
Emetko et al. 1 999
Schuler and Ghosh
1991
Timmset al. 1995
Fogeletal. 1993
Hall etal. 1994
Schuler and Ghosh
1990
Ongerth and Mutton
2001
Jacangelo et al. 1995
Seydeet al. 1998
Schaubetal. 1993
Roessler1998
5.1.1 Conventional and Direct Filtration
The typical treatment train for removing particles, including Cryptosporidium, involves addition
of coagulating chemicals (e.g., alum) to the raw water, causing the particles to coalesce into larger
particles called floe. The floe either settles to the bottom of a sedimentation basin or is removed through
subsequent filtration. Conventional filtration refers to a treatment process that includes coagulation,
flocculation, sedimentation, and filtration. Direct filtration systems do not have a sedimentation process;
coagulation and flocculation are followed by filtration.
Water systems monitor turbidity and in some cases, particle count, before and after filtration for
process control purposes. (Current regulations require systems to monitor turbidity after filtration.)
Because of the delay between the time of sampling and the receipt of test results for Cryptosporidium and
Giardia, and the challenge of quantifying low concentrations of these pathogens, measuring protozoa
concentrations is not a practical method for on-line process control. Instead, identifying a relationship
Occurrence and Exposure Assessment
for the LT2ESWTR
5-2
December 2005
-------
between turbidity or particle counts and protozoa is desirable. This is the focus of many studies discussed
in this section.
Dugan et al. (2001) evaluated the ability of conventional treatment to control Cryptosporidium
under varying water quality and treatment conditions, and assessed turbidity, particle counts, and aerobic
endospores as indicators of Cryptosporidium removal. Fourteen runs were conducted on a small pilot
scale plant that had been determined to provide equivalent performance to a larger plant. Under optimal
coagulation conditions, oocyst removal across the sedimentation basin ranged from 0.6 to 1.8 log,
averaging 1.3 log, and removal across the filters ranged from 2.9 to greater than 4.4 log, averaging greater
than 3.7 log. Removal of aerobic spores and reduction in particle counts and turbidity all correlated with
removal of Cryptosporidium by sedimentation, and these parameters were conservative indicators of
Cryptosporidium removal across filtration. Sedimentation performance under optimal conditions related
to raw water quality, with the lowest Cryptosporidium removals observed when raw water turbidity was
low. Suboptimal coagulation conditions (underdosed relative to jar test predictions) significantly reduced
plant performance. Oocyst removal in the sedimentation basin averaged 0.2 log, and removal by filtration
averaged 1.5 log. Under suboptimal coagulation conditions, low sedimentation removals of
Cryptosporidium were observed in four of the five runs, regardless of raw water turbidity.
A comprehensive study of Cryptosporidium and Giardia removal, turbidity, and particle counts
under various treatment conditions was conducted by Patania et al. (1995). These variables were studied
using four pilot plants at multiple study sites (and hence, different raw water qualities) with the objective
of providing practical information on Cryptosporidium and Giardia removal over a wide range of
treatment conditions and water qualities. Cryptosporidium and Giardia removal ranged from 1.4 to 6.2
log, with a median of 4.2 log for all studies. A comparison of two different types of treatment processes
used—conventional treatment (including sedimentation) and direct filtration—demonstrated that
Cryptosporidium removal was 1.4 to 1.8 log higher with conventional treatment than with direct filtration.
When treatment conditions were optimized for turbidity removal at four different sites, Cryptosporidium
removal ranged from 2.7 to 5.9 log, and Giardia removal ranged from 3.4 to 5.1 log.
The investigators noted that the log removal values would have been greater than reported if the
raw water oocyst concentrations had been sufficiently high to allow oocyst detection in the filtered water,
or if the detection method for finished water had a higher recovery efficiency. They also noted that
removal of Cryptosporidium was 0.4 to 0.9 log lower during filter maturation than during filter operation,
and Giardia removal was generally 0.4 to 0.5 log lower during maturation (Patania et al. 1995).
As part of the same study, the authors looked for a correlation between log removal of turbidity
and log removal of protozoa. They did not observe a one-to-one correlation, but noted that maintaining
effluent turbidity levels at 0.1 NTU was required to achieve 5.0-log removal of Cryptosporidium and
Giardia 90 percent of the time, and a filter effluent goal of 0.2 NTU resulted in removals of 4.0 log or
less for Cryptosporidium and Giardia 90 percent of the time. They also checked protozoa removal
against particle removal and found that, again, there was no direct correlation. Particle removal,
recommended as an indicator for Giardia removal (USEPA 1989c), tended to underestimate protozoa
removal under some raw water conditions, such as high protozoa concentrations and low turbidity and
particle counts. The authors felt that the one-to-one correlation between protozoa removal and particle
removal reported by others (Nieminski 1994) was due to the combination of higher turbidity raw waters, a
slightly higher range of particle removal, and lower overall protozoa removal compared to the results of
their study.
Nieminski and Ongerth (1995) evaluated performance in a pilot plant and in a full-scale plant (not
in operation during the time of the study) and examined two treatment modes: direct filtration and
Occurrence and Exposure Assessment
for the LT2ESWTR
5-3
December 2005
-------
conventional treatment. The turbidity of the raw water for the pilot plant typically was 4 NTU, with a
maximum of 23 NTU. The pilot plant achieved filtered water turbidity levels between 0.1 and 0.2 NTU,
and Cryptosporidium removals averaged 3.0 log (range = 1.9 to 4.0 log) for conventional treatment and
3.0 log (range = 1.9 to 3.8 log) for direct filtration, while the respective Giardia removals averaged 3.4
log and 3.3 log (ranges = 2.2 to 3.9 and 2.9 to 4.0 log, respectively). The turbidity of the raw water for
the full-scale plant typically was between 2.5 and 11 NTU, with a peak level of 28 NTU. The full-scale
plant achieved similar filtered water turbidity levels, and Cryptosporidium removal averaged 2.25 log for
conventional treatment and 2.8 log for direct filtration, while the respective Giardia removals averaged
3.3 log for conventional treatment and 3.9 log for direct filtration. The authors reported log removals
only for data with detections in the effluent samples; thus, reported removal efficiency is likely lower than
the actual. The differences in performance noted between direct filtration and conventional treatment in
the full-scale plant were attributed to differences in raw water quality during the respective filter runs.
A pilot plant study by Patania et al. (1999) evaluated removal of Cryptosporidium at varied raw
and filter effluent turbidity levels using direct filtration. Raw water turbidity was less than 2 NTU (low)
and 10 NTU (high). Targeted filtered water turbidity was 0.02 NTU at both low and high raw water
turbidity and 0.05 NTU for one run at the low raw water turbidity. Cryptosporidium removal was slightly
higher when the raw water turbidity was higher at equivalent levels of filtered water turbidity. Also,
Cryptosporidium removal was a mean of 1.5 log higher when steady-state filtered water turbidity was
0.02 NTU compared to 0.05 NTU.
LeChevallier and Norton (1992) evaluated protozoa removal at raw water turbidity levels ranging
from less than 1 to 120 NTU. Removals of Giardia (2.2 to 2.8 log) and Cryptosporidium (2.3 to 2.5 log)
were slightly less than those reported by other researchers, possibly because the full-scale plants studied
were operated under less ideal conditions than the pilot plants. The participating treatment plants were in
varying stages of treatment optimization.
West et al. (1994) used pilot-scale direct filtration with anthracite monomedia at filtration rates of
6 and 14 gallons per minute per square foot (gpm/tt2). Raw water turbidity was 0.3 to 0.7 NTU. Removal
efficiencies for Cryptosporidium at both filtration rates were 2 log during filter ripening (despite filter
water turbidity exceeding 0.2 NTU) and 2 to 3 log for the stable filter run, declining significantly during
particle breakthrough. When effluent turbidity was less than 0.1 NTU, removal typically exceeded 2 log.
Log removal of Cryptosporidium generally exceeded that of particle removal.
Nieminski and Bellamy (2000) studied full-scale treatment plants to determine whether water
quality parameters could be used as indicators of protozoa and virus removal. These parameters included
turbidity, particle count, bacteria, bacterial spores, and bacterial phages (viruses that infect bacteria).
Spores are the bacterial equivalent of cysts; they are formed in conditions of environmental stress to
protect the contents of a cell. In order for a parameter to be a good surrogate, it had to be detectable in
raw and filtered water. With unseeded raw water, however, the initial levels were relatively low, and
most parameters were undetectable in a large percentage of finished water samples. Nieminski and
Bellamy found spores of aerobic bacteria to be good indicators of treatment effectiveness, since they were
detectable in 85 percent of finished water samples. However, aerobic spore removal levels did not
correlate with protozoa removal. The authors also examined turbidity removal along with protozoa
removal and found that less turbid raw waters, where the concentration of protozoa was low, had slightly
lower protozoa removals than more turbid raw waters.
In a study of particle counts and their relationship to Cryptosporidium and Giardia removal at
100 plants, McTigue et al. (1998) found that particle count data were normally distributed. The median
log removal of particles larger than 2 micrometers ((am) after filtration was 2.8 log; removal depended to
Occurrence and Exposure Assessment
for the LT2ESWTR
5-4
December 2005
-------
some extent on the raw water particle concentration, There was not a significant correlation between
particle count and turbidity in raw or finished water, although correlation was better in raw water. There
was little correlation between particle counts and protozoa levels in finished water. In an analysis of
previously collected data, the authors reported that median log removal for particles larger than 3 um was
2.4 log; removal also correlated with raw water particle concentration. The authors also conducted pilot
plant studies, in water seeded with protozoa, and found that Cryptosporidium and Giardia removal after
filtration through dual media averaged 4 log and did not vary with initial concentration. They evaluated
pathogen removal under stressed filter conditions (the filtration rate was doubled) and found both
Cryptosporidium and Giardia removal decreased by approximately 2 log. When they compared pilot
plant particle removal to log removal for protozoa, there appeared to be no correlation, but when they
compared a subset of these data taken at similar temperatures and coagulation conditions, they determined
that particle and protozoa removal correlated significantly.
Ongerth and Pecoraro (1995) studied the effect of coagulation on protozoa removal by direct
filtration with low turbidity raw waters (0.35 to 0.58 NTU). With optimal coagulation, effluent turbidity
averaged less than 0.1 NTU, Cryptosporidium removal ranged from 2.7 to 3.1 log, and Giardia removal
ranged from 3.05 to 3.6 log. For one filter run, the coagulant dose was decreased by 50 percent to test
suboptimal operation conditions. Effluent turbidity increased to a mean of 0.36 NTU and
Cryptosporidium and Giardia removals decreased to 1.5 and 1.3 log, respectively.
LeChevallier et al. (1991) evaluated removal efficiencies for Giardia and Cryptosporidium at 66
surface water treatment plants in 14 States and one Canadian province. Most of the utilities achieved
between 2 and 2.5 log removals for both Giardia and Cryptosporidium. When no cysts or oocysts were
detected in the finished water, protozoan levels were set at the "detection limit" for calculating removal
efficiencies.
5.1.2 Other Filtration Technologies
5.1.2.1 Slow Sand
Slow sand filtration plants are commonly associated with smaller systems. They operate at very
low filtration rates without the use of coagulation in pretreatment and have a compacted media of smaller
grain size sand. Most of the particulate removal occurs in a thin layer on top of the sand bed called the
schmutzdecke layer. This layer not only traps particles but also provides biological treatment. EPA
(1988) lists research studies indicating that a well-designed and operated plant using these technologies is
capable of removing 3 log of Giardia and viruses. Results from more recent studies are presented below.
Schuler and Ghosh (1991) studied the removal of Cryptosporidium and Giardia by slow sand
filtration with a pilot filter. Cryptosporidium removals ranged from 3.9 to 6.1 log, with a mean of 5.4 log.
Giardia removals ranged from 2.8 to greater than 6.5 log, with a mean of 5.2 log. Lower Giardia
removals occurred during the winter months and before the filter had matured (analysis of the top layer
showed no biological activity). The I-year experiment began in January 1988; after February 1988 the
Giardia removals were all greater than 5.5 log. The sample collection for Cryptosporidium started in
March.
Timms et al. (1995) conducted a pilot-scale slow sand filtration study of Cryptosporidium
removal. No oocysts were detected in the filtered samples, indicating greater than 3.6 log removal of
Cryptosporidium.
Occurrence and Exposure Assessment
for the LT2ESWTR
5-5
December 2005
-------
Fogel et al. (1993) evaluated the removal efficiencies for Cryptosporidium and Giardia with a
full-scale slow sand filtration plant. The log removals for Cryptosporidium ranged from 0.1 to 0.5 log
and 0.9 to 1.4 log for Giardia. Raw water turbidity ranged from 1.3 to 1.6 NTU and decreased to 0.35
and 0.31 NTU respectively, after filtration. The authors attributed the low Cryptosporidium and Giardia
removals to the relatively poor grade of filter media and lower water temperature. The sand had a higher
uniformity coefficient than recommended, which creates larger pore spaces within the filter bed
subsequently retarding its biological removal capacity. Lower water temperatures (1 degree Celsius
(1 °C)) also decreased biological activity in the filter media.
Hall et al. (1994) researched the removal of Cryptosporidium with a pilot-scale slow sand
filtration plant. Cryptosporidium removals ranged from 2.8 to 4.3 log after filter maturation, with a mean
of 3.8 log (at least 1 week after filter scraping). Raw water turbidity ranged from 3.0 NTU to 7.5 NTU
for three of the four runs and 15.0 NTU for the fourth run. Filtered water turbidity was 0.2 to 0.4 NTU,
except for the fourth run with 2.5 NTU filtered water turbidity. For the fourth run, the Cryptosporidium
removal (3.9 log) did not decrease with the surge in raw and filtered water turbidity.
Hall et al, (1994) also investigated Cryptosporidium removal during start-up of the pilot-scale
slow sand filter. After scraping the sand, the filtration rate was slowly increased by steps of 100 liters per
hour (I/hour) (0.05 m/h) over a 4-day period. The results indicated high Cryptosporidium removals,
ranging from 4.3 log to greater than 6.4 log. From these results, filter ripening did not appear to affect
Cryptosporidium removal, as there was high removals and no consistent increase in removal.
5.1.2.2 Diatomaceous Earth (DE)
DE filters, or precoat filters, use a thin layer of very fine material (usually diatomaceous earth) as
a filter medium. During the filter cycle, additional filter media is metered into the influent water in
proportion to the solids being removed. The dirt particles intermingle with the additional filter media,
which maintains the permeability of the cake layer and allows for longer filter runs. DE filters can handle
only lower flow rates and therefore are used by smaller systems.
Schuler and Ghosh (1990) also investigated the removal of Cryptosporidium and Giardia with a
pilot-scale DE filter. Cryptosporidium removals ranged from 4.6 to greater than 6.4 log, with a mean of
5.3 log. The authors noted that finer grades of DE did not enhance Cryptosporidium removal and
suggested that mechanical straining may not be a significant mechanism of removal. Giardia cysts were
detected in the effluent from only one of five filter runs where a malfunctioning valve allowed air to enter
and the filter cake was cracked, although Giardia removals for this run were still relatively high at 4.3
and 4.6 log. Overall, Giardia removal ranged from 4.3 to greater than 6.7 log and averaged 5.5 log.
Ongerth and Hutton (2001) investigated Cryptosporidium removal capabilities of DE filtration
using a pilot-scale DE filter. Under normal operating conditions Cryptosporidium removals ranged from
6.1 to 6.6 log and effluent turbidity for most runs were 0.10 NTU or less. The authors also investigated
filter performance under undesirable operating conditions. They used an undamped peristaltic feed pump
that caused pressure fluctuations in the influent. (The authors noted this operating condition was unlikely
to occur in a full-scale plant). Cryptosporidium removals dropped by roughly 1 log, ranging from 5.0 to
5.8 log compared to normal operating conditions; however, effluent turbidity increased to a range of 0.14
to 0.40 NTU.
Occurrence and Exposure Assessment
for the LT2F.SWTR
5-6
December 2005
-------
5.1.2.3 Membranes
Membrane technology can achieve high levels of protozoa removal. The membranes commonly
used in municipal-scale plants are composed of cellulose acetate or various types of polymers and have a
spiral-wound or hollow-fiber structure. The process operates by forcing water through a membrane at
low to moderate pressure. Spiral-wound membranes are composed of two flat sheets rolled up like a
carpet; water passes through them into a hollow tube in the center of the roll. Hollow-fiber membranes
are bundles of aligned fibers; water flows from the outside of the bundles to a hollow core in the center
(Conlon 1990). Membrane technologies are categorized into three types by the membrane pore size:
microfiltration (largest pore size), ultrafiltration (medium pore size), and nanofiltration (smallest pore
size).
Microfiltration is becoming more prevalent in the water treatment industry due to several factors,
including increased regulation of disinfection byproducts (DBFs) and microbial contaminants. If raw
water quality is high, microfiltration can be applied without pretreatment. Other advantages of
microfiltration over conventional filtration are that less space is needed, less time is required for filter
runs, less operator expertise and labor are needed, less sludge is generated, and the cost of replacing
membranes is comparable to that of buying chemicals for conventional treatment (de Haas 1997).
A few studies relate the experiences of public water systems (PWSs) that have installed
microfiltration. A small PWS installed a pilot microfiltration unit and found that it produced turbidity
levels of 0.06 NTU, which was comparable with the levels achieved by the system's conventional
treatment plant (de Haas 1997). Another system built a small microfiltration plant for seasonal use and
found that, even with occasional raw water turbidities of 100 NTU, finished water turbidities of 0.05 NTU
could be obtained, while the system's direct filtration plant had to be shut down at 30 NTU (Gere 1997).
One significant cost associated with membrane filtration, especially at small systems, is electricity,
particularly that needed to provide pressure to force water through the membranes and to backwash the
filters (Gere 1997).
The primary disadvantage of membrane filtration is the potential for membrane fouling (clogging
of membranes by particulates and precipitates), especially since a common pore size for microfiltration is
0.2 urn. For this reason, microfiltration and ultrafiltration should be used on pretreated water or high-
quality raw water. Additionally, membranes should be chemically cleaned in place every few weeks
(Gere 1997). Some water systems are beginning to use microfiltration in combination with nanofiltration
to remove particulates and DBF precursors. The first system to install such an integrated membrane
system, a small system in Alaska, achieves mean finished water turbidities of 0.05 NTU (Lozier et al.
1997).
Membrane processes such as microfiltration have achieved greater than 4.8 log removal of both
Cryptosporidium and Giardia under bench-scale worst-case operating conditions and 6 to 7 log removal
under normal pilot-plant operating conditions (Jacangelo et al. 1995) (Exhibit 5.1). These removals are
much greater than the log removals observed by other filtration technologies such as slow-sand and DE
filtration. Seyde et al. (1998) reported 5- to 6-log removals of Cryptosporidium and Giardia using
nanofiltration.
5.1.2.4 Bag and Cartridge Filtration
Bag and cartridge filters treat lower flow rates and thus are mostly used by smaller systems. The
water is treated by passing it through a bag or porous cartridge that retains the paniculate matter. The
nature of the filter material and the direction of flow are two features that differentiate bag filtration from
Occurrence and Exposure Assessment
for the LT2ESWTR
5-7
December 2005
-------
cartridge filtration. Bag filters are typically felt or woven with materials such as polypropylene,
polyester, or nylon. Cartridge filters are typically composed of fiberglass or ceramic membranes
supported by a rigid core or are made from strings of polypropylene, acrylics, nylon, or cotton wrapped
around a filter element. The direction of flow in bag filtration is through the inside of the bag to the
outside, whereas for cartridge filtration systems the direction of flow is outside to inside.
Schaub et al. (1993) evaluated the removal of Cryptosporidium with two cartridge filters, one a
melt blown polypropylene depth filter and the second a polypropylene pleated filter (both 3-micron pore
size). Both filters were tested with "general" water and "worst case" water. For both types of influent
water and flow rates of 1 gallon per minute (gpm) and 2 gpm, there were no detects of Cryptosporidium
oocysts in the filtered water from the depth filter. Results indicated greater than 3.5 log-removal of
Cryptosporidium and no impact of increased flow on removal efficiency. The pleated filter demonstrated
lower removals, with Cryptosporidium removals ranging from 1.1 to 2.1 log and a mean of 1.6 log. The
worse case influent caused the lowest Cryptosporidium removal of 1.1 log.
Goodrich and Lykins (1995) evaluated three different bag filters and one cartridge filter for the
removal of Cryptosporidium surrogates, 4.5-micron polystyrene beads. The bag filters were tested at 25,
50, and 100 percent of their maximum flow rate (not indicated), at an inlet pressure of 50 pounds per
square inch (psi), and at various pressure drops (from 0 to 25 psi). The mean removal of the polystyrene
beads for the three bag filters were 0.5, 1.3, and 2.0 log. Changing the flow rate into the bag filters did
not make a statistically significant difference in removal rates. The cartridge filter was tested at 25 gpm
(50 percent of maximum flow rating) at an inlet pressure of 50 psi. The mean log removal of two runs
through the cartridge filter was 3.6 log.
Roessler (1998) tested the ability of a cartridge filter to remove Cryptosporidium. The test
consisted of three runs seeded with Cryptosporidium. The flow through the filter was constant at 2.4 gpm
for approximately 2 hours for each run. The cartridge filter provided 3.2 to 3.6 log removal of
Cryptosporidium.
5.1.3 Prefiltration Optimization and Filtration Characteristics
To achieve good removal efficiency, it is advantageous to optimize the water treatment process
prior to filtration. Particulate removal prior to the filtration increases the length of the filtration cycle,
reducing the need for backwashing, thereby reducing process fluctuations. Forming the oocyst/particulate
complex by flocculation/coagulation is also a key element of filtration optimization. Several studies
investigating prefiltration processes are reviewed in this section.
5.1.3.1 Coagulation Effects
In coagulation, chemicals are added to water to cause particles to coalesce and make the particles
easier to remove. Yates et al. (1997) conducted pilot-scale studies to optimize coagulation/filtration
processes for the removal of Cryptosporidium oocysts during direct filtration. Either liquid aluminum
sulfate (alum) or ferric chloride (FeCl3) was used as the primary coagulant, in combination with cationic,
anionic, and/or nonionic polymers, to arrive at the optimal coagulation conditions for turbidity and
particle removal. Each coagulant/polymer combination was evaluated only after stable, consistent filter
operation had been demonstrated, usually after two or three complete filtration cycles. In the pilot-scale
testing, FeCl3-treated water generally provided slightly greater removal of turbidity, particles, and aerobic
spores than alum-treated water; the filter runs were 32 hours, as opposed to 21 hours for alum-treated
Occurrence and Exposure Assessment
for the LT2ESWTR
5-8
December 2005
-------
water. Coagulation with FeCl} provided greater Cryptosporidium oocyst removal than alum at ambient
pH. Oocyst removals of 4.5 log were found during filter challenges when FeCl3 was used in conjunction
with prechlorination and addition of filter aid (polyDADMAC). Similar conditions using alum yielded
about a 3.7-log removal (Yates et al. 1997).
Huck et al. (2000) studied filtration efficiency during optimal and sub-optimal coagulation
conditions at two direct filtration pilot-scale plants. One plant employed a high coagulation dose (38
milligrams per liter (mg/L) alum and 2 mg/L silicon dioxide (SiO2)) for both total organic carbon (TOC)
and particle removal, and the second plant used a low dose (5 mg/L alum) for particle removal only. Both
plants were operated to maintain turbidity below 0.1 NTU during optimal conditions. The high
coagulation dose plant achieved a median (mean was not reported) of 5.6 log removal of
Cryptosporidium, while the low dose plant achieved a median of 3 log. The authors noted that the
difference between the two plants' results was not fully understood but may be due to the difference in
coagulation doses. When sub-optimal conditions were induced by decreasing the coagulation dose 40 to
60 percent for a target turbidity of 0.2 to 0.3 NTU, the median Cryptosporidium removals dropped to 3.2
log and 1 log at the high dose and low dose plants, respectively. However, the high coagulation dose
plant showed a much wider range of Cryptosporidium removals, with the 25th and 75th percentiles
approximately 1.7 log and 5.2 log, respectively. The authors also tested conditions of total coagulation
failure, and results indicated minimal Cryptosporidium removal; from two runs the median was
approximately 0.25 log.
5.1.3.2 Filter Breakthrough Effects
A 1996 pilot-plant study by Hall and Croll assessed aspects of rapid gravity filter operations that
can influence the risk of breakthrough (in which particles and microbes pass through filters) into filtered
water. They observed quality changes that can occur during a filter run. Initial peaks in particle counts,
turbidity, and oocyst concentrations (mean = 6.3 oocysts/liter (L)) occurred in the first hour of the filter
run. Stopping and restarting after backwashing the filters also produced peaks in particle counts and
turbidity, but these were less significant than the peaks at the beginning. The initial peaks were a
consistent feature of all filter runs monitored at the plant and were attributed to backwashing and filter
ripening. Hall's and Croll's observations demonstrated that higher oocyst concentrations are detected in
finished water during the first hour of a filter run, consistent with turbidity and particle count data,
indicating that breakthrough of particles gives a good indication of potential exposure to Cryptosporidium
(Hall and Croll 1996).
Emelko et al. (2000) conducted further investigations of Cryptosporidium removals during
vulnerable filtration periods using a pilot-scale direct filtration system. The authors investigated four
different operational conditions: stable; early breakthrough—when filter effluent turbidities were
approximately between 0.1 and 0.3 NTU; late breakthrough—when filter effluent turbidity reached 0.3
NTU; and end of run—from when subtle changes in effluent turbidity and particle counts were noticed to
approximately 0.1 NTU. The authors presented their results for stable operation in combination with data
from previous work they conducted at the same facility (Coffey et al. 1999). Cryptosporidium removals
during stable operation ranged from 4.7 to 5.8 log, with a mean of 5.5 log. At these times, turbidity was
consistently low, approximately 0.04 NTU. The early breakthrough period started with effluent
turbidities from 0.04 to 0.08 NTU and increased to approximately 0.2 NTU. Cryptosporidium removals
diminished during this period; results from three experiments ranged from 1.7 to 2.8 log removal, with a
mean of 2.1 log. For the late breakthrough period, turbidities were consistently 0.25 to 0.3 NTU at the
start of seeding periods. (The authors provided ending effluent turbidity data for one of two
experiments—0.35 NTU where the initial effluent turbidity was 0.25 NTU). Cryptosporidium removals
Occurrence and Exposure Assessment
for the LT2ESWTR
5-9
December 2005
-------
further decreased to a mean of 1.4 log and a range of 1.3 to 1.8 log. Two experiments tested
Cryptosporidium removal during the end-of-run operation, when effluent turbidities generally start
increasing. Turbidity started at about 0.04 NTU for both experiments and ended at 0.06 NTU for the first
experiment and 0.13 NTU for the second experiment. The authors reported a range of 1.8 to 3.3 log and a
mean of 2.5 log Cryptosporidium removal for both experiments. During the one-hour seeding period of
the first experiment, turbidity increased steadily from 0.04 NTU to 0.06 NTU while Cryptosporidium
removal decreased from 3.3 to 2.3 log.
Emelko et al. (1999) studied design and operational conditions for maximizing Cryptosporidium
removal by filters during both optimal and challenged operating conditions. A bench-scale experiment
was conducted using a dual media filter (anthracite and sand). Water containing a suspension of kaolinite
was coagulated in-line at pH 6.9 with 5 mg/L alum and filtered.
After an hour of normal operation the coagulant addition was terminated, resulting in elevated
effluent turbidities and decreased removal of Cryptosporidium. Turbidity increased from 0.08 NTU to
0.15 NTU while corresponding Cryptosporidium removals decreased from 4.2 log to 3,3 log.
5.1.3.3 Sedimentation and Dissolved Air Flotation
One disadvantage of the flocculation/coagulation and sedimentation process is the buoyancy of
oocysts and their low settling velocity, which contribute to incomplete removal (Gregory 1994;
Swabby-Cahill et al. 1996). Dissolved air filtration (DAF) takes advantage of these properties by floating
the oocyst/particle complex to the surface for removal. In DAF, air is dissolved in pressurized water,
which is then released into a flotation tank containing the flocculated particles. As the water enters the
tank, the dissolved air forms small bubbles, which collide with and attach to floe particles and float to the
surface. The smaller the air bubbles, the more bubbles can form, increasing the chances of collision with
floe (Gregory and Zabel 1990).
Plummer et al. (1995) investigated the effectiveness of DAF for the removal of Cryptosporidium
parvum oocysts, comparing removal levels to those achieved by conventional treatment with
sedimentation. As shown in Exhibit 5.2, comparison of the prefiltration steps of the conventional
treatment process showed only 0 to 0.81 log removal of oocysts by sedimentation and 0.38 to 3.7 log
removal by DAF processes, depending on coagulant dose (Plummer et al.,1995).
Edzwald and Kelley (1998) demonstrated a 3 log removal of oocysts using DAF, compared with
a 1 log removal using sedimentation in the clarification process before filtration. Braghetta et al. (1997)
did not evaluate oocyst removal, but they showed that DAF lengthened microfiltration filter runs and
decreased turbidity from mean raw water levels of 2.7 NTU to 0.3 to 1.0 NTU.
Harrington et al. (2001) studied the removal of Cryptosporidium and emerging pathogens by
filtration, sedimentation, and DAF, using bench-scale jar tests and pilot-scale conventional treatment
trains. In the bench-scale experiments, mean log removal of Cryptosporidium was 1.2 by sedimentation
and 1.7 by DAF, with all experiments run at optimized coagulant doses. Cryptosporidium removal was
not significantly affected by lower pH or coagulant aid addition, and was similar in all four water sources
that were evaluated. However, removal of Cryptosporidium was greater at 22°C than at 5°C, and was
observed to be higher with alum coagulant than with either polyaluminum hydroxychlorosulfate or ferric
chloride. In the pilot scale experiments, water treated with coagulation, flocculation, sedimentation, and
granular media filtration had a mean log removal of Cryptosporidium of 1.9 in filtered water with
turbidity of 0.2 NTU or less. Tenth percentile removal increased as filtered water turbidity dropped
Occurrence and Exposure Assessment
for the LT2ESWTR
5-10
December 2005
-------
below 0.3 NTU. The authors noted there was no apparent effect of filtration rate on removal efficiency.
In comparing Cryptosporidium removal by sand, dual media (anthracite/sand), and trimedia
(anthracite/sand/garnet) filters, no difference was observed near neutral pH in samples collected from
ripened filters. However, at pH 5.7, removal increased significantly in the sand filter, and it outperformed
the other filter media configurations. There was no observable effect of a turbidity spike on
Cryptosporidium removal.
Exhibit 5.2 Data Comparing Sedimentation and Dissolved Air Flotation Removal
of Cryptosporidium
Raw Water Oocyst
Concentration
(no./L)
3.5 x 105
3.5 x 105
3.5 * 105
3.5 x 105
3.5 x105
Ferric Chloride
Concentration
(mg/L)
2
3
3.5
4
5
Removal by Dissolved
Air Flotation (log units)
0.38
2
2.6
NR
3.7
Removal by Sedimentation
(log units)
0
NR
0.61
0.81
NR
Source: Plummeretal. 1995.
NR = No marked reduction.
As part of its rule development process, EPA is studying other prefiltration steps systems can use
to make filtration more effective (USEPA 2000f). One possible step is installation of pre-settling basins
to allow additional settling before or while coagulant is added. Storage has the added advantage of
providing time for oocysts and cysts to die off. Systems can also implement additional filtration before
and after their usual filtration processes. These include the use of in-bank filtration (passing water
through an infiltration gallery or shoreline before withdrawing water), roughing filters (preceding primary
filtration), and secondary filtration (including granular activated carbon filters after conventional
filtration).
5.1.3.4 Solids Contact Clarifiers
Solids contact clarifiers, also referred to as sludge blanket clarifiers, are processes that incorporate
coagulation, flocculation, and sedimentation into one unit process. These processes can provide efficient
clarification by recirculating flocculated water, or mixing raw water and coagulants with flocculated
water, which enhances chemical precipitation and larger floe formation. Additionally, the flocculated
water generally passes directly through the sludge blanket to allow adsorption of the floes by the sludge
blanket. A pilot-scale study of a pulsating solids contact clarifier by New Jersey American Water
Company showed over 4-log removal ofGiardia cysts (Pennsylvania American Water Company 1989).
The pulsating design allows the sludge blanket to be uniformly expanded over the entire surface area,
which minimizes short circuits.
Occurrence and Exposure Assessment
for the LT2ESWTR
5-1!
December 2005
-------
5.2 Removal of Viruses
This section discusses the factors influencing virus removal in filtered systems. Viruses that are
adsorbed to particulate matter may settle out during sedimentation (USEPA 1999a). Sedimentation can
remove about 95 percent of poliovirus and coxsackievirus (USEPA 1999b). However, viruses removed
as a result of the sedimentation process have not been inactivated (SDWC 1977). They are still capable
of causing infection if the settled sludge is improperly disinfected. If untreated sludge is disturbed,
viruses can be redistributed into the overlying water (USEPA 1999a).
Removal of viruses in treatment plant filters is highly variable and depends on the filter design
and operation, as well as the type of pretreatment provided. Virus particles that are not bound to
particulates are usually too small to be retained on sand filters or alternate media used to filter water.
Virus retention by such media depends on association of the virus with suspended matter that is large
enough to be trapped mechanically. This is because the sand has no affinity for the virus and the virus
particles are small enough to pass through the filter pores (Montgomery 1985). When sand filtration
follows coagulation and sedimentation, viruses that are sorbed on fine floe particles can be retained
efficiently by sand. Under these conditions, virus removal by the filter media ranges from 90 percent to
more than 99 percent (SDWC 1977; USEPA 1999a). Payment and Armon (1989) evaluated the
percentage removal of indigenous viruses at operational water treatment plants for a combined
fiocculation and settling step. Results ranged from 0 to 74 percent removal of rotavirus, enterovirus, and
bacteriophage during the dry and rainy seasons.
Adsorption is the phenomenon whereby molecules adhere to a surface, such as granular activated
carbon (GAC) or resins, with which they come into contact because offerees of attraction at the surface.
The degree of adsorption of a virus is pH-dependent, with stronger retention occurring at lower pH
(SDWC 1977). Results from bench- and pilot-scale studies performed in 1977 suggest that virus removal
onto GAC is inconsistent and hard to control (Montgomery 1985; USEPA 1999a).
More recently, Jancangelo et al. (1995) studied the effectiveness of micro- and ultrafiltration on
MS2 bacteriophage viruses. They tested seeded water from three different sources at bench and pilot
scales to determine possible mechanisms of action for membrane filtration. At bench scale, with the use
of distilled water and worst-case scenario operating conditions, microfiltration provided about 1-log
removal of MS2. At pilot scale using raw water, microfiltration consistently achieved 2 or slightly above
2 log of removal for some waters and membranes, but less than 1 log for other sources, suggesting that
other mechanisms besides physical straining or adsorption to the membrane are at work. Varying the
initial virus concentration had no impact. The authors also tested the impact of adding a clay suspension
to the water at bench scale before passing it through a microfilter. Removal was not significantly
impacted, suggesting that adsorption (at least to clay suspensions) is not a mechanism. However,
precoating the microfilter with clay before filtration, causing the formation of a cake layer, increased log
removal from about 1.2 to 3.7, depending on the amount of clay applied. In natural waters at pilot scale,
natural cake layer formation had less of an effect, increasing removal by only 0.5 log after 5 hours of
operation without back washing, probably due to the much lower density of the layer. Removal returned
to initial levels after backwashing. The authors also found that chemical precipitation, which decreased
the rate of flow through the membrane, increased virus removal. The composition of the precipitate was
not determined. Ultrafiltration was found to be more effective than microfiltration, exhibiting log
removals at or above 6 at both bench-and pilot-scale plants.
Adham et al. (1998) conducted a bench-scale study of virus removal via reverse osmosis. They
tested five brands of membranes on water seeded with 10" to 107 plaque-forming units per milliliter
(pfu/mL) of MS2 viruses. Reverse osmosis membranes have nominal cutoffs of 0.001 urn; they block
Occurrence and Exposure Assessment
for the LT2ESWTR
5-12
December 2005
-------
most particles larger than 0.001 urn in diameter, but variations may occur depending on particle shape
and charge. MS2 viruses are 0.025 \im, but some membranes still allowed viruses to pass through. The
membranes achieved removes from 1.4 to >7.4 log.
5.3 Conclusion
The studies described above indicate that conventional and direct filtration systems, when
operating under appropriate coagulation conditions and achieving a filtered water turbidity level of less
than 0.3 NTU, should achieve at least 2 log of Cryptosporidium removal. Removal rates vary widely, to
over 5 log, depending on water matrix conditions, filtered water turbidity levels, and stage of the filtration
cycle.
According to the literature reviewed in this chapter, the highest pathogen removal rates occurred
in pilot plants and systems that achieved finished water turbidities of less than 0.1 NTU. Studies of
Cryptosporidium and Giardia removal at times of filter breakthrough and end-of-run show that removal
can decline before a significant change in turbidity is detected. Although turbidity is an indicator of
filtration performance, it is not a direct indicator of pathogen removal.
DAF may be an effective prefiltration alternative to sedimentation. In two experiments,
sedimentation resulted in less than 1 log removal of Cryptosporidium, while DAF resulted in up to 3 or
more log removal. DAF also has been shown to reduce turbidity.
Virus removal depends heavily on whether adsorption takes place. Viruses adsorb to floes and
either settle or filter out in the treatment process. Other technologies (e.g., slow sand, DE filtration,
membranes) can provide equivalent or better pathogen removal than conventional filtration. Slow sand,
DE filtration, and bag and cartridge filtration are more suited for smaller systems. Membrane
technologies can provide excellent pathogen removal.
Occurrence and Exposure Assessment
for the LT2ESWTR
5-13
December 2005
-------
6. Observed Finished Water Occurrence
The ICR required finished water sampling by PWSs that detected certain levels of
microorganisms in their source water. PWSs were required to conduct finished water monitoring at any
treatment plant at which they detected 10 or more Giardia cysts, 10 or more Cryptosporidium oocysts, or
1 or more total culturable viruses, per liter of any microbial raw water sample (USEPA 1996a). The PWS
was required to analyze finished water samples for all organisms, not just those detected in the source
water. For each sample of finished water, the PWS reported the concentrations of Cryptosporidium,
Giardia, total culturable viruses, and other indicators (total coliforms, fecal coliforms, and/or E. colt).
The rule did allow some exceptions to the monitoring requirements. This section presents the finished
water microbial occurrence data. Sections 6.1 through 6.4 discuss the Cryptosporidium, Giardia, virus,
and indicator data, respectively.
Along with the systems that reached the trigger levels, other PWSs voluntarily tested their
finished water. But because not every system was required to test its finished water, the total number of
finished water samples is far less than the number of source water samples.
Exhibit 6.1 presents a summary of sample volumes and volumes analyzed for the two groups of
microorganisms of concern: protozoa and viruses. Of the microorganisms tested for in the finished water
analysis, protozoa—which include Cryptosporidium and Giardia—were sampled almost twice as many
times as viruses. Coliforms were also sampled more frequently than viruses. Of water sampled each
time, only a portion was tested; this was the volume analyzed. The range of the volume analyzed,
however, was not significantly different for the two groups. The volume analyzed for viruses was 35
percent lower than the sample volume. For protozoa, the volume analyzed was significantly different
from the sample volume; the volume analyzed was 22 percent of the sample volume.
Exhibit 6.1 Summary of Volumes Sampled and Analyzed
Microorganism
Protozoa
Viruses
Number of
Samples (N)
1,058
573
Sample Volume (L)
Median
1.015
1,518
Mean
1,072
1,550
Range
100-8,509
200 - 2,998
Volume Analyzed (L)
Median
110
1,000
Mean
228
995
Range
1 -1,855
1 -1,489
6.1 Cryptosporidium
Exhibit 6.2 summarizes the results of sampling for various microorganisms. During the 18
months in which ICR data were collected, 1,058 samples of finished water were collected to test for
Cryptosporidium oocysts. The volumes collected to analyze for Cryptosporidium were all close to
1,000 liters (L) (6 percent of samples were 1,200 L or greater). Of the total sample, 11 samples (from 7
plants) had positive detects. These positive detections were 1.04 percent of the total number of samples
taken. The average and median concentrations for all 11 positive detections were less than 0.1 oocyst/L.
The range of positive concentrations for the Cryptosporidium samples was much smaller than that for any
of the other microorganisms, except for viruses.
Occurrence and Exposure Assessment
for the LT2£S(yTR
6-1
December 2005
-------
Exhibit 6.2 Summary of Sample Results
Microorganism
Cryptosporidium
Giardia
Viruses
Total coliform
Fecal coliform
£. co//
Number of
Samples
1,058
1,058
573
3,452
2,208
2,102
Number of
Detects (Percent)
11 (1.04)
16 (1.51)
9 (1.57)
24 (0.70)
12 (0.54)
5 (0.24)
Concentration1
Mean2
0.0057
0.033
0.0011
22.83
2.42
1
Median2
0.0049
0.012
0.001
1.5
1
1
Range of Positive
Concentrations
0.001 - 0.01 1
0.002 - 0.26
0.001 - 0.002
1-330
1-13
1-1
Concentrations are in units of number of oocysts/L for Cryptosporidium, cysts/L for Giardia, most probable
number (MPN) of plaque-forming units/L for viruses, and density/100 milliliters (ml) for bacteria.
2 Mean and median for positive samples.
One plant detected Cryptosporidium four times in its finished water; another detected it twice.
The other plants with positive detects found Cryptosporidium only once. All of the plants with positive
detects use conventional filtration and chlorine as primary and secondary disinfectants, except for one that
used chloramine as a secondary disinfectant. One plant was found to have oocysts with internal structures
still intact. The majority of positive samples, however, had oocysts that contained no internal structures
and instead were either amorphous or empty (see Exhibit 6.3).
Exhibit 6.3 Summary of Crypfosporfcff'um-Positive Sample Results
Volume
Analyzed (L)
202
800
180
237
213
409
213
427
454
293
802
Total
Cryptosporidium
Concentration1
0.01
0.001
0.006
0.004
0.0094
0.0049
0.0047
0.007
0.011
0.0034
0.001
Oocyst Count
Total
2
1
1
1
2
2
1
3
5
1
1
Internal
Structures
0
0
0
0
0
0
0
0
3
1
0
Amorphous
1
1
1
0
0
1
0
1
2
0
0
Empty
1
0
0
1
2
1
1
2
0
0
1
1 Oocysts per liter.
Occurrence and Exposure Assessment
for the LT2ESWTR
6-2
December 2005
-------
6.2 Giardia
During the 18 months in which ICR data were collected, 1,058 samples of finished water were
collected and tested for Giardia cysts. Of these, 16 samples tested positive. These positive detects
constitute 1.51 percent of the total number of samples taken. The mean and median concentrations for all
16 positive detects are more distinct from one another than for Cryptosporidium. The range of the
positive concentrations for Giardia is much broader than the range of detects for Cryptosporidium, even
though the same total number of samples were analyzed (see Exhibit 6.2).
The 16 Gianfla-positive samples were generated by 6 water systems. Five of the plants with
positive samples also had positive samples for Cryptosporidiiim. All of the plants with positive samples
used conventional filtration and chlorine as primary and secondary disinfectants. The volumes collected
to analyze for Giardia were all close to 1,000 L, but only a small volume of the 1,000 L of water
collected was analyzed. The number of detects per liter of water ranged from 0.002 to 0.26, which is a
considerably broader range than for Cryptosporidium and for viruses. Counts for Giardia were generally
higher than for Cryptosporidium. This reflects the fact that Giardia concentrations are higher than
Cryptosporidium concentrations in source water (see Chapter 4). Three Giardia samples had cysts with
one internal structure, and three had cysts with more than one internal structure (see Exhibit 6.4).
Occurrence and Exposure Assessment
for the LT2ESWTR
6-3
December 2005
-------
Exhibit 6.4 Summary of G/ard/a-Positive Sample Results
Volume
Analyzed (L)
202
172
202
237
212
102
213
213
427
205
454
293
34
120
210
802
Total Giardia
Concentration1
0.069
0.017
0.01
0.004
0.0094
0.02
0.047
0.0094
0.0023
0.0097
0.026
0.0068
0.26
0.017
0.014
0.002
Cyst Count
Total
14
3
2
1
2
2
10
2
1
2
12
2
9
2
3
2
One
Internal
Structure
0
0
0
0
0
0
1
0
0
0
6
0
1
0
0
0
Two or More
Internal
Structures
0
0
0
0
0
0
0
0
0
0
3
0
2
0
1
0
Amorphous
0
0
1
0
1
2
6
2
1
2
1
1
5
2
2
0
Empty
14
3
1
1
1
0
3
0
0
0
2
1
1
0
0
2
Cysts per liter.
6.3 Viruses
During the 18 months in which ICR data were collected, 573 finished water samples were
collected to test for viruses. Of these, nine were positive; this constitutes about 1.6 percent of the total
number of samples. The average and median concentrations are remarkably similar. Therefore, it can be
assumed that very few of the data are outliers. This assumption is confirmed by the range of positive
concentrations (0.001-0.002 MPN/L), which is much smaller for virus samples than for any of the other
microorganisms (see Exhibit 6.2).
The nine positive sample results were generated by seven plants. The plants with the positive
samples used a variety of different water system treatment and disinfection methods (see Exhibit 6.5).
The sample volumes collected were approximately 1,500 L, and the volumes analyzed were
approximately 1,000 L (see Exhibit 6.1). One of the plants also had positive Cryptosporidium and
Giardia samples.
Occurrence and Exposure Assessment
for the LT2ESWTR
6-4
December 2005
-------
Exhibit 6.5 Summary of Virus Positive Results
MPN virus/1.1
0.001
0,001
0.001
0.001
0.001
0.001
0.002
0.001
0.001
Treatment Type
Direct Filtration
Conventional
Conventional
Conventional
Two Stage Softening
Two Stage Softening
Two Stage Softening
Two Stage Softening
Conventional
Disinfection
Primary
Ozone
Chlorine/Chloramine
Chlorine
Chlorine
Chlorine/Chloramine
Chlorine/Chloramine
Ozone
Chloramine
Chlorine
Secondary
Chlorine
Chloramine
Chlorine
Chlorine
Chloramine
Chloramine
Chloramine
Chloramine
Chloramine
Most Probable Number of viruses per liter.
6.4 Indicators
The ICR required all PWSs to monitor their source water for total coliforms. Operators could,
however, choose between fecal coliform or E. coli; some plants monitored for both. Whenever a PWS
exceeded a trigger concentration and was required to monitor finished water for protozoa and viruses, it
was also required to monitor finished water for total coliform and either fecal coliform or E. coli.
6.4,1 Total Coliform
During the 18 months in which ICR data were collected, a total of 3,452 finished water samples
were collected to be tested for total coliform. Of that total number, 24 had positive detects. These
positive detects account for less than 1 percent of the total number of samples. The mean and median
concentrations are quite different (see Exhibit 6.2) because a few outliers (shown in Exhibit 6.6) skew the
distribution.
Occurrence and Exposure Assessment
for the LT2ESWTR
6-5
December 2005
-------
Exhibit 6.6 Summary of Total Conform Positive Results
Density/100 mi.
1
1
1
1
1
1
1
1
2
3
8
8
11
19
21
90
330
37
1
1
1
1
2
5
Treatment Type
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Direct Filtration
In-Line Filtration
Softening
Softening
Softening
Softening
Softening
Disinfection
Primary
None Listed
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine/Chloramine
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine/Chloramine
Chlorine
Chlorine
Chlorine
Chlorine/Chloramine
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine/Chloramine
Ozone
Chlorine
Ozone
Secondary
None Listed
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
Chloramine
Chloramine
Chlorine
Chlorine
Chloramine
Chloramine
Chlorine
Chlorine
Chloramine
Chloramine
Chlorine
Chlorine
Chlorine
Chlorine
Chloramine
Chlorine
Chlorine
Chloramine
6.4.2 Fecal Coliform
During the 18 months in which ICR data were collected, 2,208 finished water samples were
collected and tested for fecal coliform. Of that number, 12 samples were positive. These positive detects
Occurrence and Exposure Assessment
for the LT2ESWTR
6-6
December 2005
-------
account for less than 1 percent of the total number of samples. The results of the positive samples are
shown below in Exhibit 6.7.
Exhibit 6.7 Summary of Fecal Conform Positive Results
Density/100 mL
1
1
1
1
1
1
1
2
2
4
13
1
Treatment Type
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
Conventional
In-Line Filtration
Disinfection
Primary
None Listed
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine/Chloramine
Chlorine/Chloramine
Chlorine
Chlorine
Chlorine
Chlorine
Secondary
None Listed
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
Chloramine
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
6.4.3 E. coli
During the 18 months in which ICR data were collected, 2,102 finished water samples were
collected and tested for E. coli. Of that number, five samples were positive. These positive detections
account for less than 0.5 percent of the total number of samples. The positive results are shown in Exhibit
6.8.
Occurrence and Exposure Assessment
for the LT2ESWTR
6-7
December 2005
-------
Exhibit 6.8 Summary of E. coli Positive Results
Density/100 mL
1
1
1
1
1
Treatment Type
In-Line Filtration
Conventional
Conventional
Conventional
Conventional
Disinfection Type
Primary
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
Secondary
Chlorine
Chlorine
Chlorine
Chlorine
Chlorine
6.5 Finished Water Occurrence—Summary
Of the plants that detected a particular organism in their raw water, less than 2 percent detected
the same organism in their finished water. The concentrations of Cryptosporidium and Giardia in
finished water were quite low, with Cryptosporidium concentrations never greater than 0.01 oocysts/L
and Giardia concentrations (with one exception) below 0.07 cysts/L. There was a strong correlation
between Cryptosporidium and Giardia, most plants with positive Giardia samples also had positive
Cryptosporidium samples. Several of these plants also had multiple positive samples, which may indicate
frequent process upsets or poorer performance at these plants. Virus detects were also low. Coliform
densities in positive samples were generally at or close to 1 per 100 mL for fecal coliform and E. coli,
while total coliform concentrations for several samples were up to a few hundred times higher. The
overall microbial occurrence was low; however, the limitations of the sampling and analysis techniques
affected the finished water data as well as source water. As a result, the true finished water occurrence
may differ from what the observed data indicate.
Occurrence and Exposure Assessment
for the LT2ESWTR
6-8
December 2005
-------
7. Modeled Pre-LT2ESWTR Occurrence of Cryptosporidium in Finished
Water
This chapter estimates the expected occurrence of Cryptosporidium in finished water of surface
water supplies following implementation of the IESWTR and LT1ESWTR. This expected occurrence
(referred to as the Pre-LT2ESWTR occurrence) is used in additional analyses to support the regulatory
development of the LT2ESWTR. The Economic Analysis for the LT2ESWTR uses this finished water
occurrence estimate as a baseline to predict the treatment cost and health benefits of different regulatory
scenarios.
It is difficult to quantify Cryptosporidium levels in treated drinking water due to the low and
variable concentrations typically present after filtration and the relatively high "detection limit" of
analytical methods for protozoa. Consequently, EPA estimated finished water occurrence levels based on
source water occurrence data and assumptions about the performance of treatment in removing oocysts.
Because EPA assesses risk using mean annual exposure, the specific end-point modeled is the expected
annual average Cryptosporidium concentration in finished water at the plant level.
The modeling of Pre-LT2ES WTR occurrence in finished water of surface water plants was
conducted using two-dimensional Monte Carlo simulation procedures structured to address the variability
and uncertainty in both annual average source water occurrence and treatment effectiveness among plants.
The algorithm for calculating the finished water Cryptosporidium level at an individual water
plant is relatively simple, as shown below:
CsxlO'L
(Equation 7.1)
where CF is the finished water Cryptosporidium concentration, Cs is the source water Cryptosporidium
concentration, and L is the log removal achieved by the treatment performed at the plant.
In the Monte Carlo procedure, Equation 7.1 is computed many times. In each iteration of the
calculation, different values are selected for Cs and L, reflecting (1) the variability in source water mean
occurrence levels and treatment effectiveness expected to occur from one plant to another and (2) the
uncertainty in the characterization of both of these factors. The output of this procedure is a set of
distributions for CF. Each of the individual CF distributions represents one possible depiction of the plant-
to-plant variability of finished water Cryptosporidium levels. The overall set of those distributions
reflects the uncertainty in the "true" plant-to-plant variability. It is important to note that the occurrence
modeling performed here, for use in the LT2 rule Economic Assessment, focuses on endemic risk and
normal operating conditions at treatment plants and not outbreaks due to extraordinary occurrence levels
or breakdowns in the treatment system. It is for this reason that the modeling effort focuses on plant
average concentrations in the sources water and not extreme values that may occur rarely. Therefore, it
was also considered appropriate to use expected values or means for the treatment effectiveness input
value, L, that reflect typical operating conditions at a plant over the long term.
The remainder of this chapter is organized as follows:
• Section 7.1 presents relevant information on the source water Cryptosporidium occurrence
distributions obtained from each of the four primary data sets (ICR Filtered, ICR Unfiltered,
ICRSSM, and ICRSSL).
Occurrence and Exposure Assessment
for the LT2ESWTR
December 2005
-------
Section 7.2 presents information on estimated treatment effectiveness distributions following
full implementation of the IESWTR and LT1ESWTR.
Section 7.3 describes the structure and implementation of the two-dimensional Monte Carlo
simulation model.
Section 7.4 presents the finished water Cryptosporidium occurrence distributions resulting
from this modeling effort.
7.1 Source Water Occurrence
As discussed in Chapter 4, four data sets are used to characterize the occurrence of
Cryptosporidium in source water for the LT2ESWTR regulatory analysis: the ICR filtered systems data
set, the ICR unfiltered systems data set, the ICRSSL data set, and ICRSSM data set. A statistical model
was developed and fit to these four data sets to characterize from each a national distribution of source
water Cryptosporidium concentrations, reflecting both the variability from one plant to another and the
uncertainty in the estimates. This statistical model is discussed in Chapter 3 and also in Appendix B.
In the statistical modeling, the four data sets were treated separately, each serving as an
alternative picture of the "true" source water occurrence distribution. The data sets were not combined
because of differences among the surveys in plant population sampled as well as in sampling and testing
methods. The ICR and 1CRSS, for example, were conducted at different times, over different numbers of
months and numbers of samples per month, and used different laboratory analysis methods. Within the
1CRSS, there were two independent surveys of large and medium plants leading to the ICRSSL and
ICRSSM data sets, respectively, that characterize these two classes of plants. The ICR surveyed only
large plants, but included plants with both filtered and unfiltered source water and is broken into two data
sets accordingly. (The ICRSS included too few unfiltered plants to model separately.)
Among the four data sets, ICR unfiltered systems are modeled differently from the others.
Because unfiltered systems are required to have watershed protection programs that achieve low turbidity
and coliform limits, they generally have high source water quality and are not required to provide
physical treatment (they do provide chemical disinfection that does not reliably inactivate
Cryptosporidium). As a result, their finished water occurrence is the same as their source water
occurrence and no further modeling is required. Thus, the source water occurrence modeling described in
this section applies to all four data sets, but the finished water modeling described in Sections 7.2 through
7.4 applies only to filtered systems data from the ICR, ICRSSM, and ICRSSL.
Using the simplified, or reduced-form, occurrence model described in Appendix B (section B.6),
1,000 pairs of parameter values—log mean and log standard deviation—were generated by the model
based on each data set, each pair defining a national distribution of plant-mean concentrations consistent
with that data set. The resulting collections of 1,000 national distribution curves, one collection from
each data set, served as input to the cost and benefit models developed for the LT2ESWTR Economic
Analysis. Each curve in a set of 1,000 depicts the expected variability in mean source water
Cryptosporidium concentration among plants, where corrections for test method recovery specific to each
data set have been incorporated. The differences from curve to curve, within a collection, represent
uncertainty in the estimated plant-mean distribution.
Occurrence and Exposure Assessment
for the LT2ESWTR 7-2 December 2005
-------
Exhibits 7.1 through 7.5 summarize these four sets of 1,000 distribution curves. A graph with all
1,000 curves would appear as a solid, filled space instead of individual curves and thus provide limited
information. Instead, each set is represented by three percent!les—the 50th, representing the median
curve, and the 5lh and 95th percentiles—comprising a bound that captures all but 10 percent of the values
in the 1,000 curves.
None of these percentile curves, however, is an actual curve from among the 1,000 that make up a
given set. To generate these bounds, each collection of curves is cross-sectioned vertically at regular
intervals. At each evaluated point along the horizontal axis (which shows mean concentration), a single
vertical "slice" captures the associated range of vertical axis values (cumulative percentage of plants)
across all 1,000 curves. From each slice, the following ordered values are saved: 50th, 500th, and 950th.
These saved values represent the 5th, 50th, and 95th percentile at a particular concentration level. To
generate the 95-percent bound, for example, a curve is drawn through the saved 95"1 percentile values,
across the full horizontal range of concentration values.
Exhibit 7.1 shows a comparison of ICR, ICRSSL, and ICRSSM median curves across 1,000
modeled source water occurrence distributions. A comparison of the distributions for the four data sets
shown in Exhibit 7.1 indicate both some overall differences seen in the mean or median results among the
underlying data sets as well as differences in the variability seen from location to location within each of
these data sets. These differences among data sets can be considered as the first level of uncertainty that
was characterized in the occurrence modeling by virtue of modeling these data sets, especially the three
filtered system data sets, separately. A comparison of the curves shows the lower predicted occurrence
for the unfiltered systems. It also shows, among the filtered systems data sets, greater variability in the
ICR-estimated plant-means. Relative to the ICRSSL and ICRSSM filtered curves, the more gradual slope
of the ICR filtered curve results in a higher frequency of both very high and very low plant means.
Exhibits 7.2 through 7.5 capture the second level of uncertainty, specific to each data set, by
adding the upper and lower 90-percent bounds to each of the median distribution curves. These bounds
reflect uncertainty due to sampling and measurement error (not all public water systems were sampled,
small fraction of total source water volume sampled at each participating location, and variable recovery
in the laboratory). While differences among data sets and curves reflect uncertainty, each individual
curve, by itself, captures the estimated variability in mean source water Cryptosporidium concentration
from plant-to-plant.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-3
December 2005
-------
Exhibit 7.1 Comparison of ICR, ICRSSL, and ICRSSM Median Curves Across
1,000 Modeled Source Water Occurrence Distributions
100%
80% -
40%
o 20%-
1e-005 0.0001 0.001 D01 0.1 1 10
Plant Mean Cryplosporidium Concentration (Total oocysts/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
7-4
December 2005
-------
Exhibit 7.2 ICR Filtered Systems
Summary of 1,000 Modeled Source Water Distribution Curves
This graph knows modeled variability
gnd incwrtwnty m Mure* weier
Oyp'Oipartduj'tt occurrence lor ttttered
systems ruumnwilM 1.000 plaunble
curve? u$*d in the Economic Analysis
Ninety percent of th» modeled curve values
(all within toe daahad-lme uncertainty
bounds. The heavy center line reflects the
central tendency icroit ill 1.000 curves
Among th« 1,000 modtlM cufv«s. any tingle
Curve describe* vartabilKy in occurrence
aacu al tterftd syst*m* Steeper curves
ndicate less venabiliry from syitem lo system
Modeled Percent or Plants with
Average Source Water
CrypfoipcrKtum Concentration
6«Mmr 0 01 oocysts/L
- 5tti %Me Bound = 2ftH
Median Estimate = 23S
95th %tile Bound = 1fl%
1e-005
0.0001 0001 0.01 0.1 1
Plant Mean Cryptosporidium Concentration (Total oocysts/l)
10
Exhibit 7.3 ICR Unfiltered Systems
Summary of 1,000 Modeled Source Water Distribution Curves
100%
Modeled Percent of Plant* wttt
Average Source Water
Cfyptospondium Concentration
gelow 0 01 oocysts/L:
" 5th %tile Bound « 74%
~" Median Estimate = 5B%
95 MbM Bound = 39%
o%-
1e-005 00001 0001 001 0.1 1 10
Plant Mean Cryptosporidium Concentration (Total oocysts/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
7-5
December 2005
-------
Exhibit 7.4 ICR Supplemental Survey—Large Systems
Summary of 1,000 Modeled Source Water Distribution Curves
100% -
80%
60% '
0.
S 40%
0 20% -
0%
1e-005 0.0001 0.001 0.01 0.1 1 10
Plant Mean Cryplosporidium Concentration (Total oocysts/L)
Exhibit 7.5 ICR Supplemental Survey— Medium Systems
Summary of 1,000 Modeled Source Water Distribution Curves
100% -
1
•5
60%
40%
]
20% -
0% -
1e-005 0.0001 0.001 0.01 0.1 1 10
Plant Mean Cryptospondium Concentration (Total oocysts/L)
Exhibit 7.6 provides four alternative views of the 1,000 modeled curves from ICR filtered plants
data. Starting at the upper left and going clockwise it shows: 1) a copy of Exhibit 7.2, the 5th, and 95"1
Occurrence and Exposure Assessment
for the LT2ESWTR
7-6
December 2005
-------
percentiles for the ICR filtered curves; 2) Exhibit 7.2 with 25th and 75th percentile curves added; 3)
Exhibit 7.2 with 1st and 100"1 percentile curves added; and 4) the 10 actual curves that define the 100
percent bounds in the third plot. Although these four plots are all drawn from ICR filtered data, they
highlight two characteristics that are common to the sets of 1,000 curves modeled from all four data sets.
First, the actual curves do not cover the bounded 90 percent range uniformly. In each vertical
"slice" through the curves, points are distributed in a roughly symmetric, bell-shaped pattern with a
higher proportion near the median and fewer out towards the bounds. Comparing the second and third
plots in Exhibit 7.2, the relatively narrow range bounded by the 25* and 75lh percentiles (top right)
contains half of all curve points, while only 10 percent of curve points fall in the wider region between
the 90th and 100th percentile bounds (bottom and top region combined).
Second, the 1,000 curves in each set are not arranged in a regular, "parallel" pattern as suggested
by the percentile curves. To show this, the fourth graph (bottom right) plots the 10 curves that together
define the 100-percent bounds shown in the third graph (bottom left). In other words, each of these 10
curves is either higher or lower, in the vertical direction, than all the other curves in the collection of
1,000 at some point along the horizontal axis, but none of them is higher or lower everywhere. In
general, then, percentiles are defined by points from many individual curves that ascend at different rates
and cross over one another.
Section 7.3 describes how the collections of modeled distribution curves depicted in the Exhibits
7.2 through 7.5 are used in the Monte Carlo analysis to predict finished water occurrence levels.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-7
December 2005
-------
(A
£
3A|ie|nLuno
ro
o
m
a
I
o
o
0)
a>
1
5 B
I
SI
-------
7.2 Pre-LT2ESWTR Removal of Cryptosporidium
EPA estimated average annual removal of Cryptosporidium by treatment plants and combined
these estimates with the modeled source water occurrence distributions (described in Section 7.1) to
derive average finished water Cryptosporidium concentrations. The IESWTR and the LT1ESWTR
establish filtration requirements designed to provide finished water with a minimum 2 log (99 percent)
reduction in Cryptosporidium concentrations relative to source water levels. Although systems are
required to meet only the 2 log removal target, it is recognized that most systems will achieve greater
levels of Cryptosporidium removal. For example, some systems have unit processes in addition to
conventional treatment that provide higher Cryptosporidium removals. Based on studies described below
and in Chapter 5, EPA estimates that average yearly Cryptosporidium removal levels vary from 2 log to 4
log (99 percent to 99.99 percent) for small surface water systems (serving fewer than 10,000 people), and
from 2 log to 5 log (99 percent to 99.999 percent) for medium and large systems (serving at least 10,000
people).
EPA has characterized the distribution of these average removal values among systems as
triangular distributions. These triangular distributions are shown in Exhibits 7.7 and 7.8 for large and
small systems, respectively. The three points of any single triangular distribution represent the assumed
minimum, maximum, and mode (the most likely value) log removal.
The top and bottom figures in Exhibits 7.7 and 7.8 show two different sets of triangular
distributions for each system size category. The "lower-end" distributions represent conventional and
direct filtration plants that have no additional treatment processes that could enhance Cryptosporidium
removal and that minimally meet the effluent turbidity requirements of the IESWTR and LT1ESWTR.
To account for those plants that have or will have additional treatment processes prior to the LT2ESWTR
or that achieve very low filter effluent turbidity (e.g., combined filter effluent sO.15 NTU), a second
"higher-end" distribution was developed by shifting the lower-end triangle 0.5 log higher.
EPA accounts for uncertainty in the distributions of average removal values by allowing the
mode to vary. This creates a set of triangular distributions for each of the four system groups broken out
in Exhibits 7.7 and 7.8, with all possible distributions in a set equally likely. For example, the first set of
triangular distributions shown in Exhibit 7.7 represents the low-end distributions for medium and large
systems. It shows identical end-points for all possible distributions in this set, minimum and maximum
values of 2.0 and 4.5. The mode, however, varies uniformly between 2.5 and 3.0, with each possible
value defining a different triangular distribution.
The cumulative distributions shown in Exhibits 7.9 and 7.10 bound the range of log removals
expected for small and large systems, respectively. Within the range of cumulative distributions shown,
all possible curves are equally likely, and the modeling never generates a distribution that falls outside of
the depicted bounds.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-9
December 2005
-------
Exhibit 7.7 Triangular Removal Distributions for Medium and Large Systems
Lower-End Distributions
! 77
3 4
Log Rtmovalt
Higher-End Distributions— Estimate With 0.5 Log
Removal Credit
e
a
3 4
Log Removal*
Occurrence and Exposure Assessment
for the LT2ESWTR
7-10
December 2005
-------
Exhibit 7.8 Triangular Removal Distributions for Small Systems
Lower-End Distribution
Higher-End Distribution—Estimate With 0.5 Log
Removal Credit
Occurrence and Exposure Assessment
for the LT2ESWTR
7-11
December 2005
-------
Exhibit 7.9a Cumulative Probability Low-End Distributions for Large and Medium
Systems
D.
1
3.
Exhibit 7.9b Cumulative Probability High-End Distributions for
Large and Medium Systems
2 3
Log Removals
Occurrence and Exposure Assessment
for the LT2ESWTR
i-n
December 2005
-------
Exhibit 7.1 Oa Cumulative Probability Low-End Distributions for Small Systems
Log Removals
Exhibit 7.10b Cumulative Probability High-End Distributions for Small Systems
2 3
Log Removals
Occurrence and Exposure Assessment
for the LT2ESWTR
7-13
December 2005
-------
These triangular log-removal distributions were estimated from analysis of various filtration
studies described in Chapter 5. These studies covered a wide range of source waters and treatment
conditions. There are many challenges in estimating the Cryptosporidium removal achieved by full-scale
water treatment plants. Most full-scale studies show few detections of Cryptosporidium in finished water
due to the low source water occurrence of Cryptosporidium and insensitivity of the available analytical
methods. Pilot studies can overcome the source water concentration limitations by seeding the influent
with very high concentrations; however, the relationship between full-scale plant and pilot plant
performance is questionable. Another approach for addressing the challenge of estimating
Cryptosporidium removal is to assess the removal of other particles, such as aerobic spores and total
particle counts, as indicators of Cryptosporidium removal. Pilot studies have established strong
correlations between Cryptosporidium removal and the removal of spores and particles (McTigue et al.
1998, Dugan et al. 2001), and it is often possible to quantify removal of these indicators in full-scale
plants.
The key studies that support the minimum, modes, and maximum log removal values of the
triangular distributions are provided below. The values are based on indicator measurements (spores or
particles) at full-scale plants or for Cryptosporidium at pilot-scale plants, with the exception of one study
by Nieminski and Ongerth, (1995) which seeded Cryptosporidium in a full-scale plant that was not in
operation at the time of the study.
7.2.1 Large and Medium System Triangular Distributions
The large and medium system lower-end triangle represents conventional or direct filtration
plants that do not have additional unit processes to enhance Cryptosporidium removal and that minimally
meet IESWTR effluent turbidity requirements. Several studies support the log removal estimations
presented in Exhibit 7.7 with 2.5 and 3.0 log modes (being the most likely), 2.0 log minimum, and 4.5 log
maximum.
7.2.1.1 Modes
McTigue et al. (1998) conducted an on-site survey of 100 treatment plants and reported a median
particle removal of 2.8 log for particles greater than 0.2 urn, with removals ranging of 0.04 to 5.5 log.
Part of their investigation also included a pilot study that showed a significant correlation between
particle and protozoa removal when operating at similar temperature and coagulation conditions.
Nieminski and Bellamy (2000) sampled raw and finished water at 24 utilities and found that the median
removal of aerobic spores was 2.8 log. They also tested for Cryptosporidium and Giardia, but these
protozoa were rarely detected in finished water due to their relatively low occurrence and the insensitivity
of analytical methods. Nieminski and Ongerth (1995) reported average Cryptosporidium removal of 3
log for conventional and direct filtration pilot studies at relatively low influent and effluent turbidities of 4
NTU and 0.1 - 0.2 NTU, respectively. They also conducted full-scale studies that yielded lower average
Cryptosporidium removals of 2.25 log for conventional and 2.8 log for direct filtration. The reported data
may be lower than the actual removals because the authors reported Cryptosporidium log removals only
when Cryptosporidium were detected in the effluent (six of eight trials).
These studies support the choice of 2.5 and 3.0 log as modes of the triangular distribution for
those plants that achieve relatively lower effluent turbidity levels and lack additional unit processes. Data
from both the McTigue et al. (1998) and Nieminski and Bellamy (2000) studies covered a large number
Occurrence and Exposure Assessment
for the LT2ESWTR
7-14
December 2005
-------
of utilities, which indicates their median tog removals are more likely to be representative of national
averages than studies that sample one or two different utilities or source waters.
7.2.1.2 Maximum
Dugan et al. (2001) conducted several seeded runs at varying effluent turbidities. Under optimal
coagulation conditions, total Cryptosporidium removals (after sedimentation and filtration) was greater
than 5.0 log for runs with effluent turbidity ranging from 0.02 to 0.15 NTU. McTigue (1998) conducted
pilot studies and reported average Cryptosporidium removals of 4 log during stable operation. Patania et
al. (1995) also reported average and median values of approximately 4 log removal and a range of 1.4 to
6.2 log removal for Cryptosporidium from pilot studies at four different locations. These studies show
that greater than 5.0 log removal of Cryptosporidium can be achieved by filtration during optimal
conditions. The maximum log removal of 4.5 log represents an upper estimate of the mean log removal
full scale plants could achieve over the course of a year.
7.2.1.3 Minimum
The minimum 2 log removal is based on studies cited above and IESWTR and LT1ESWTR
requirements. The IESWTR and LT1ESWTR require an effluent turbidity of 0.3 NTU for 95 percent of
the time; plants meeting this limit are assumed for purposes of regulatory compliance to achieve at least a
2 log removal of Cryptosporidium. The minimum 2.0 log removal in the triangular distribution represents
the plants that operate frequently under stressed or sub-optimal conditions or with overall poorer
performance. Investigations by Huck et al. (2000) showed that Cryptosporidium removals diminished
when the filters were operating under sub-optimal conditions. Coagulant dose was decreased in two
direct filtration pilot-scale plants to achieve a target turbidity of 0.2 to 0.3 NTU. Median
Cryptosporidium removals decreased from 5.6 to 3.2 log and from 3.2 to 1 log. Emelko et al. (2000) also
studied Cryptosporidium removal during vulnerable filtration periods with a direct filtration system.
During an early breakthrough period, effluent turbidity increased from very low (0.04-0.08 NTU) to
approximately 0.2 NTU while the Cryptosporidium removal decreased from an average of 5.5 log to 2.1
log.
7.2.1.4 High-End Distribution
The high-end triangular distribution represents treatment plants with unit processes in addition to
conventional filtration and/or plants that achieve very low effluent turbidity. For example, EPA estimates
a pre-sedimentation basin can provide an additional 0.5 log Cryptosporidium removal. EPA also has
assumed that under the LT2ESWTR, plants that achieved combined filter effluent turbidity <;0.15 NTU in
95 percent of samples would receive an additional 0.5 log credit towards Cryptosporidium treatment
requirements. Considering that many systems have or will have these processes in place prior to the
LT2ESWTR or will achieve low effluent turbidity levels through programs like the Partnership for Safe
Water, a second triangular distribution (high-end removal) was developed by increasing the log removal
modes and endpoints by 0.5 log.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-15
December 2005
-------
7.2.2 Small System Triangular Distributions
For small systems, the predicted maximum and most likely Cryptosporidium removal levels are
lower than the large and medium system values. Smaller plants have fewer filters than larger plants;
therefore, fluctuations in the operation of a single filter have larger impacts on plant performance. For
example, backwashing and shutting down a filter for maintenance cause hydraulic fluctuations that can
substantially affect operational stability. In addition, if one filter performs poorly, the effect is greater in
a plant with fewer filters. Plant control is generally less automated at smaller plants, which makes water
quality control more difficult. The minimum estimated log removal is the same for small and larger
systems since systems are assumed to be in compliance with the IESWTR and LT1ESWTR and achieve
at least 2.0 log Cryptosporidium removal. Note, source water quality for small systems is assumed to be
the same as medium and large systems.
7.3 Description of Monte Carlo Model Used to Predict Finished Water Occurrence
As noted above, the predictions of finished water occurrence levels of Cryptosporidium were
derived through a two-dimensional Monte Carlo simulation model. This section describes the structure
and implementation of that model.
Two-dimensional Monte Carlo models are used in simulations where it is important to
differentiate between model inputs that describe uncertainty in a variable and those that describe
variability. As shown in Equation 7.1, presented at the beginning of this chapter, the basic algorithm for
calculating finished water occurrence at the individual system (or plant) level is a fairly simple function of
two variables: source water occurrence (Cs) and log removal (L). There is, however, plant-to-plant
variability in both of these inputs, and uncertainty in the "true" characterization of that plant-to-plant
variability. As noted previously, the Monte Carlo model considers normal operating conditions at each
plant, and therefore uses a constant treatment efficacy for the modeled time period of one year (i.e., a
given treatment efficacy was selected at random from the distribution of L values that vary from plant to
plant and then applied to all measurements in a year for that plant).
As noted earlier, one level of uncertainty considered in this analysis is embodied in the three
separate sets of filtered system data on source water occurrence. The use of these three data sets as
alternative views of the true occurrence distributions falls outside of the Monte Carlo simulation analysis.
These three source water occurrence data sets are used to produce three alternative views of finished
water occurrence for filtered systems.
Within each data set, uncertainty is incorporated in the Monte Carlo simulations through the
selection of alternative distributions as depicted in Exhibits 7.2 through 7.4. In similar fashion, the
uncertainty in the effectiveness of treatment is captured through the selection of alternative log removal
distributions described by the "space" depicted in the cumulative distributions shown in Exhibits 7.9 and
7.10.
For each combination of data set and plant size group, the Monte Carlo simulation was carried
out in two loops. The first was an uncertainty loop with 100 iterations, each with two steps—select a
distribution of plant-mean concentrations at random and select a single distribution of log removals. For
example, for the combination of the ICR filtered data set and the small plant category, the plant-mean
distribution was selected from among the 1,000 curves summarized in Exhibit 7.2, and the log-removal
distribution was obtained from the range of curves shown in Exhibit 7.10.
Occurrence and Exposure Assessment
for the LT2ESUTR 7-16 December 2005
-------
With these two distributions (plant-mean concentrations and log removals) defined in the
uncertainty loop, 250 values were then chosen at random from each distribution in a variability loop, the
second loop in the Monte Carlo simulation. From these values, 250 finished water values were computed
as shown in Equation 7.1. In total, 25,000 finished water concentrations were simulated in 100
uncertainty loops, with each collection of 250 values representing one possible national distribution of
mean finished water Cryplosporidium concentrations and the differences among sets of 250 representing
uncertainty about the true national distribution.
From each distribution of 250 estimated finished water plant-mean concentrations, summary
statistics including the mean, median, and percentiles were saved. The result was 100 estimates of each
distribution statistic, one from each uncertainty loop. To summarize this collection of estimates, the 5th,
50lh, and 95th value across the uncertainty dimension (100 estimates per percentile) was taken at every 5th
percentile across the plant-to-plant variability dimension. Connecting all these 50th percentiles, for
example, results in the median curves displayed in Exhibit 7.11. This method for summarizing the data is
similar to the method described in Exhibit 7.1 for summarizing collections of source water occurrence
curves. The finished water occurrence curves are presented for large and small systems and all four data
sets (three filtered and one unfiltered) in the next section (Exhibits 7.12 through 7.19).
7.4 Estimates of Pre-LT2 Finished Water Occurrence
Exhibits 7.11 and 7.12 compare the estimated finished water distribution curves across all three
filtered system and the one unfiltered system data sets, for small and large systems, respectively. The
distribution curves are the 50lh percentile or median curves. Exhibits 7.13 through 7.19 present all the
combinations of data set and system size category one at a time, adding the 5th and 95th percentile curves
to show 90-percent confidence bounds. Comparison of these finished water curves to the analogous
source water curves in Exhibits 7.2 through 7.5 reveals the impact of the 2.5 to 3.5 log removal for
Cryptosporidium modeled in Section 7.2.
Exhibit 7.20 gives a table of summary statistics for the estimated distributions for plant-mean
finished water Cryptosporidium concentration in Exhibits 7.13 through 7.19. Like the exhibits, the table
is broken out by both data set and system size category. The rows of the table reflect the variability
within each distribution curve for each data set. The first three rows for the ICR filtered data set, for
example, show the estimated mean, median, and 90lh percentile finished water concentrations for each of
the three distribution curves for large and small systems. Columns reflect uncertainty in the distributions.
The first column in Exhibit 7.20 shows the summary statistics for the median occurrence distribution for
large systems for each of the four data sets. The second column shows the same statistics for the 5th
percentile distribution (the lower end of the 90-percent confidence interval) and the third column shows
statistics for the 95th percentile distribution (the upper end of the 90-percent confidence interval). For
example, looking at the first row of the table, the mean concentration of the median distribution for large
plants, based on ICR data, is 1.54 * 10'4 oocysts/L, and a 90-percent confidence interval for this
concentration is 5.38 * 10's to 3.83 xKT4.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-11
December 2005
-------
Exhibit 7.11 All Data Sets, Small Systems
Comparison of Median Finished Water Distribution Curves
1.00E-10 1 OOE-09 1 OOE-Ot 10OE-07 10OE-Oe VOOE-05 1006-04 1OOE-03 10OE-02 1 OOE-01 1.00E«00 VOOEX1
FlnKhm Witor CiyplMportdium (
Exhibit 7.12 All Data Sets, Large Systems
Comparison of Median Finished Water Distribution Curves
10QE-10 IOOE-09 100E-W I.OOE-07 VOOE-OB 1.00E-DS 1.00E-04 VOOE-03 1.00E-02 1.00E-01 10OE«OO 10OE»01
Flntolwd W«r
Occurrence and Exposure Assessment
for the LT2ESWTR
7-18
December 2005
-------
Exhibit 7.13 ICR Filtered Data, Small Systems
Median Curve and 90-Percent Confidence Bounds
1QOE-10 1.00E-OT
1.0(je-07 1K>E-06 lOOE-OS l.OOE-04 LODE-OS 1.00E-02 100E-01 1.00E+00 1.00E*01
Flnlll»dW»tar CryptolporldilMli I0«y*t»rt.)
Exhibit 7.14 ICR Filtered Data, Large Systems
Median Curve and 90-Percent Confidence Bounds
100E-10 100E-OS 100E-08 1 OOE-07 1 OOE-M 10OE-05 1 OOE-M 1 OOE-03 100E-02 1.00E-Ot 1 00£*00 1,OOE»01
Flol«h»4 W*t*r Crypto.p«rtdlum [
Occurrence and Exposure Assessment
for the LT2ESWTR
7-19
December 2005
-------
Exhibit 7.15 ICRSSL Data, Small Systems
Median Curve and 90-Percent Confidence Bounds
1006-10 100E-09 1DOE-C8 1006-07 1.00E-M 1 OOE-05 10OE-04 1.00E-03 10OE-02 10CE-01 1 OOE»00 1 DOE-OI
Flnkhad Water CiypMtperMlum {
Exhibit 7.16 ICRSSL Data, Large Systems
Median Curve and 90-Percent Confidence Bounds
1.00E-10 1-OOE-Oe 1 DCE-08 100E-D7 1.00E-M 1.00E-05 1.00E-04 1.DOE-03 1.00E-02 10OE-01 10OE*M 10OE*01
Occurrence and Exposure Assessment
for the LT2ESWTR
7-20
December 2005
-------
Exhibit 7.17 ICRSSM Data, Small Systems
Median Curve and 90-Percent Confidence Bounds
3 «*
1006-10 1 OOE-09 1 DOE-OS 1.00E-07 1 OOE-06 10OE-05 10OE-04 10OE-03 1006-02 1 OOE-01 1.006*00 10OE*01
Exhibit 7.18 ICRSSM Data, Large Systems
Median Curve and 90-Percent Confidence Bounds
100E-IO 100E-06 100E-OI 100E-07 1 OOE-06 1.00E-OS 1 OOE-04 10OE-03 I OCE-C2 1 OOE-01 1 OOE-00 1 OOE.01
FlnlBhvd Watw CryptollKMMIyrn p>0«ytun.(
Occurrence and Exposure Assessment
for the LT2ESWTR
7-21
December 2005
-------
Exhibit 7.19 ICR Unfiltered Data, Large Systems
Median Curve and 90-Percent Confidence Bounds
100E-10 1 OOE-09 I.ME-Ot 1 OOE-07 100E-M 1.00E-49 1.ME-04 100E-03 I ME-02 100E-01 1 OOEXXI tOOEXll
tf W«l*r Cryptasportdhim (OocyiuA.)
Occurrence and Exposure Assessment
for the LT2ESWTR
7-22
December 2005
-------
Exhibit 7.20. Summary of Finished Water Occurrence Distributions
by Data Source and System Size
Data Source
ICR
ICRSSL
ICRSSM
Unfiltered
Statistic
Mean
Median
90th %ile
Mean
Median
90th %ile
Mean
Median
90th %ile
Mean
Median
90th %ile
Larae Systems
Median
Distribution
1.54x10^
5.05x10*
1.63x10""
3.50X10"5
7.75x10*
7.74X10"5
6.24x10*
8.28x10*
1.17x10^
1.95X10"2
1.47x10"3
1.39X10"2
5th
Percentile
Distribution
5.38x10*
2.79x10*
8.22x10*
2.34x1 0'5
5.48x10*
5.17x10"s
3.66x1 05
5.32x1 0*
7.20x10"5
2.71x10"3
4.36x10""
5.05x10"3
95th
Percentile
Distribution
3.83x10""
8.22x10*
2.66x10""
5.28x10*
1.09x10*
1.23x10""
9.64x10*
1.20x10*
1.80x10""
2.00x1 0"2
2.97x1 0"3
3.11x10"2
Median
Distribution
3.62x10""
1.63x10"5
3.85X10"4
8.33X10"5
2.60x10*
1.82x10^
1.46x10""
2. 56x1 0"5
2.85x10""
1.95x10'2
1.47X10'3
1.39x10"2
Small Svstem
5th
Percentile
Distribution
1.33x10^
9.75x10*
2.10x10^
4.86x10"5
1.58x10"5
1.19x10""
7. 87x1 0"5
1.57X10"5
166x10""
2.71x1 0"3
4.36x10""
5.05x10"3
95th
Percentile
Distribution
7.70x10""
2.44x10'5
6.24X10"4
1.30x10""
3.73X10"5
2.60x10""
2.61x10""
3.75x10"5
4.46x10""
2.00X10"2
2.97x1 0"3
3.11X10"2
Note: Data provided in Exhibit 7.20 are in oocysts per L.
7.5 Comparison of EPA Estimates with Aboytes et al. (2000)
A study by Aboytes et al. (2000) provides an alternative perspective on Cryptosporidium
occurrence in finished drinking water and the efficacy of treatment. This study involved collecting 100-L
finished water samples on a monthly basis from 80 surface water utilities. Samples were analyzed for
infectious Cryptosporidium parvum with a cell culture-PCR (CC-PCR) method. The objective of the
study was to "assess the adequacy of treatment to protect against infectious Cryptosporidium in drinking
water." All utilities in the study were enrolled in the Partnership for Safe Water, a voluntary cooperative
program that seeks to optimize treatment plant performance. Most samples had turbidity below 0.1 NTU
and all were below 0.3 NTU, the standard set by the 1ESWTR. Among 1,674 samples of 100 L each, 24
were positive for infectious C. parvum (LeChevallier 2001). The authors reported an average CC-PCR
recovery efficiency of 32.3 percent. Hence, if it is assumed that one infectious oocyst accounted for each
positive sample, and the oocyst count is adjusted for average recovery, these results produce a mean
concentration of infectious oocysts of 4.4 * 10"4 oocysts/L, or 0.044 oocysts/100 L.
To compare results from Aboytes et al. with EPA finished water Cryptosporidium estimates
based on results from the ICR and 1CRSS, it is necessary to consider the fraction of oocysts that are
infectious. Because oocysts lose viability in the environment, it is expected that infectious oocysts are
only a fraction of the total number of oocysts in a water sample. While the CC-PCR method registers
only infectious oocysts, the ICR Method and EPA Methods 1622/23 count total oocysts without regard to
whether they are viable and infectious. To estimate the fraction of oocysts that may be infectious, EPA
evaluated a study by LeChevallier et al. (2003) that analyzed several hundred source water samples from
six utilities using both the CC-PCR method and Method 1623. Oocysts were detected in 60 of 593
samples (10.1 percent) by Method 1623, and infectious oocysts were detected in 22 of 560 samples (3.9
percent) by the CC-PCR procedure. Recovery efficiencies for the two methods were similar. According
to the authors, these results suggest that approximately 37 percent of the oocysts detected by Method
Occurrence and Exposure Assessment
for the LT2ESWTR
7-23
December 2005
-------
1623 were viable and infectious. Based on these results, as well as consideration of the structure of
counted oocysts, EPA assumes that the fraction of oocysts that is infectious may average from 15 to 25
percent for the ICR and 30 to 50 percent for the ICRSS (as described in Section 5.2.4 of the Economic
Analysis for the LT2ESWTR).
Jf the estimates of mean large plant finished water oocyst concentrations from the median
distributions from Exhibit 7.20 are multiplied by 40 percent to adjust for the fraction of oocysts that are
infectious, the mean finished water concentrations of infectious oocysts are as follows:
ICR Mean = 6.1 x IO'5; ICRSSM Mean = 2.5 x 10'5; ICRSSL = 1.4 x 10'5 (oocysts/L)
Thus, the mean finished water infectious oocyst level reported in Aboytes et al. of 4.4 x 10"4
oocysts/L is a factor of 7 greater than the EPA mean estimate based on the ICR and a factor of 30 greater
than the ICRSSL estimate. While the reason for this significant discrepancy is unknown, it cannot be
fully attributed to potential error in factors such as the fraction of oocysts that are infectious. Rather, it
may indicate that the ICR and ICRSS underestimate source water Cryptosporidium occurrence and/or that
EPA has overestimated treatment efficacy, as discussed below.
Unfortunately, source water data are not available for plants in the Aboytes et al. study, so it is
not possible to directly determine their average treatment efficiency. However, removal efficiency may
be estimated based on other survey data. The ICRSS was conducted during a time frame similar to
Aboytes et al., and the filtration and separation steps used in the CC-PCR method of Aboytes et al. are
similar to those in Method 1622/23 used in the ICRSS. Consequently, the ICRSS source water data may
be somewhat comparable to the Aboytes et al. finished water data. The mean source water
Cryptosporidium concentration of plants in the ICRSS, adjusted for average recovery of 43 percent, was
0.14 oocysts/L. If this value is multiplied by 40 percent as an estimate of the fraction of oocysts that are
infectious, the result is a mean source water concentration of 0.056 infectious oocysts/L. If this were the
mean source water concentration in the Aboytes et al. study, then the plants in that survey achieved an
average removal of 1.7 log to produce the mean finished water concentration of 0.0011 oocysts/L that
was measured. For the plants in the Aboytes et al. survey to have achieved a mean oocyst removal of 2.5
log, which was the lowest mean removal assumed in EPA estimates, the mean source water oocyst
concentration would have to have been over twice that measured during the ICRSS. Either hypothesis
indicates that EPA may have underestimated the risk from Cryptosporidium in drinking water by
underestimating finished water oocyst levels.
7.6 Summary
Through modeling efforts that encompass the uncertainty and variability of Cryptosporidium
concentrations in source waters, treatment efficiencies, sampling, and analytical methods, the occurrence
of Cryptosporidium in finished water was estimated for Pre-LT2ESWTR conditions. While estimated
finished water occurrence levels are very low, the presence of any Cryptosporidium in drinking water
poses the risk of consumption, which can lead to adverse health effects. In the Economic Analysis for the
LT2ESWTR, EPA uses these finished water occurrence distributions to estimate risk prior to the
LT2ESWTR, and subsequently, to estimate potential health benefits realized from the LT2ESWTR.
Occurrence and Exposure Assessment
for the LT2ESWTR
7-24
December 2005
-------
8. Population Profile for Exposure Assessment
Accurately assessing the risk associated with the LT2ESWTR regulatory options requires
estimating and characterizing the population served by PWSs affected by the rule. This chapter describes
the data sources, assumptions, and methodology used to estimate the population at risk from exposure to
microbial contaminants.
8.1 Population Profile
The LT2ESWTR applies to all PWSs that use surface water or ground water under the direct
influence of surface water (GWUD1) and, by extension, the populations they serve. Characterization of
population served is available from two major sources: the Safe Drinking Water Information System
(SDWIS), as summarized in the Drinking Water Baseline Handbook (USEPA 200Ib), and the 1996 ICR.
Detailed system information, including system counts and population served, as reported by states to
EPA, is maintained within SDWIS. The SDWIS inventory is "frozen" near the end of the calendar year
and distributed to state drinking water programs for verification of the number and types of systems.
These data are incorporated into the Baseline Handbook for use in conducting cost-benefit assessments.
The second source of population data is the 1996 ICR. Treatment plant data, including
population served, were collected from nearly 300 large surface and ground water systems. The ICR
provided EPA with extensive information on chemical byproducts, microbial contaminants, and treatment
capabilities to control these contaminants and was used to develop Cryptosporidium occurrence profiles
for LT2ESWTR risk analyses. Population served by the large water systems as reported under the ICR
does not correlate with SDWIS data, mainly due to the way in which wholesale populations were
assigned. Under the ICR, systems reported total population served, including retail and wholesale
portions. Under SDWIS, systems report only their retail population served—systems that purchase water
from other systems and distribute it to their retail populations are considered stand-alone systems in
SDWIS. To maintain consistency in estimating population exposed across all system size categories, the
SDWIS population inventory was selected for all LT2ESWTR risk analyses.
Exhibit 8.1 presents the total population served by public water from SDWIS (4th quarter 2000
data) for community water systems (CWSs), nontransient noncommunity water systems (NTNCWSs),
and transient noncommunity water systems (TNCWSs). Because a person need ingest only one
Cryptosporidium oocyst to become infected, LT2ESWTR risk analyses must consider all types of PWSs,
including those that provide drinking water only part of the time or to transient populations. The total
population served presented in Exhibit 8.1 (279.8 million) is 99.4 percent of the total population in 2000,
as reported by the United States Census Bureau. The inclusion of TNCWSs in population estimates may
result in some double-counting; for instance, people who drink the water at a rest area TNCWS may be
served by a CWS at home. These estimates represent the broad universe of populations and systems to be
considered in the risk assessment.
Occurrence and Exposure Assessment
for the LT2KSWTR
8-1
December 2005
-------
Exhibit 8.1 Estimates of Population Served By System Size (2000)
Size Category
Large PWSs
(serving over 100,000)
Medium PWSs
{serving 10,001-100,000)
Small PWSs
(serving 10.000 or fewer)
Total Population
Population
Served by
PWSs
115,597,748
96,660,320
67,536,316
279,794,384
Population Served by
Surface Water and
GWUDI Systems
97,821,628
60,639.692
18,980,512
177,441,832
Percent of
Total
Population
Served by
PWSs
35.0%
21.7%
6.7%
63.4%
Percent of
Surface Water
and GWUDI
Population
Served in Each
Size Category
55.3%
34.0%
10.7%
100%
Source: USEPA 2000g
In additional to system type, source water and system size are important distinctions in evaluating
the risks of microbial contamination. Source water for drinking water treatment systems can be ground
water, surface water, ground water under the direct influence of surface water (GWUDI), or a
combination of the three. GWUDI is any ground water source with significant occurrence of
microorganisms or significant and relatively rapid shifts in water characteristics that closely correlate with
surface water conditions. It is important to note that system classification in SDWIS is not by
predominant water source. A system with any input of surface water, even if it is a very small portion of
the total flow, is categorized as a surface water system. Systems using surface water or GWUDI are
susceptible to microbial contamination and are targets of the LT2ESWTR. They provide water to more
than 63 percent of the total population served by PWSs.
8.2 Characterization of Population, Including Sensitive Subpopulations
Several population subgroups are of particular importance with respect to the potential health
risks posed by microbial contaminants, especially Cryptosporidium and Giardia. In general, people with
compromised immune systems, infants, and the elderly (if already weakened by other conditions) are at
higher risk for the negative health effects caused by Cryptosporidium and Giardia. This section discusses
the size of these sensitive subpopulations and possible future population trends within these subgroups.
Individuals with compromised immune systems, most notably three subgroups comprised of
people undergoing treatment for cancer, recipients of organ transplants, and those with acquired
immunodeficiency syndrome (AIDS), have a greater risk than immunocompetent individuals of
developing severe, life-threatening illness if they become infected (Gerbaet al. 1996). Estimates of the
number of persons in each subgroup, compiled from Centers for Disease Control and Prevention (CDC)
data and other surveys, are presented in Exhibit 8.2. The portion of immunocompromised persons served
by surface and GWUDI systems is estimated to be 0.6 percent of all people served by these systems. The
numbers in Exhibit 8.2 were calculated by multiplying the number of immunocompromised people in
each category by the percentage of the total U.S. population served by surface water and GWUDI systems
and by the percentages of the surface water and GWUDI population served by large, medium, and small
systems. These estimates assume random distribution of the sensitive subpopulation among ground
Occurrence and Exposure Assessment
for the LT2ESWTR
8-2
December 2005
-------
water, surface water, and GWUDI systems, among systems of different sizes, and between individuals
served and not served by PWSs. They also do not account for those who use alternative water sources
because of their illness.
Exhibit 8.2 Estimates of Immunocompromised Population
Served By Surface Water and GWUDI Systems By System Size
Size Category
Large PWSs
(serving over 1 00,000)
Medium PWSs
(serving 10,001-100,000)
Small PWSs
(serving 10,000 or fewer)
Total
Population
Served by
Surface Water
and GWUDI
97,821,628
60,639,692
18,980,512
177,441,832
Percentage of Population Served by
Surface Water and GWUDI Systems That
ts Immunocompromised
Percentage of Total U.S. Population
Number of
People
Living with
AIDS1
112,227
69,570
21,776
203,572
0.11%
0.07%
Number of
New
Cancer
Cases2
440,754
273,224
85,520
799,498
0.45%
0.28%
Number of
Organ
Recipients
Living3
49,848
30,367
9,167
89,382
0.05%
0.03%
Total
Immuno-
Compromised
602,829
373,160
116,463
1,092,452
0.62%
0.39%
Source: 1. CDC 2000c
2. U.S. Census Bureau 2001a
3. United Network of Organ Sharing 2000
Notes: Not all new cancer cases are immunocompromised, since not all undergo chemotherapy. However, for
purposes of simplicity, all new cases were assumed to be immunocompromised. In addition, cancer patients
diagnosed in previous years may be immunocompromised but are not included here. Cancer cases do not include
certain skin cancers and most in situ carcinomas.
Organ recipients is the number of recipients living who received a donation from Oct. 1. 1987 to Dec. 31, 1999.
Infectious diseases can have a greater impact on the elderly because immune function declines
with age, antibiotic treatment is less effective, and malnutrition is more common (Meyers 1989).
However, not all elderly are at increased risk for Cryptosporidium; only those with underlying health
problems are. Because it is difficult to determine the percentage of elderly that are at increased risk, the
entire elderly population is shown in Exhibit 8.3, as a very conservative estimate of those that could be at
risk. As the relatively large generation born between 1946 and 1964 (the "Baby Boomers") ages, the age-
based population groups at risk for infections also will increase. Exhibit 8.3 provides estimates of the
current and future population age 65 and older served by surface and GWUDI systems, derived from U.S.
Census Bureau (200 Ib) projections. Population projections assume the proportion of the total population
served by surface water and GWUDI systems is constant through time, along with the percentages of that
population served by small, medium, and large systems. These estimates also assume random distribution
of this population subgroup among ground water, surface water, and GWUDI systems and among systems
of different sizes.
As shown in Exhibit 8.3, the population age 65 and older affected by the LT2ESWTR is expected
to more than double from 2000 to 2050. Based strictly on this size increase in a sensitive subpopulation,
Occurrence and Exposure Assessment
for the LT2ESWTR
8-3
December 2005
-------
waterborne disease cases in the elderly could double in the next 50 years. Also, the percentage of those
65 and older served by surface and GWUDI systems is expected to increase from 7.8 to 10.4 percent of
the total U.S. population by 2020, indicating an increase in the relative proportion of the elderly
population to the general population.
Exhibit 8.3 Estimates of U.S. Population 65 and Older
Served By Surface Water and GWUDI Systems by System Size
Size Category
Large PWSs (>100,000)
Medium PWSs (10,001-100,000)
Small PWSs {10,000 or fewer)
Total
Percent of Total Projected U.S.
Population
2000
12,156,774
7,545,584
2,360,753
22,063,110
7.8%
2010
13,797,619
8,564,039
2.679,392
25,041,050
8.3%
2020
18,676,693
11,586.844
3,625,124
33,879,661
10.4%
2050
28,487,748
17,682,051
5,532,104
51,701,902
12.8%
Source: Derived from U.S. Census Bureau 2001 b and Exhibit 8.1
Young children are a vulnerable population, subject to more severe health effects caused by
Giardia, Cryptosporidium, and other waterborne pathogens. In particular, infants who have not yet
developed immune responses are vulnerable to exposure and are at higher risk of severe dehydration
caused by diarrhea. Although direct exposure to Cryptosporidium in tap water is minimal for infants (tap
water used for formula is generally boiled) and low for children under 5 (USEPA 200Ib), children are
more vulnerable to severe health effects when they are exposed. Exhibit 8.4 estimates the population of
children under 5 years of age by system size for those served by surface and GWUDI systems and the
projected increases in this subpopulation. Population projections assume the proportion of the total
population served by surface water and GWUDI systems is constant through time, as are the percentages
of that population served by small, medium, and large systems. These estimates also assume random
distribution of this population subgroup among ground water, surface water, and GWUDI systems and
among systems of different sizes.
Exhibit 8.4 Estimates of U.S. Population under Age 5
Served by Surface Water and GWUDI Systems by System Size
Size Category
Large PWSs (>1 00.000)
Medium PWSs (10,001-100,000)
Small PWSs (10,000 or fewer)
Total
Percent of Total Projected U.S.
Population
2000
6,662,045
4.135,063
1,293.718
12,090,826
4.3%
2010
6,982,710
4,334,096
1,355,989
12,672,795
4.2%
2020
7,626,124
4.733.456
1,480,935
13,840,516
4.3%
2050
9,350,349
5,803,665
1,815.766
16,969,780
4.2%
Source: Derived from U.S. Census Bureau 2001b and Exhibit 8.1
Occurrence and Exposure Assessment
for the LT2ESWTR
8-4
December 2005
-------
8.3 Population Profile for Exposure Assessment—Summary
Risk assessments of various LT2ESWTR regulatory scenarios must take into account the
characteristics of the population affected. The total population served by public water systems is
estimated to be 279.8 million, based on SDWIS data. More than 63 percent of this population (177.4
million) is served by surface water or GWUDI systems and is affected by the LT2ESWTR. Sensitive
subpopulations, namely the immunocompromised, children under the age of 5, and the elderly, make up
approximately 35.2 million of the 177.4 million individuals, or 20 percent of the population served by
surface water and GWUDI systems. Not everyone in these groups is at increased risk, since not all
elderly have the same vulnerability and since children may have limited exposure to drinking water.
However, the number of immunocompromised individuals is expected to increase over the next several
decades, making accurate risk assessment for these sensitive subpopulations all the more critical.
Occurrence and Exposure Assessment
for the LT2ESWTR
8-5
December 2005
-------
9. References Cited
Aboytes, Ramon, G. Di Giovanni, F. Abrams, J. Nieroda, N. Shaw, R. Kozik, and M.W. LeChevallier.
2000. Advances in monitoring finished water for Cryptosporidium parvum. Proceedings from
A WWA Water Quality Technology Conference.
Adham, S.S., R.S. Trussell, P.P. Gagliardio, and R.R. Trussell. 1998. Rejection of MS-2 virus by RO
membranes. J AWWA 90(9): 130-135.
Alhajjar, B. J., S. L. Stramer, D. O. Cliver, and J. M. Harkin. 1988. Transport modelling of biological
tracers from septic systems. Water Research 22 (7):907-916.
Alvarez, M.E., M. Aguilar, A. Fountain, N. Gonzalez, O. Rascon, and D, Saenz. 2000. Inactivation of
MS-2 phage and poliovirus in groundwater. Can. J. Microbiol. 46(2): 159-65.
Anderson, B.C. 1986. Effect of drying on the infectivity of Cryptosporidia-lzden calf feces for 3-to 7-day
old mice. Amer. J. Vet, Res. 47: 2272.
Anderson, B.C. 1985. Moist heat inactivation of Cryptosporidium sp. Amer. J. Publ. Health 75:1433.
Archer, G.P., W.B. Belts, and T. Haigh. 1993. Rapid differentiation of untreated, autoclaved and ozone-
treated Cryptosporidium parvum oocysts using dielectrophoresis. Microbios 73: 165. (Cited in
Payer etal. 1997)
Arora, H., M. LeChevallier, R. Aboytes, E. Bouwer, C. O'Melia, W. Ball, W. Weis, and T. Speth. 2000.
Full-scale evaluation of riverbank filtration at three Midwest water treatment plants. Proceedings
of the A WWA Water Quality Technology Conference, Salt Lake City, Utah, November, 2000.
Aulicino, F. A., A. Mastrantonio, P. Orsini, C. Bellucci, M. Muscillo, and G. Larosa. 1996. Enteric
viruses in a wastewater treatment plant in Rome. Water, Air, & Soil Pollution 91 (3-4):327-334.
AWWARF. 1997. AWWARF: Drinking water inspectorate fact sheet-Astroviruses. Website visited
August 3,2000. Website last updated March 1997.
Barcina, I., J. M. Gonzalez, J. Iriberri, and L. Egea. 1989. Effect of visible light on progressive dormancy
of Escherichia coli cells during the survival process in natural fresh water. Appl. Environ.
Microbiol. 55 (1):246-251.
Barwick, R.S., D.A. Levy, G.F. Craun, M.J. Beach, and R.L. Calderon. 2000. Surveillance for
waterborne-disease outbreaks-United States, 1997-1998. CDC Surveillance Summaries, May 26,
2000. MMWR 49(SS-4): 1-35.
Battigelli, D.A., M.D. Sobsey and D.C. Lobe. 1993. The inactivation of hepatitis A virus and other model
viruses by UV irradiation. Wat. Sci. Tech. 27(3-4): 339-342.
Baxby, D., C.A. Hart, and C. Taylor. 1983. Human cryptosporidiosis: a possible case of hospital cross
infection. Br. Med. J. 287: 1760-1761.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-1
December 2005
-------
Berg, G., D. R. Dahling, G. A. Brown, and D. Berman. 1978. Validity of fecal coliforms, total coliforms,
and fecal streptococci as indicators of viruses in chlorinated primary sewage effluents. Appl.
Environ. Microbiol. 36 (6):880-4.
Bergmire-Sweat, D., J. Morgan, K. Wilson, K. Von Alt, L. Marengo, T. Bennett, Y.M. Lee, V.C.W.
Tsang, W.R. MacKenzie, B. Furness. 1998. Cryptosporidiosis at Brushy Creek: Describing the
epidemiology and causes of a large outbreak in Texas. Epidemiology in Texas. Website visited
September 29, 2000.
Berman, D., R. Sullivan, C.J. Hurst. 1992. Effect of the method of preparing monochloramine upon
inactivation of MS2 coliphage, Escherichia coli, and Klebsiella pneumonias. Can. J. Microbiol.
38(1): 28-33.
Bingham, A.K., E.L. Jarroll, E.A. Meyer, and S. Radulescu. 1979. Giardia sp.: Physical factors of
excystation in vitro, and excystation vs. eosin exclusion as determinants of viability. Exp.
Parasitol. 47: 284-291.
Blackmer, F., K.A. Reynolds, C.P. Gerba, and I.L. Pepper. 2000. Use of integrated cell culture-PCR to
evaluate the effectiveness of poliovirus inactivation by chlorine. Appl. Environ. Microbiol. 66(5):
2267-68.
Blewett, D.A. 1989. Disinfection and oocysts. Cryptosporidiosis. Proceedings of the First International
Workshop, 1988. Ed. K.W. Angus and D.A. Blewett. Edinburgh: The Animal Disease Research
Association. 107-116.
Braghetta, A., J.G. Jacangelo, S. Chellam, M.L. Hotaling, and B.A. Utne. 1997. DAF pretreatment: its
effect on MF performance. J. AWWA. 89(10): 90-101.
Brown, E.A.E., D.P. Casemore, A. Gerken, I.F. Greatorex. 1989. Cryptosporidiosis in Great Yarmouth -
the investigation of an outbreak. Public Health 103: 3-9.
Brush, C.F., W.C. Ghiorse, L.J. Anguish, J.-Y. Parlange, and H.G. Grimes. 1999. Transport of
Cryptosporidium oocysts through saturated columns. J. Env. Qual. 28: 809-815.
Brush, C.F., M.F. Walter, L.J. Anguish, and W.C. Ghiorse. 1998. Influence of pretreatment and
experimental conditions on electrophoretic mobility and hydrophobicity of Cryptosporidium
parvum oocysts. Appl. Env. Microbiol. 64: 4439—4445.
Bukhari, Z., T.M. Hargy, J.R. Bolton, B. Dussert and J.L. Clancy. 1999. Medium-pressure UV light for
oocyst inactivation. J. AWWA 91(39): 86-94.
Bukhari, Z., M.M. Marshall, D.G. Korich, C.R. Fricker, H.V. Smith, J. Rosen, J.L. Clancy. 2000.
Comparison of Cryptosporidium parvum viability and infectivity assays following ozone
treatment of oocysts. Appl. Environ. Microbiol. 66(7): 2972-2980.
Campbell, A.T. 2000. Determination of the effect of UV irradiation on human-derived Giardia lamblia.
Final report. Report # ATC00105. Prepared for Safewater Solutions, Inc., Scottsdale, AZ. May
2000.
Occurrence ami Exposure Assessment
for the LT2ESWTR
9-2
December 2005
-------
Campbell, A.T., L.J. Robertson, and H.V. Smith. 1992. Viability of Cryptosporidium parvum oocysts:
Correlation of in vitro excystation with inclusion or exclusion of fluorogenic vital dyes. Appl.
Environ. Microbiol 58: 3488-3493.
Campbell, A.T., L.J. Robertson, M.R. Snowball, and H.V. Smith. 1995. Inactivation of oocysts of
Cryptosporidium parvum by ultraviolet irradiation. Water Res. 29: 2583.
Casemore, D.P. 1990. Epidemiological aspects of human cryptosporidiosis. Epidemiol. Infect. 104: 1-28.
Casemore, D.P., S.E. Wright, and R.L. Coop. 1997. Cryptosporidiosis—human and animal epidemiology.
Cryptosporidium and Cryptosporidiosis. Ed. R. Payer. Boca Raton: CRC Press. 65-92.
CDC. 200la. National respiratory and enteric virus surveillance system: Adenoviruses. Website visited
December 7,2001. Website last reviewed August 20,2001.
CDC. July 2001b. Hepatitis E: Clinical features. Website visited October 17, 2002.
CDC. July 200Ic. Hepatitis E: Epidemiologic features. Website visited October 17, 2002.
CDC. 2001d. Respiratory and enteric virus branch: Rotavirus. Website visited April 16,2003. Website
last updated August 21,2001.
CDC. 2000a. Viral Hepatitis A - Fact sheet. Website visited 2 Aug. 2000. Website last updated May 1,
2000.
CDC. 2000b. Hepatitis A frequently asked questions. Website visited 2 Aug. 2000. Website last updated
May 1, 2000.
CDC. 2000c. HIV/AIDS surveillance report. 12(2). Year end edition.
CDC. 1999. Public health dispatch: outbreak of Escherichia coli O157:H7 and Campylobacter among
attendees of the Washington County Fair-New York, 1999. Morbidity and Mortality Weekly
Report 48(36): 803.
CDC. 1998a. Non-polio enterovirus infections. Website visited 2 Aug. 2000. Website last updated June 9,
1998.
CDC. 1998b. Foodborne outbreak of cryptosporidiosis-Spokane, Washington, 1997. .7. Amer. Med.
Assoc. 280(7): 595-596.
Champliaud, D., P. Gobet, M. Naciri, O. Vagner, J. Lopez, J.C. Buisson, I. Vargas, G. Harly, R.
Mancassola, and A. Bonnin. 1998. Failure to differentiate Cryptosporidium parvum from C.
meleagridis based on PCR amplification of eight DNA sequences. Appl. and Environ. Microbiol.
64(4): 1454-58.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-3
December 2005
-------
Chappell, C.L., P.C. Okhuysen, C.R. Sterling, and H.L. DuPont. 1996. Cryptosporidiumparvum:
Intensity of infection and oocyst excretion patterns in healthy volunteers. Jour. Infect. Dis. 173:
232-236.
Chappell, C.L, P.C. Okhuysen, C.R. Sterling, C. Wang, W. Jakubowski, H.L. DuPont. 1999. Infectivity
of Cryptosporidium parvum in healthy adults with pre-existing anti-C. parvum serum
immunoglobulin G. Am. J. Trap. Med. Hyg. 60(1): 157-64.
Chen, Y.S., and J.M. Vaughn. 1990. Inactivation of human and simian rotaviruses by chlorine dioxide.
Appl. and Environ. Microbiol. 56(6): 1363-66.
Clancy, J.L., Z. Bukhari, T.M. Hargy, J.R. Bolton, B.W. Dussert, M.M. Marshall. 2000. Using UV to
inactivate Cryptosporidium. J. A WWA 92(9): 97-104.
Clavel, A., A.C. Arnal, B.C. Sanchez, J. Cuesta, S. Letona, J.A. Amiguet, F.J. Castillo, M. Varea, and
R. Gomez-Lus. 1996. Respiratory cryptosporidiosis: case series and review of the literature.
Infection 24(5): 341-346.
Cliver, D.O. 1997. Virus transmission via food. World Health Stat. 50(1-2): 90-101. (Cited in USEPA
1999c).
Coffey, B.M, Huck P.M., Maurizo, D.D., Emelko, M.B., Douglas, I.P., and J. Van Den Oever. 1999. The
effect of optimizing coagulation on the removal of Cryptosporidium parvum and Bacillus
subtillis. Proceedings from A WWA Water Quality Technology Conference.
Conlon, W.J. 1990. Membrane processes. Water Quality and Treatment: A Handbook of community
water supplies. F.W. Pontius, ed. American Water Works Association. 4th ed. New York:
McGraw-Hill. 709-746.
Cordell, R.L., and D.G. Addiss. 1994. Cryptosporidiosis in child care settings: a review of the literature
and recommendations for prevention and control. Pediatr. Infect. Dis. Jour. 13(4): 310-317.
Craft, J.C. 1982. Experimental infection with Giardia lamblia in rats. J. Infect. Dis. 145:495-498.
Craik, S.A., D. Weldon, G.R. Finch, J.R. Bolton, and M. Belosevic. 2001. Inactivation of
Cryptosporidium parvum oocysts using medium- and low-pressure ultraviolet radiation. Wat. Res.
35(6): 1387-1398.
Craun, G.F. 1990. Waterborne giardiasis. Giardiasis. E.A. Meyer, ed. Elsevier Science Publishers. 267-
293.
Craun, G.F. 1986. Waterborne giardiasis in the United States, 1965-84. Lancet Aug. 30, 1986. 513-514.
Crawford, F.G., and S.H. Vermund. 1988. Human cryptosporidiosis. CRC Crit. Rev. Microbiol. 16(2):
113-159.
Current, W.L. 1986. Cryptosporidium: Its biology and potential for environmental transmission. CRC
Crit. Rev. Environ. Control 17(1): 21-51.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-4
December 2005
-------
Dahling, D. R., and R. S. Safferman. 1979. Survival of enteric viruses under natural conditions in a
subarctic river. Appl, Environ. Microbiol. 38 (6):1103-10.
Daly, J.J. 1983. Parasitic protozoa of humans. Microbiology: Basic Principles and Clinical Applications.
Ed. N,R. Rose and A.L. Barron. New York: Macmillan Publishing Co.
Davenport, C. V., E. B. Sparrow, and R. C. Gordon. 1976. Fecal indicator bacteria persistence under
natural conditions in an ice-covered river. Appl. Environ. Microbiol. 32 (4):527-36.
Davies, C. M, S. C. Apte, and S. M. Peterson. 1995. Beta-D-galactosidase activity of viable,
non-culturable coliform bacteria in marine waters. Letters in Applied Microbiology 21 (2):99-102.
Davis, E. M., M. T. Oarrett, and T. D. Skinner. 1995. Significance of indicator bacteria changes in an
urban stream. Water Science & Technology 31 (5-6):243-246.
Dee, S.W., and J.C. Fogleman. 1992. Rates of inactivation of waterborne coliphages by monochloramine.
Appl. Environ. Microbiol. 58(9): 3136-41.
De Haas, R.1997. Microfiltration bests conventional treatment. J. AWWA 89(10): 10.
deRegnier, D.P., L. Cole, D.G. Schupp, and S.L. Erlandsen. 1989. Viability ofGiardia cysts suspended in
lake, river, and tap water. Appl. Environ. Microbiol. 55:1223-1229.
Di Giovanni, G.D., F.H. Hashemi, N.J. Shaw, F.A. Abrams, M.W. LeChevallier, and M. Abbaszadegan.
1999. Detection of infectious Cryptosporidiwn parvum oocysts in surface and filter backwash
water samples by immunomagnetic separation and integrated cell culture-PCR. Appl. Environ.
Microbiol. 65(8): 3427-32.
Domek, M. J, M. W. LeChevallier, Cameron. S. C., and G. A. McFeters. 1984. Evidence for the role of
copper in the injury process of coliform bacteria in drinking water. Appl. Environ. Microbiol.
48:289-293.
Dugan, N.R, K.R. Fox, R.J. Miltner, and J.H. Owens. 2001. Control of Cryptosporidium oocysts by
conventional treatment. J. AWWA. 93: 64-76.
Dunn, J., T. Ott, and W. Clark. 1995. Pulsed light treatment of food and packaging. Food Technol. 49: 95.
Dupont, C., M.E. Bougnoux, L. Turner, E. Rouveix, and M. Dorra. 1996. Microbiological findings about
pulmonary cryptosporidiosis in two AIDS patients. J. Clin. Microbiol. 34: 227-229.
DuPont, H.L., C. Chappell, C. Sterling, P. Okhuysen, J. Rose, and W. Jakubowski. 1995. The infectivity
of Cryptosporidium parvum in healthy volunteers. New England J. Med. 332: 855-859.
Du Preez, M., P. Kfir, and P. Coubrough. 1995. Investigation of injury of coliforms after chlorination.
Water Science & Technology 31 (5-6):l 15-118.
Edberg, S.C., E.W. Rice, R.J. Karlin, MJ. Allen. 2000. Escherichia coli: the best biological drinking
water indicator for public health protection, J. Appl. Microbiol. 88. Suppl:106S-l 16S.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-5
December 2005
-------
Edzwald, J.K. and M.B. Kelley. 1998. Control of Cryptosporidium from reservoirs to clarifiers to filters.
Water Sci. Technol. 37(2): 1-8.
Emelko, M.B., Huck P.M, Douglas, I.P, and J. Van Den Oever. 2000. Cryptosporidium and microsphere
removal during low turbidity end-of-run and early breakthrough filtration. Proceedings from
A WWA Water Quality Technology Conference.
Emeiko, M., P. Huck, and R. Slawson. 1999. Design and operational strategies for optimizing
Cryptosporidium removal by filters. Proceedings of 1999 Water Quality Technology Conference.
American Water Works Association, Denver, CO.
Evison, L. M., and E. Tosti. 1980. An appraisal of bacterial indicators of pollution in sea water. Prog
Water Technol. 12:591-599.
Ey, P.L., J.M. Darby, and G. Mayrhofer. 1998. Comparison of tsa417-\\ke variant-specific surface protein
(VSP) genes in Giardia intestinalis and identification of a novel locus in genetic Group II
isolates. Parasitology 117: 445-455.
Payer, R. 1995. Effect of sodium hypochlorite exposure on infectivity of Cryptosporidium parvum
oocysts for neonatal BALB/c mice. Appl. Environ. Microbiol. 61: 844.
Payer, R. 1994. Effect of high temperature on infectivity of Cryptosporidium parvum oocysts in water.
Appl. Environ. Microbiol. 60: 2732-2735.
Payer, R., U.M. Morgan, and S.J. Upton. 2000. Epidemiology of Cryptosporidium: transmission,
detection, and identification. Int. J. Parasitol. 30:1305-1322.
Payer, R., and T. Nerad. 1996. Effects of low temperatures on viability of Cryptosporidium parvum
oocysts. Appl. Environ. Microbiol. 62: 1431-1433.
Payer, R., C.A. Speer, and J.P. Dubey. 1997. General biology of Cryptosporidium. Cryptosporidium spp.
and Cryptosporidiosis. Ed. R. Payer. Boca Raton: CRC Press. 1-41.
Payer, R., C.A. Speer, and J.P. Dubey. 1990. General biology of Cryptosporidium. Cryptosporidiosis of
Man and Animals. Ed. J.B. Dubey, C.A. Speer, and R. Payer. CRC Press. 2-29.
Payer, R., and B.L.P. Ungar. 1986. Cryptosporidium spp. and Cryptosporidiosis. Microbiol. Rev. 50(4):
458-483.
Filice, P.P. 1952. Studies on the cytology and life history of a Giardia from the laboratory rat. U. Calif.
Publ. Zool. 57(2): 53-146.
Finch, G.R., E.K. Black, L. Gyiirek, and M. Belosevic. 1993a. Ozone inactivation of Cryptosporidium
parvum in demand free phosphate buffer determined by in vitro excystation and animal
infectivity. Appl. Environ. Microbiol. 59: 2403.
Finch, G.R., E.K. Black, C.W. Labatiuk, L. Gyiirek, and M. Belosevic. 1993b. Comparison of Giardia
lamblia and Giardia muris cyst inactivation by ozone. Appl. Environ. Microbiol. 59(11):
3674-80.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-6
December 2005
-------
Finch, G.R. and N. Fairbairn. 1991. Comparative inactivation of poliovirus type 3 and MS2 coliphage in
demand-free phosphate buffer by using ozone. Appl. Environ. Microbiol. 57(11): 3121-26.
Finch, G. R., L. L. Gyurelc, L. R. J. Liyanage, and M. Belosevic. 1997. Effect of various disinfection
methods on the inactivation of Cryptosporidium. AWWA Research Foundation and AWWA:
Denver. Distributed to Technical Working Group Meeting June 2, 1997.
Finch, G.R., and M. Belosevic. 1999. Inactivation of C. parvum and G. muris with medium pressure
ultraviolet radiation. Presentation at EPA Workshop at Ultraviolet Disinfection of Drinking
Water, Arlington, VA, April 28, 1999.
Fogel, D., J. Isaac-Renton, R. Guasparini, W. Moorehead, and J. Ongerth. 1993. Removing Giardia and
Cryptosporidium by slow sand filtration. J. AWWA 85(11): 77-82.
Fujioka, R. S., H. H. Hashimoto, E. B, Siwak, and R. H. Young. 1981. Effect of sunlight on survival of
indicator bacteria in seawater. Appl. Environ. Microbiol. 41(3):690-6.
Fujioka, R. S., and O. T. Narikawa. 1982. Effect of sunlight on enumeration of indicator bacteria under
field conditions. Appl. Environ. Microbiol. 44 (2):395-401.
Furness, B.W., M.J. Beach, J.M. Roberts. 2000. Giardiasis surveillance—United States 1992-1997.
MMWR Surveillance Summaries 49.SS07: 1-13.
Gerba., C.P., J.B. Rose, and C.N. Haas. 1996. Sensitive populations: who is at the greatest risk? Int. J.
Food Microbiol. 30(1-2): 113-123.
Gere, A.R. 1997. Microfiltration operating costs. J. AWWA 89(10): 40-49.
Goodrich, J. A., Li, S. Y., and Lykins, Jr., B. W. 1995. Cost and performance of alternative filtration
technologies for small communities. Proceedings of the 1995 A WWA Annual Conference.
Goyal, S. M, and C. P. Gerba. 1979. Comparative adsorption of human enteroviruses, simian rotavirus,
and selected bacteriophages to soils. Appl. Environ. Microbiol. 38:241-247.
Graczyk, T.K., MR. Cranfield, R. Payer, J. Trout, and H.J. Goodale. 1997. Activity ofCryptosporidium
parvum oocysts is retained upon intestinal passage through a migratory water-fowl species
(Canada goose, Branta canadensis). Trop. Med. Int. Health 2(4): 341-347.
Graczyk, T.K., R. Payer, J.M. Trout, E.J. Lewis, C.A. Farley, 1. Sulaiman, and A.A. Lai. 1998. Giardia
sp. cysts and infectious Cryptosporidium parvum oocysts in the feces of migratory Canada geese
(Branta canadensis). Appl. Env. Microbiol. 64(7): 2736-2738.
Gregory, J. 1994. Cryptosporidium in water: Treatment and monitoring methods. Filtr. Sep. 31(3): 283-
289.
Gregory, R., and T.F. Zabel. 1990. Sedimentation and flotation. Water Quality and Treatment: A
Handbook of community water supplies. F.W. Pontius, ed. American Water Works Association.
4th ed. New York: McGraw-Hill. 367-453.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-7
December 2005
-------
Grigsby, P., and J. Calkins. 1980. Inactivation of a natural population of coliform bacteria by sunlight.
Photochemistry & Photobiology 3 1 (3):291-294.
Gyiirek, L.L., H. Li, M. Belosevic, and G.R. Finch. 1999. Ozone inactivation of Cryptosporidium in
phosphate buffer../. Env. Eng. 125(10): 913-923.
Haas, C.N, J.B. Rose, and C.P Gerba. Quantitative microbial risk assessment. New York: John Wiley &
Sons, Inc. 280-319.
Hall, T., Pressdee, J., and E. Carrington.1994. Removal of Cryptosporidium oocysts by water treatment
processes. Foundation for Water Research.
Hall, T. and B. Croll. 1996. The UK approach to Cryptosporidium control in water treatment.
Proceedings A WWA Water Quality Technology Conference. Boston: AWWA.
Hancock, C.M., J.B. Rose, G.J. Vasconcelos, S.I. Harris, P.T. Klonicki, and G.D. Sturbaum. 1999. Using
microbiota to predict Giardia and Cryptosporidium occurrence in ground water. Proceedings of
the A WWA Water Quality Technology Conference.
Hancock, C.M., J.B. Rose, G.J. Vasconcelos, S.I. Harris, P.T. Klonicki, and G.D. Sturbaum. 1998a.
Correlation of Cryptosporidium and Giardia occurrence in groundwater with surface water
indicators. American Water Works, Annual Conference and Exposition, 1998. Abstract.
Hancock, CM., J.B. Rose, and M. Callahan. 1998b. Cryptosporidium and Giardia in U.S. groundwater.
Hansen, J.S., and J.E. Ongerth. 1991 . Effects of time and watershed characteristics on the concentration
of Cryptosporidium oocysts in river water. Appl. Environ. Microbiol. 57(10): 2790-2795.
Harp, J.A., R. Payer, B.A. Pesch, and G.J. Jackson. 1996. Effect of pasteurization on infectivity of
Cryptosporidium parvum oocysts in water and milk. Appl. Environ. Microbiol. 62: 2866-2868.
Harrington, G., H. Chen, A. Harris, I. Xagoraraki, D. Battigelli, and J. Standridge. 2001. Removal of
emerging waterborne pathogens. American Water Works Association Research Foundation,
Denver, CO.
Harris, G.D., V.D. Adams, D.L. Sorensen and M.S. Curtis. 1987. Ultraviolet inactivation of selected
bacteria and virus with photoreactivation of the bacteria. Wat. Res. 21(6): 687-692.
Harter, T., S. Wagner, and E.R. Atwill. 2000. Colloid transport and filtration of Cryptosporidium parvum
in sandy soils and aquifer sediments. Env. Sci. Tech. 34(1): 62-70.
Harvey, R.W., N.E. Kinner, A. Bunn, D. MacDonald, and D. Metge. 1995. Transport behavior of
groundwater protozoa and protozoan-sized microspheres in sandy aquifer sediments. Appl.
Environ. Microbiol. 61 : 209-217.
Heijbel, H., K. Slaine, B. Seigel, P. Wall, S.J.N. McNabb, W. Gibbons, and G.R. Istre. 1987. Outbreak of
diarrhea in a day care center with spread to household members: the role of Cryptosporidium.
Pediatr. Infect. Dis. J. 6: 532-535.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-8
December 2005
-------
Heisz, M.. 1997. In vitro survival of Cryptosporidium oocysts in natural waters. International Symposium
on Waterborne Cryptosporidium. Newport Beach, March 1997.
Hibler, C.P., C.M. Hancock, L.M. Perger, J.G. Wegrzyn, K.D. Swabby. 1987. Inactivation ofGiardia
cysts with chlorine at 0.5°C to 5.0°C. American Water Works Association Research Foundation.
Hoffman, R.M., J.H. Standridge, A.F. Prieve, J.C. Cucunato, and M. Bernhardt. 1995. The use of flow
cytometry for detection of Cryptosporidium and Giardia in water samples. Proc. Water Qua!.
Techno!. Conf. New Orleans. 883-903.
Hopkins, R.M., B.P. Meloni, D.M. Groth, J.D. Wetherall, J.A. Reynoldson, and R.C. Thompson. 1997.
Ribosomal RNA sequencing reveals differences between the genotypes ofGiardia isolates
recovered from humans and dogs living in the same locality. J. Parasitol. 83(1): 44-51.
Horwitz, M.S. 1996. Adenoviruses. Fields Virology. Ed. B.N. Fields, D.M. Knipe, P.M. Howley, and
R.M. Chanock et al. 3rd ed. Philadelphia: Lippincott-Raven Publ. 2149-2171. (Cited in USEPA
1999b)
Hoxie, N.J., J.P. Davis,J.M. Vergeront, R.D. Nashold, and K.A. Blair. 1997. Cryptosporidiosis-associated
mortality following a massive waterborae outbreak in Milwaukee, Wisconsin. Amer. J. Publ.
Health 87(12): 2032-2035.
Huck, P.M., Coffey, B.M., O'Melia, C.R., and M.B. Emelko. 2000. Removal of Cryptosporidium by
filtration under conditions of process challenge. Proceedings from AWWA Water Quality
Technology Conference.
Hurst, C. J., C. P. Gerba, J. C. Lance, and R. C. Rice. 1980a. Survival of enteroviruses in rapid-infiltration
basins during the land application of wastewater. Appl. Environ. Microbiol. 40(2): 192-200.
Hurst, C. J., C. P. Gerba, and 1. Cech. 1980b. Effects of environmental variables and soil characteristics
on virus survival in soil. Appl. Environ. Microbiol. 40 (6): 1067-1079.
Hurst, C.J. 1997. Modeling the fate of microorganisms in water, wastewater, and soil. Manual of
Environmental Microbiology. Ed. C.J. Hurst, G.R. Knudsen, M.J. Mclnerney, L.D. Stetzenback,
and M.V. Walter. Ch. 22. Washington, DC: American Society for Microbiology.
ICTV (International Committee on Taxonomy of Viruses). 1995. Sixth report of the International
Committee on Taxonomy of Viruses. F.A. Murphy, C.M. Fauquet, D.H.L. Bishop, S.A. Ghabrial,
A.W. Jarvis, G.P. Martelli, M.A. Mayo, and M.D. Summers, eds. Springer-Verlag, Vienna,
Austria and New York, NY. (Cited in USEPA 1999c)
Jacangelo, J.G., S.S. Adham, and J.-M. Laine. 1995. Mechanism of Cryptosporidium, Giardia, and MS2
virus removal by MF and UF. J. A WWA 87(9): 107-121. (Cited in USEPA 1999b)
Jakubowski, W. 1990. The control ofGiardia in water supplies. In: Giardiasis. Ed. E.A. Meyer. New
York: Elsevier Science. 335-353.
Jarroll, E.L. 1988. Effect of disinfectants on Giardia cysts. CRC Crit. Rev. Env. Control 18: 1-28.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-9
December 2005
-------
Jarroll, E.L., J.C. Hoff, and E.A. Meyer. 1984. Resistance of cysts to disinfection agents. In: Giardia anrf
Giardiasis. S.L. Erlandsen and E.A. Meyer, eds. Plenum: New York. 311-328.
Jarroll, E.L., A.K. Bingham, and E.A. Meyer. 1981. Effect of chlorine on Giardia lamblia cyst viability.
Appl. Environ. Microbiol. 41(2): 483-487.
Johnson, D.C. C.E. Enriquez, I.L. Pepper, T.L. Davis, C.P. Gerba, and J.B. Rose. 1997. Survival of
Giardia, Cryptosporidium, poliovirus, and Salmonella in marine waters. Wat. Sci. Tech. 35(11-
12):261-268,
Juranek, D. 1995. Cryptosporidiosis: Sources of infection and guidelines for prevention. Clin. Infect. Dis.
21(Suppl. 1):S57-S61.
Kansas State University. 1997. Basic biology of Cryptosporidium. Website of Parasitology Laboratory,
Division of Biology, Kansas State University. 5 April
2000.
Kapikian, A.Z., M.K. Estes, and R.M. Chanock. 1996. Norwalk group of viruses. Fields Virology. Ed.
B.N. Fields, D.M. Knipe, P.M. Howley, and R.M. Chanock et al. 3rd ed. Philadelphia:
Lippincott-Raven Publ. 783-810. (Cited in USEPA 1999b)
Karetnyi, Y.V., M.J. Gilchrist, and S.J. Naides. 1999. Hepatitis E virus infection prevalence among
selected populations in Iowa. J. Clin. Virol. 14(1): 51-55.
Kayed, D., and J.B. Rose. 1987. Development of a method for the detection of Cryptosporidium in
sewage sludge. Abstr. Ann. Meeting American Society of Microbiology; 296.
Kelley, M.B., P.K. Warrier, J.K. Brokaw, K.L. Barrett, and S.J. Komisar. 1995. A study of two U.S.
Army installation drinking water sources and treatment systems for the removal of Giardia and
Cryptosporidium. Proc. Water Qual. Technol. Conf. New Orleans. 2197-2230.
Keswick, B. H., C. P. Gerba, S. L. Secor, and I. Cech. 1982. Survival of enteric viruses and indicator
bacteria in groundwater. Journal of Environmental Science & Health Part A-Environmental
Science & Engineering 17 (6):903-9I2.
Keswick, B.H., T.K. Satterwhite, P.C. Johnson, H.L. DuPont, S.L. Secor, J.A. Bitsura, W.G. Gary, and
J.C. Hoff. 1985. Inactivation of Norwalk virus in drinking water by chlorine. Appl. Environ.
Microbiol. 50(2): 261-264. (Cited in USEPA 1999b, 1999c)
Khramtsov, N.V., K.M. Woods, M.V. Nesterenko, C.C. Dykstra, S.J. Upton. 1997. Virus-like, double-
stranded RNAs in the parasitic protozoan Cryptosporidium parvum. Mol. Microbiol. 26(2): 289-
300.
Korich, D.G., J.R. Mead, J.S. Madore, N.A. Sinclair, and C.R. Sterling. 1990. Effects of ozone, chlorine
dioxide, chlorine, and monochloramine on Cryptosporidium oocyst viability. Appl. Env.
Microbiol. 56: 1423-1428.
Kramer, M.H., Herwaldt, B.L., Calderon, R.L., and Juranek, D.D. 1996. Surveillance for waterborne-
disease outbreaks-United States, 1993-1994.MMWR 45(SS-l):l-33.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-10
December 2005
-------
Kurtz, J.B., and T.W. Lee. 1987. Astroviruses: human and animal. Novel Diarrhea Viruses. Ed. G. Bock
and J. Whelan. Ciba Foundation Symposium 128. Chicester: Wiley. 92-107. (Cited in USEPA
1999b)
Labatiuk, C.W., F.W. Schaefer, III, G.R. Finch, and M. Belosevic. 1991. Comparison of animal
infectivity, excystation and fluorogenic dye as measure of Giardia tnuris cyst inactivation by
ozone. Appl. Environ. Microbiol. 57: 3187-3192.
LeChevallier, M.W. 2001. Personal communication, 10/19/2001.
LeChevalHer, M.W., H. Arora, D. Battigelli, and M. Abbaszadegan. 1996. Chlorine dioxide for control of
Cryptosporidiwn and disinfection by-products. American Water Works Association, Water
Quality Technology Conference, Boston, MA, November 1996. Boston: AWWA, 1996. 17-21.
LeChevallier, M.W., and W.D. Norton. 1995. Giardia and Cryptosporidium in raw and finished water. J.
AWWA 87(9):54-68.
LeChevallier, M. W., and W.D. Norton. 1992. Examining relationships between particle counts and
Giardia, Cryptosporidium and turbidity. J. AWWA 85(12): 54-60.
LeChevallier, M.W., W.D. Norton, and R.G. Lee. 1991. Giardia and Cryptosporidium spp. in filtered
drinking water supplies. Appl. Environ. Microbiol. 57(9): 2617-2621.
Lee, S.H., D.A. Levy, G.F. Craun, M.J. Beach, and R.L. Calderon. 2002. Surveillance for waterborne
disease outbreaks, United States, 1999-2000. Morbidity and Mortality Weekly Report,
Surveillance Summaries 51 (SS08): 1 -28.
Levine, N.D. 1985. Phylum II. Apicomplexa. Illustrated guide to the protozoa. Ed. J.J. Lee, S.H. Hutner,
and E. C. Bovee. Lawrence: Society for Protozoology. 322-374.
Levy, D.A., M.S. Bens, G.F. Craun, R.L. Calderon, and B.L. Herwaldt. 1998. Surveillance for waterborne
disease outbreaks - United States, 1995-1996. MMWR 47(SS-5):l-34.
Li, H., G.R. Finch, N. Neumann, and M. Belosevic. 1998. Inactivation of Cryptosporidium by chlorine
dioxide at 1°C. Proceedings of the 1998 AWWA Water Quality Technology Conference. AWWA
QTC98234.
Linden, Karl. 2000. Personal communication.
Lisle, J.T., and J.B. Rose. 1995. Cryptosporidium contamination of water in the U.S. and UK: a mini-
review. J. Water SRT—Aqua 44(3): 103-117.
Lozier, J.C., G. Jones, and W. Bellamy. 1997. Integrated membrane treatment in Alaska. J. AWWA
89(10): 50-64.
Ma., J.F., T.M. Straub, I.L. Pepper, and C.P. Gerba. 1994. Cell culture and PCR determination of
poliovirus inactivation by disinfectants. Appl. Environ. Microbiol. 60(11): 4203-06.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-11
December 2005
-------
Madore, M.S., J.B. Rose, C.P. Gerba, M.J. Arrowood, and C.R. Sterling. 1987. Occurrence of
Cryptosporidium oocysts in sewage effluents and selected surface waters. J. Parasit. 73(4): 702-705.
Maier, A., D. Tougianidou, A. Wiedenmann, K. Botzenhart. 1995. Detection of poliovirus by cell culture
and by PCR after UV disinfection. Wat. Sci. 7V?cfc. 31 (5-6): 141-145.
Malley, J.P., A.B. Margolin, G.A. Snicer, and S.P. Hogan. 1999. UV inactivation of viruses in natural
waters (summary). American Water Works Association Research Foundation: Denver. Website
visited on 10-23-02.
Marzouk, Y., S. M. Goyal, and C. P. Gerba. 1979. Prevalence of enteroviruses in ground water of Israel.
Ground Water 17 (5):487-491.
Matson, D.O., M.K. Estes, R.I. Glass, A.V. Bartlett, M. Penaranda, E. Calomeni, T. Tanaka, S. Nakata,
and S. Chiba. 1989. Human calicivirus-associated diarrhea in children attending day care centers.
J. Infect. Dis. 159(1): 71-78. (Cited in USEPA 1999b)
Mawdsley, J.L., A.E. Brooks, and R.J. Merry. 1996. Movement of the protozoan pathogen
Cryptosporidium parvum through three contrasting soil types. Biol. Fertil. Soils 21(1-2): 30-36.
McLauchlin, J., S. Pedraza-Diaz, C. Amar-Hoetzeneder, and G.L. Nichols. 1999. Genetic characterization
of Cryptosporidium strains from 218 patients with diarrhea diagnosed as having sporadic
Cryptosporidiosis. J. Clinical Microbiol. 37(10): 3153-3158.
McTigue, N. E., M. LeChevallier, H. Arora, J. Clancy. 1998. National assessment of particle removal by
filtration. Denver: American Water Works Association Research Foundation and AWWA. 266 p.
Medema, G., F. Schets, P. Teunis, and A. Havelaar. 1998. Sedimentation of free and attached
Cryptosporidium oocysts and Giardia cysts in water. Appl. Environ. Microbiol. 64(1): 4460-
4466.
Melnick. J.L. 1996. Enteroviruses: Poliovirus, coxsackieviruses, echoviruses, and newer enteroviruses.
Fields Virology. Ed. B.N. Fields, D.M. Knipe, and P.M. Howley. Philadelphia: Lippincott Raven
Publishers. (Cited in USEPA 1999c)
Melo Cristino, J.A.G., M. Isabel, P. Carvalho, and M. Jose Salgado. 1988. An outbreak of
cryptosporidiosis in a hospital day-care centre. Epidem. Inf. 101: 355-359.
Meloni, B. P., AJ. Lymbery, and R.C.A. Thompson. 1995. Genetic characterization of isolates of Giardia
duodenalis by enzyme electrophoresis: Implication for reproductive biology, population
structure, taxonomy and epidemiology. J. Parasitol. 81: 368-383.
Meng, Q.S. and C.P. Gerba. 1996. Comparative inactivation of enteric adenovirus, poliovirus and
coliphages by ultraviolet irradiation. Wat. Res. 30(11): 2665-2668.
Messner M., 2000. Precision, Accuracy, and Detection Capability of the ICR Method for Measuring
Cryptosporidium and Giardia in Water - Results of the ICR Laboratory Spiking Study, in the
1999 Joint Statistical Meetings, Section on Statistics and the Environment. Alexandria, VA.
American Statistical Assocation.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-12
December 2005
-------
Metcalf, T. G., J. L. Melnick, and M. K. Estes. 1995. Environmental virology: from detection of virus in
sewage and water by isolation to identification by molecular biology—a trip of over 50 years.
Annual Review of Microbiology 49:461-487.
Meyer, E.A., ed. 1990. Giardiasis. Human Parasitic Diseases. Vol. 3. Amsterdam: Elsevier.
Meyers, B.R. 1989. Infectious diseases in the elderly: an overview. Geriatrics 44(Supplement A):4-6.
August.
Mifsud, A.J., D. Bell, and M.S. Shafi. 1994. Respiratory cryptosporidiosis as a presenting feature of
AIDS. Jour. Infect. 28(2): 227-229.
Modlin, J.F. 1997. Enteroviruses: Coxsackie, echoviruses, and newer enteroviruses. Principles and
Practice of Pediatric Infectious Diseases. Ed. S.S. Long, et al. New York: Churchill Livingstone.
(Cited in USEPA 1999c)
Mofidi, A. H. Baribeau, J. Green. 1999. Inactivation of Cryptosporidium parvum with polychromatic UV
systems. Presented at AWWA Water Quality Technology Conference, UV Special Topic Session,
Tampa Bay. Denver: AWWA,Nov 1-4, 1999.
Montgomery, J.H. 1985. Water treatment principles and design. John Wiley & Sons Press. James M.
Montgomery, Consulting Engineers, Inc. (Cited in USEPA 1999a, 1999b)
Moore, A.C., B.L. Herwaldt, G.F. Craun, R.L. Calderon, A.K. Highsmith, and D.D. Juranek. 1993.
Surveillance for waterborae disease outbreaks - United States, 1991-1992. MMWR. 42(SS-5):1-
22 (November 19).
Morgan, U.M., and R.C.A. Thompson. 1998. PCR detection of Cryptosporidium: the way forward?
Parasitology Today 14(6): 241-245.
Morgan, U.M., L. Xiao, R. Payer, A.A. Lai, R.C. Andrew Thompson. 1999. Variation in
Cryptosporidium: towards ataxonomic revision of the genus. Int. J. Parasitol. 29: 1733-1751.
Morgan, U.M., R. Weber, L. Xiao, I. Sulaiman, R. C. Andrew Thompson, W. Ndtritu, A. Lai, A. Moore,
and P. Deplazes. 2000a. Molecular characterization of Cryptosporidium isolates obtained from
human immunodeficiency virus-infected individuals living in Switzerland, Kenya, and the United
States./. Clin. Microbiol. 38(3): 1180-1183.
Morgan, U.M., L. Xiao, B.D. Hill, P. O'Donoghue, J. Limor, A. Lai, and R.C. Thompson. 2000b.
Detection of the Cryptosporidium parvum "human" genotype in a dugong (Dugong dugong). J.
Parasitol. 86(6): 1352-1354.
Najm, I., M. Aieta, J. Oppenheimer, and B. Gallagher. 1998. Impact of turbidity on the inactivation of
Giardia cysts with ozone. Water Supply 16: 419-423.
Neumann, N.F., L.L. Gyiir^k, L. Gammie, G.R. Finch, and M. Belosevic. 2000. Comparison of animal
infect! vity and nucleic acid staining for assessment of Cryptosporidium parvum viability in water.
Appl. Environ. Microbiol. 66(1): 406-412.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-13
December 2005
-------
Nieminski, B.C. 1994. Giardia and Cryptosporidium cyst removal through direct filtration and
conventional treatment. In: Proceedings oftheAWWA Annual Conference, New York, NY.
Denver: AWWA. (cited in Patania et al. 1995)
Nieminski, E.G., and W.D. Bellamy. 2000. Application of surrogate measures to improve treatment
performance. Project 363. Denver: AWWARF and AWWA. 24-58.
Nieminski, E.G., and J. E. Ongerth. 1995. Removing Giardia and Cryptosporidium by conventional
treatment and direct filtration. J. AWWA 87(9): 96-106.
O'Donoghue, P.J. 1995. Cryptosporidium and cryptosporidiosis in man and animals. Int. J. Parasitol. 25:
39-195.
Okhuysen, P., C.L. Chappell, J.H. Crabb, C.R. Sterling, and H.L. DuPont. 1999. Virulence of three
distinct Cryptosporidium parvum isolates for healthy adults. The Journal of Infectious Diseases
180: 1275-81.
Olson, M.E., J. Goh, M. Phillips, N. Guselle, and T.A. McAllister. 1999. Giardia cyst and
Cryptosporidium oocyst survival in water, soil, and cattle feces. J. Environ. Qual, 28(6): 1991-
1996.
Ongerth, J.E. and J.P. Pecoraro. 1995. Removing Cryptosporidium using multimedia filters. J. AWWA
87(12): 83-89.
Ongerth, J.E, and P.E. Mutton. 2001. Testing of diatomaceous earth filtration for removal of
Cryptosporidium oocysts J. AWWA. 93: 54-63.
Oppenheimer, J.A., E.M. Aieta, J.G. Jacangelo, and I. Najm. 1997. CT requirements for disinfection of
Cryptosporidium in natural waters. Proceedings A WWA Water Quality Technology Conference,
Denver, November. Denver: AWWA.
Overturf, Gary D. 1994. Editorial response: Endemic giardiasis in the United States— Role of the day-
care. Clinical Infectious Diseases 18: 764-5.
Owens, J.H., R.J. Miltner, F.W. Schaefer, III, and E.W. Rice. 1994. Pilot-scale ozone inactivation of
Cryptosporidium and Giardia. Proc. of Water Qual. Techno!. Conf. San Francisco. 1319-1324.
Paintlia, Ajaib S., Manjeet K. Paintlia, Ramesh C. Mahajan, Anuradha Chakraborti, and Nirmal K.
Ganguly. 1999. A DNA-based probe for differentiation of Giardia lamblia group A and B from
northern India. Clinical Infectious Diseases 28: 1178-80.
Parker, J., and H. Smith. 1993. Destruction of oocysts of Cryptosporidium parvum by sand and chlorine.
Water Res. 27:729.
Patania, Nancy; Mazounie, Pierre; Bernazeau, Francois; Maclean, Greg; Alia, Pierre. 1999. Removal of
Cryptosporidium by contact filtration: The Sydney experience. AWWA Conference Proceedings.
Patania, N.L., J.G. Jacangelo, L. Cummings, A. Wilczak, K. Riley, and J. Oppenheimer. 1995.
Optimization of filtration for cyst removal. Denver: AWWA Research Foundation and AWWA.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-14
December 2005
-------
Pavlasek, I. 1984. Effect of disinfectants in infectiousness of oocysts of Cryptosporidium sp. Cs. Epidem.
Mikrobiol. Immunol. 33: 97.
Payment, P. 1999. Poor efficacy of residual chlorine disinfectant in drinking water to inactivate
waterborne pathogens in distribution systems. Can. J. Microbiol. 45(8): 709-715.
Payment, P., and R. Armon. 1989. Virus removal by drinking water treatment process. Crit. Rev. Env.
Control 19(1): 15-32. (Cited in USEPA 1999b)
Pedraza-Diaz, S., C. Amar, A.M. Iversen, P.J. Stanley, and J. McLauchlin. 2001. Unusual
Cryptosporidium species recovered from human faeces: first description of Cryptosporidium felis
and Cryptosporidium 'dog type' from patients in England. J. Med. Microbiol. 50(3):293-296.
Peelers, J.W., E. Ares Mazas, W.J. Masschelein, I. Villacorta, and E. Debacker. 1989. Effect of
disinfection of drinking water with ozone or chlorine dioxide on survival of Cryptosporidium
parvum oocysts. Appi Environ. Microbiol 55: 1519.
Peng, M.M., L. Xiao, A.R. Freeman, M.J. Arrowood, A.A. Escalante, A.C.Weltman, C.S.L.Ong, W.R.
Mac Kenzie, A.A. Lai, and C.B. Beard. 1997. Genetic polymorphism among Cryptosporidium
parvum isolates: Evidence of two distinct human transmission cycles. Emerging Infectious
Diseases 3(4): 9 Aug. 1999. .
Pennsylvania American Water Company. Capital Division. 1989. Results of superpulsator pilot trial,
March 22, 1989 to May 18, 1989. 23 pp.
Pesaro, F. I. Sorg, and A. Metzler. 1995. In situ inactivation of animal viruses and acoliphage in
nonaerated liquid and semiliquid wastes. Appl. Env. Microbiol. 61(1): 92-97.
Peterson, D.A., T.R. Hurley, J.C. Hoff, and L.G. Wolfe. 1983. Effect of chlorine treatment on infectivity
of hepatitis A virus. Appl. Env. Microbiol. 45(1): 223-227. (Cited in USEPA 1999c)
Pieniazek, N.J., F.J. Bornay-Llinares, S.B. Slemenda, AJ. da Silva, I.N.S. Moura, M.J. Arrowood, O.
Ditrich, and D.G. Addiss. 1999. New Cryptosporidium genotypes in HIV-infected persons.
Emerg. Infect. Dis. 5(3) May-June.
Piummer, J.D., J.K.. Edzwald, and M.B. Kelley. 1995. Removing Cryptosporidium by dissolved-air
flotation. J. AWWA 87(9): 85-95.
Quinn, C.M., and W.B. Betts. 1993. Longer term viability status of chlorine-treated Cryptosporidium
oocysts in tap water. Biomed. Lett. 48: 315.
Ransome, M.E., T.N. Whitmore, and E.G. Carrington. 1993. Effect of disinfectants on the viability of
Cryptosporidium parvum oocysts. Water Supply 11: 75.
Rennecker, J. L., B. J. Marifias, J. H. Owens, and E. W. Rice. 1999. Inactivation of Cryptosporidium
parvum oocysts with ozone. Wat. Res. 33(11): 2491-88.
Ribeiro, C.D., and S.R. Palmer. 1986. Family outbreak of cryptosporidiosis. Br. Med. J. 292: 377.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-15
December 2005
-------
Rice, E. W., and J. C. Hoff. 1981. Inactivation of Giardia lamblia cysts by ultraviolet irradiation. Appl.
Environ. Microbiol 42(9): 546-7.
Robertson, L.J., A.T. Campbell, and H.V. Smith. 1998. Viability of Cryptosporidiumparvum oocysts:
assessment of the dye permeability assay. Appl. Environ. Microbiol. 64(9): 3544-3545.
Robertson, L.J., A.T. Campbell, and H.V. Smith. 1992. Survival of Cryptosporidiumparvum oocysts
under various environmental pressures. Appl. Environ. Microbiol. 58: 3494-3500.
Robertson, J.B., and S.C. Edberg. 1997. Natural protection of spring and well drinking water against
surface microbial contamination. I. Hydrogeological parameters. Crit. Rev. Microbiol. 23(2): 143-
78.
Rodriguez-Hernandez, J., A. Canut-Blasco, and A.M. Martin-Sanchez. 1996. Seasonal prevalences of
Cryptosporidium and Giardia infections in children attending day care centres in Salamanca
(Spain) studied for a period of 15 months. Eur. J. Epidemiol. 12(3): 291-295.
Rose, J.B. 1997. Environmental ecology of Cryptosporidium and public health implications. Annual Rev.
Public Health 18:135-161.
Rose, J.B., L. J. Lisle, and M. LeChevallier. 1997. Waterborne cryptosporidiosis: Incidence, outbreaks,
and treatment strategies. Cryptosporidium and Cryptosporidiosis, Ed. R. Payer. Chapter 4. Boca
Raton: CRC Press.
Roessler, N. 1998. Control of Cryptosporidium in bottled water using cartridge filtration systems.
Filtration & Separation. 35(l):37-39.
Rubin, A.J., D.P. Evers, C.M. Eyman, E.L. Jarroll. 1989. Inactivation of gerbil-cultured Giardia lamblia
cysts by free chlorine. Appl. Environ. Microbiol. 55(10):2592-2594.
Sachs, L. 1984. Applied statistics: A handbook of techniques. 2nd ed. New York: Springer-Verlag.
Sagik, B. P., B. E. Moore, and C. A. Sorber, eds. 1978. Infectious disease potential of land application of
wastewater. Vol. I, State of knowledge in land treatment qfwastewater. Washington, DC: U. S.
Environmental Protection Agency.
Schaub, S.A., Hargett, H.T., Schmidt, M.O., and Burrows, W.D. 1993. Reverse osmosis water
purification unit: Efficacy of cartridge filters for removal of bacteria and protozoan cysts when
RO elements are bypassed. US Army Biomedical Research and Development Laboratory, Fort
Detrick, Frederick, Maryland.
Schuler, P.P., and Ghosh, M.M. 1990. Diatomaceous earth filtration of cysts and other particulates using
chemical additives. J. AWWA December.
Schuler, P.P. and Ghosh, M.M. 1991. Slow sand filtration of cysts and other particulates. Proc. of AWWA
Annual Conference. Philadelphia, PA. June 23-27, 1991. 235-252.
SDWC (Safe Drinking Water Committee). 1980. Drinking Water and Health. Vol. 2. Washington, DC:
National Academy Press. (Cited in USEPA 1999b)
Occurrence and Exposure Assessment
for the LT2ESWTR
9-16
December 2005
-------
SDWC. 1977. Drinking Water and Health. Vol. 3. Washington, DC: National Academy Press. (Cited in
USEPA 1999b)
Seyde, V., D.L. Clark, E. Akiyoshi, A. Kalinsky, and C. Spangenburg. 1998. Nanofiltration membrane
systems for control of microbials. Abstracts of the 1998 General Meeting of the American Society
for Microbiology. 446.
Shaw, M. K., A. G. Marr, and J. L. Ingraham. 1971. Determination of minimal temperature for growth of
Escherichia colt, J. Bacterial. 105:683-684.
Sherwood, D., K.W. Angus, D.R. Snodgrass, and S. Tzipori. 1982. Experimental cryptosporidiosis in
laboratory mice. Infect. Immun. 38: 471-475.
Shin, G., K. G. Linden, G. Faubert, M. Arrowood, and M. D. Sobsey. 2000a. Inactivation of
Cryptosporidium parvum oocysts and Giardia lamblia cysts by monochromatic UV radiation.
Proc. IWA First International Congress, Health Related Water Microbiology. Paris, July 1 -4,
2000.
Shin, G., K..G. Linden, and M.D. Sobsey. 2000b. Comparative inactivation of Cryptosporidium parvum
oocysts and coliphage MS-2 by monochromatic UV radiation. Proc., Disinfection 2000, WEF
Specialty Conference. New Orleans, March 15-17, 2000.
Shin, G., K.G. Linden, and M.D. Sobsey. 1999. Inactivation of Cryptosporidium oocysts with low
pressure UV radiation. Proc., Water Quality Technology Conference. Tampa Bay: AWWA, Nov
1-3, 1999.
Sil, A.K,. P. Das, S. Bhattacharyya, S. Ghosh, D.J. Chattopadhyay. 1998. Cloning of ribosomal RNA
genes from an Indian isolate of Giardia lamblia and the use of intergenic nontranscribing spacer
regions in the differentiation of Giardia from other enteric pathogens. J. Biosci. 23(5): 557-564.
Smith, H.V., A.L. Smith, R.W.A. Girdwood, and E.G. Carrington. 1990. The effect of free chlorine on the
viability of Cryptosporidium oocysts isolated from human feces. Cryptosporidium in Water
Supplies. Ed. J. Badenoch. London: Her Majesty's Stationery Office. 185.
Sobsey, M.D. 1989. Inactivation of health-related microorganisms in water by disinfection processes.
Wat. Set. Tech2\(3): 179-195. (Cited in USEPA 1999b)
Sobsey, M.D., T. Fuji, and R.M. Hall. 1991. Inactivation of cell-associated and dispersed hepatitis A virus
in water. J. AWWA 83: 64-67.
Spano, Furio, L. Putignani, A. Crisanti, P. Sallicandro, U.M. Morgan, S.M. LeBalanco, L. Tchack, S.
Tzipori, and G. Widmer. 1998. Multilocus genotypic analysis of Cryptosporidium parvum
isolates from different hosts and geographical origins. Journal of Clinical Microbiology Nov.:
3255-3259.
Sreter, T., and Z. Szell. 1998. The low sedimentation speed of Cryptosporidium oocysts: a further
explanation for waterbome outbreaks. J. Protozool. Res. 8: 58-63.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-17
December 2005
-------
Sulaiman, Irshad M, L. Xiao, C. Yang, L. Escalante, A. Moore, C.B. Beard, M.J. Arrowood, and A.A.
Lai. 1998. Differentiating human from animal isolates of Cryptosporidium parvum CDC and
Prevention, Atlanta, Georgia, USA. Emerging Infectious Diseases 4(4) Oct-Dec.
Sundermann, C.A., D.S. Lindsay, and B.L. Blagburn. 1987. Evaluation of disinfectants for ability to kill
avian Cryptosporidium oocysts. Comp. Anim. Pract. 36.
Swabby-Cahill, K.D., G.W. Clark, and A.R. Cahill. 1996. Buoyant qualities of Cryptosporidium parvum
oocysts. AWWA Water Quality Technology Conference. Boston: AWWA.
Tangermann, R., H.S. Gordon, P. Wiesner, and L. Kreckman. 1991. An outbreak of cryptosporidiosis in a
day-care center in Georgia. Am. J. Epidemiol. 133: 471-476.
Thompson, R.C., and A.J. Lymbery. 1996. Genetic variability in parasites and host-parasite interactions.
Parasitol 112: (Suppl)S7-S22.
Tierney, J. T., R. Sullivan, and E. P. Larkin. 1977. Persistence of poliovirus I in soil and on vegetables
grown in soil previously flooded with inoculated sewage sludge or effluent. Appl. Environ.
Microbiol. 33:109-113.
Timms, S., J.S. Slade, and C.R. Fricker. 1995. Removal of Cryptosporidium by slow sand filtration. Wat.
Sci. Technol. 31(5-6): 81-84.
Toranzos, G.A., and G.A. McFeters. 1997. Detection of indicator microorganisms in environmental
freshwaters and drinking waters. Manual of Environmental Biology. Ed. Christon J. Hurst et al.
Washington, DC: ASM Press. 184-184.
Tzipori, S. 1985. Cryptosporidium: Notes on epidemiology and pathogenesis. Parasitol. Today 1: 159-
165.
Tzipori, S. 1983. Cryptosporidiosis in animals and humans. Microbiol. Rev. 47: 84-96.
Tzipori, S. and J.K. Griffiths. 1998. Natural history and biology of Cryptosporidium parvum. Adv.
Parasitol. 40: 5-36.
Tzipori, S., M. Smith, C. Birch, G. Barnes, R. Bishop. 1983. Cryptosporidiosis in hospital patients with
gastroenteritis. Am. J. Trap. Med. Hyg. 32(5): 931-934.
United Network of Organ Sharing. 2000. .
U.S. Census Bureau. 200 la. Statistical abstract of the United States: 200 J. Table 194. Cancer—estimated
new cases, 2001, and survival rates, 1980-82 to 1989-1996.
U.S. Census Bureau 2001b. Statistical Abstract of the United States: 2001. Tables 11 and 13. Resident
population by sex and age group: 1990 and 2000, and Resident population projections by sex and
age: 2005 to 2050.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-18
December 2005
-------
U.S. Court of Appeals, DC Circuit. 2000. Chlorine Chemistry Council et al. v. Environmental Protection
Agency. No. 98-1627. March 31, 2000. http://pacer.cadc.uscourts.gov/common/opinions/200003/
98-1627a.txt Website visited August 14,2000.
USEPA. 2003. National primary drinking water regulations: Stage 2 disinfectants/disinfection byproducts
rule. Draft. April 28, 2003.
USEPA. 2002. National primary drinking water regulations: Long term 1 enhanced surface water
treatment rule; final rule. Federal Register 67(9): 1811-1844. January 14, 2002.
USEPA. 200la. National primary drinking water regulations: Filter backwash recycling rule; Final rule.
Federal Register 66(111):31085-31105. June 8, 2001.
USEPA. 200Ib. Draft drinking water baseline handbook, second edition. Prepared by International
Consultants, Fnc. Contract 68-C6-0039. March 17, 2000.
USEPA. 2000a. Removal of the maximum contaminant level goal for chloroform from the National
primary drinking water regulations. Federal Register 65(104):34404-34405. May 30,2000.
USEPA. 2000b. National primary drinking water regulations: Ground water rule; Proposed rules. Federal
Register 65(91): 30194-30274. May 10, 2000.
USEPA. 2000c. Stage 2 microbial and disinfection byproducts federal advisory committee agreement in
principle. Federal Register 65(25l):&30\ 5-83024. December 29, 2000.
USEPA. 2000d. Occurrence assessment for the long-term 1 enhanced surface water treatment and filter
backwash rule. Prepared for Office of Ground Water and Drinking Water by Science
Applications International Corporation. McLean, VA.
USEPA. 2000e. ICR data analysis plan.
USEPA. 2000f. Technical working group for LT2ESWTR rule options toolbox.
USEPA. 2000g. SDWISDatabase, September 2000 SDWIS freeze data, adjusted for population errors.
USEPA. 1999a. Cryptosporidium spp. systematics and waterborne challenges in public health. Summary
Report. Prepared for Office of Water, Health and Ecological Criteria Division, by ISSI
Consulting Group Silver Spring, MD, under Contract No. 68-C98-195. April 29-30.
USEPA. 1999b. Drinking water criteria document for viruses: An addendum. Prepared for Health and
Ecological Criteria Division, Office of Science and Technology by ISSI, Inc., Silver Spring, MD.
Final Draft. EPA 822-R-98-042. January 15.
USEPA. 1999c. Drinking water criteria document for enteroviruses and hepatitis A: An addendum.
Prepared for Health and Ecological Criteria Division by Nena Nwachuku, Health and Ecological
Criteria Division, Office of Science and Technology. Final Draft. EPA 822-R-98-043. January
15.
USEPA. 1999d. Method 1622: Cryptosporidium in water by filtration/IMS/FA. Office of Water,
Washington, DC. EPA 821-R-99-001. January 1999.
Occurrence and Exposure Assessment
for the LT2ESWTR 9-19 December 2005
-------
USEPA. 1999e. Method 1623: Cryptosporidium and Giardia in water by filtration/IMS/FA. Office of
Water, Washington, DC. EPA 821-R-99-006. April 1999.
USEPA. 1998a. National primary drinking water regulations: Disinfectants and disinfection byproducts;
Final rule. Federal Register 63(241):69389-69476. December 16, 1998.
USEPA. 1998b. National primary drinking water regulations: Interim enhanced surface water treatment;
Final rule. Federal Register 63(241):69477-69521. December 16, 1998.
USEPA. 1998c. Occurrence and exposure assessment for the interim enhanced surface water treatment
rule. Prepared for the Office of Ground Water and Drinking Water, Washington, DC, by Science
Applications International Corporation, McLean, VA.
USEPA. I998d. Addendum to draft drinking water criteria document for Cryptosporidium. Office of
Water. EPA 815-B-98-011. July 1998.
USEPA. 1998e. Giardia human health criteria document. EPA 823-R-99-002. Office of Water.
Washington, D.C. August.
USEPA. 1997. National primary drinking water regulations; disinfectants and disinfection byproducts;
Notice of data availability; Proposed rule. Federal Register 62(212):59388-59484. November 3,
1997.
USEPA. 1996a. National primary drinking water regulations: monitoring requirements for public drinking
water supplies; Final rule. Federal Register 61 (94):24353-24388.
USEPA. I996b. ICR microbiai laboratory manual. U.S. EPA Office of Research and Development.
Washington, DC. EPA 600-R-95-178. April 1996.
USEPA. 1989a. National primary drinking water regulations; Filtration, disinfection, turbidity, Giardia
lamblia, viruses, Legionella, and heterotrophic bacteria; Final rule. Part II. Federal Register
54(124):27486.June29, 1989
USEPA. 1989b. National primary drinking water regulations; Total coliforms (including fecal coliform
and E. coli); Final rule. Federal Register 54(124):27544. June 29, 1989.
USEPA. 1989c. Guidance manual for compliance with the filtration and disinfection requirements for
public water systems using surface water sources. Federal Register 54(124):27544. June 29,
1989.
USEPA. 1985. Drinking water criteria document for viruses. United States Environmental Protection
Agency, Washington, DC. ECAO-CIN-451.
USEPA. 1979. National interim primary drinking water regulations; Control of trihalomethanes in
drinking water. Federal Register 44{231):68624. November 29, 1979.
Vaughn, J. M., E. F. Landry, M. Z. Thomas, T. J. Vicale, and W. F. Penello. 1979. Survey of human
enterovirus occurrence in fresh and marine surface waters on Long Island. Appl. Environ.
Microbiol. 38(2):290-6.
Occurrence and Exposure Assessment
for the LT2ESWTR
9-20
December 2005
-------
Venczel, L.V., M. Arrowood, M. Hurd, and M.D. Sobsey. 1997. Inactivation of Cryptosporidiumparvum
oocysts and Clostridium per/ringens spores by a mixed-oxidant disinfectant and by free chlorine.
Appl. Environ. Microbiol. 63(4): 1598-1601.
Walker, F.R., and J.R. Stedinger. 1999. Fate and transport model of Cryptosporidium. J. Env.
Engineering 125(4): 825-335.
Walker, M.J., C.D. Montemagno, and M.B. Jenkins. 1998. Source water assessment and nonpoint sources
of acutely toxic contaminants: A review of research related to survival and transport of
Cryptosporidium parvum. Wat. Resovr. Res. 34(12): 3383-3392.
Wellings, F, M, A. L. Lewis, C. W. Mountain, and L. V. Pierce. 1975. Demonstration of virus in
groundwater after effluent discharge onto soil. Appl Environ. Microbiol. 29:751-757.
West, T., P. Daniel, P. Meyerhofer, A. DeGraca, S. Leonard, and C. Gerba. 1994. Evaluation of
Cryptosporidium removal through high-rate filtration. Proc. AWWA 1994 Annual Conference,
Water Quality. New York: AWWA, June 19-23.
White, G.C. 1986. The handbook ofchlorination. New York: Van Nostrand Reinhold.
Wickramanayake, G.B., A.J. Rubin, and O.J. Sproul. 1985. Effects of ozone and storage temperature on
Giardia cysts. J. AWWA 77(8): 74-77.
Widmer, G., D. Akiyoshi, M.A. Buckholt, X. Feng, S.M. Rich, K.M. Deary, C.A. Bowman, P. Xu, Y.
Wang, X. Wang, G.A. Buck, S. Tzipori 2000 Animal propagation and genomic survey of a
genotype 1 \so\nleofCryptosporidiumparvum. Molecular and Biochemical Parasitology
108:187-197.
Wiedenmann, A., B. Fischer, U. Straub, C.-H. Wang, B. Flehmig and D. Schoenen. 1993. Disinfection of
hepatitis A virus and MS-2 coliphage in water by ultraviolet irradiation: Comparison of U V-
susceptibility. Wat. Sci. Tech. 27(3-4): 335-338.
Wolfson, J.S., J.M. Richter, M.A. Waldron, D.J. Weber, D.M McCarthy, C.C. Hopkins. 1985.
Cryptosporidiosis in immunocompetent patients. N. Engl. J. Med. 312(20): 1278-1282.
Xiao, L., U.M. Morgan, J. Limor, A. Escalante, M. Arrowood, W. Shulaw, R.C. Thompson, R. Payer, and
A. A. Lai. 1999. Genetic diversity within Cryptosporidium parvum and related Cryptosporidium
species. Appl. Environ. Microbiol. 65(8): 3386-91.
Xiao, L., I. Sulaiman, R. Payer, and A.A. Lai. 1998. Species and strain-specific typing of
Cryptosporidium parasites in clinical and environmental samples. Mem. Inst. Oswaldo Cruz
93(5): 687-91.
Yates, R.S., J.F. Green, S. Liang, R.P. Merlo, and R. DeLeon. 1997. Optimizing coagulation/filtration
processes for Cryptosporidium removal. Proceedings International Symposium on Waterborne
Cryptosporidium. Newport Beach.
Zhang, C. Y., X. F. Li, and Z. Q. Wang. 1991. The comparative study of the levels of viruses and
indicator bacteria in source water and tap water. Virologica Sinica 6 (1):65-70.
Occurrence and Exposure Assessment
for the LT2ESWTR 9-21 December 2005
-------
***t
Appendices to the
Occurrence and Exposure
Assessment for the Final
Long Term 2 Enhanced
Surface Water Treatment Rule
-------
Office of Water (4606-M) EPA 815-R-06-002 December 2005 www.epa.gov/safewater
-------
Appendix A. Waterborne Outbreaks Caused by Microbial Agents in Public
Water Systems 1991-2000
Year
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1991
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1992
1993
1993
1993
1993
1993
1993
1993
1993
1993
1994
1994
1994
1994
1994
State
CA
IL
Ml
Ml
MN
MN
MN
NM
PA
PA
PA
PA
PA
PR
PR
ID
MN
NV
NY
NC
OH
OR
OR
PA
PA
PA
PA
PA
PA
WY
MN
MO
NV
NY
PA
PA
SD
3D
Wl
IN
ME
MN
NH
NH
Cases Etioloav System
15 Giardia NC
386 AGI NC
1,320 AGI NC
33 AGI NC
30 AGI NC
30 AGt NC
17 AGI NC
38 AGI NC
170 AGI NC
8 AGI NC
1 3 Giardia NC
551 Cryptosporidium NC
300 AGI NC
202 AGI C
9,847 AGI C
1 5 Giardia C
250 AGI NC
80 Giardia C
107 AGI NC
200 AGI NC
129 AGI NC
3 000 CryPtosPoridium c
Cryptosporidium C
5 AGI NC
28 AGI C
42 AGI NC
50 AGI NC
57 AGI NC
80 AGI NC
150 Shigelta sonnei NC
27 Cryptosporidium NC
625 Salmonella C
serotype
Typhimurium
103 Cryptosporidium C
172 Campylobacter C
jejuni
20 Giardia lamblia C
65 AGI NC
7 Giardia C
40 AGI NC
403,000 Cryptosporidium C
118 AGI NC
72 AGI NC
19 Campylobacter NC
jejuni
18 Giardia C
36 Giardia C
Deficiency
4
5
2
2
2
4
2
2
3
3
3
3
3
4
3
2
3
3
4
2
4
3
3
3
5
3
3
3
3
2
5
4
5
5
3
3
2
2
3
2
2
2
3
3
Location
Recreation area
School
Campground
Resort
Campground
Resort
Restaurant
Camp
Picnic Area
Restaurant
Park
Picnic Area
Camp
Penitentiary
Community
Trailer Park
Restaurant
Community
Restaurant
Restaurant
Campground
Community
Community
Restaurant
Park
Camp
Camp
Camp
Camp
Park
Resort
Community
Community
Subdivision
Trailer Park
Ski Resort
Subdivision
Resort
Community
Restaurant
Camp
Park
Community
Community
Source
Spring
Well
Well
Well
Well
Well
Well
Well
Well
Weil
Well
Well
Well
River
River
Well
Lake
Lake
Well
Well
Well
Spring
River
Well
River
Well
Well
Well
Well
Well
Lake
Well
Lake
Well
Well
Well
Well
Well
Lake
Well
Well
Well
Reservoir
Lake
Occurrence and Exposure Assessment
for the LT2ESWTR
A-l
December 2005
-------
Year
1994
1994
1994
1994
1994
1995
1995
1995
1995
1995
1995
1995
1995
1995
1995
1995
1996
1996
1996
1997
1997
1997
1997
1997
1997
1997
1998
1998
1998
1998
1998
1999
1999
1999
1999
1999
1999
1999
1999
1999
2000
2000
2000
2000
2000
State
NY
Saipan
PA
TN
WA
ID
ID
MN
MT
NY
OK
PA
SD
WA
Wl
Wl
ID
NY
Wl
CO
NM
NY
NY
OR
SD
WA
FL
MN
OH
TX
WY
CA
FL
FL
FL
MO
NM
NY
TX
WA
CA
CO
FL
FL
FL
Cases Etioloav System
230 Shigella sonnet NC
11 Non-01 Vibrio C
cholerae
200 AGI NC
304 Giardia C
1 34 Cryptosporidium C
83 Shigella sonnei NC
18 AGI C
33 E. co//O1S7:H7 NC
450 AGI NC
1 ,449 Giardia C
10 Shigella sonnei NC
19 AGI NC
48 AGI NC
87 Giardia C
26 AGI NC
148 Small round C
structured virus
94 AGI NC
60 Plesiomonas NC
shigelloides
21 AGI NC
9 AGI NC
123 AGI NC
1 ,450 Norwalk-like NC
virus
50 Giardia C
100G/artf/a NC
16 AGI NC
4E. co//O157:H7 NC
7 Giardia C
83 Shigella sonnei C
10 AGI C
1 ,400 Cryptosporidium C
157E co//0157:H7 C
31 AGI NC
4 AGI C
6 AGI C
3 AGI C
124Sa/mone//a C
typhimurium
70 Small round- NC
structured virus
781 £. co//O157:H7/ NC
Campylobacter
jejuni
22 E. CO// 01 57: H7 C
68 AGI NC
147 Norwalk-like NC
virus
27 Giardia NC
19 AGI C
21 AGI C
5 Cryptosporidium C
Deficiency
2
5
3
4
2
2
3
3
2
3
3
2
2
4
3
4
3
3
4
3
4
3
3
4
3
3
2
4
4
3
2
2
2
4
4
3
3
2
3
2
2
3
3
3
4
Location
Camp
Bottled Water
Resort
Correctional Facility
Community
Resort
Community
Camp
Campground
Water utility
Store
Inn
Camp
Community
Restaurant
School
Camp
Restaurant
Restaurant
Cabins
Country club
Ski resort
Community
Campground
Campground
Trailer Park
Community
Fairgrounds
Treatment plant
Subdivision
Community
Camp
Community
Apartment
Community
Community
Camp
Fairground
Community
Soccer match
Camp
Resort
Trailer park
Trailer park
Community
Source
Weil
Wells
Well
Reservoir
Well
Well
Well
Spring
Well
Lake
Well
Well
Well
Well
Well
Lake
Well
Spring
Well
Spring
Well
Well
Lake
Well/Spring
Well
Well
Well
Well
Surface
Well
Well/SDrinq
Well
Well
River/stream
Well
Well
Spring
Well
Well
Well
Well
River
Well
Well
Well
Occurrence and Exposure Assessment
for the LT2ESWTR
A-2
December 2005
-------
Year State Cases
Etiology System Deficiency
Location
Source
2000 ID
2000 ID
2000 KS
2000 MN
2000 OH
2000 WV
15 Campylobacter NC
jejuni
65 AGI NC
86 Norwalk-like NC
virus
^2Giartiia NC
29 E. co//O157:H7 C
123 Norwalk-like NC
virus
2
2
2
4
3
Camp
Restaurant
Reception hall
Camp
Fairground
Camp
Spring
Well
Well
Well
Surface
Well
AGI = Acute gastrointestinal illness of unknown etiology.
NC = Non-community; C = community
Definitions of deficiencies: (1) Untreated surface water; (2) untreated ground water; (3) treatment deficiency (e.g.,
temporary interruption of disinfection, chronically inadequate disinfection, and inadequate or no filtration); (4)
distribution system deficiency (e.g., cross-connection, contamination of water mains during construction or repair,
and contamination of a storage facility); and (5) unknown or miscellaneous deficiency (e.g.. contaminated bottled
water.
Sources: Moore et al. 1993, Kramer et ai. 1996, Levy et al. 1998, Barwick et al. 2000, and Lee et al. 2002
Occurrence and Exposure Assessment
for the LT2ESWTR
A-3
December 2005
-------
Appendix B. Modeling Microbial Source Water Occurrence
Chapter 3 of this document discusses the primary sources of measurement data used by EPA to
characterize the occurrence of Cryptosporidium and other pathogens in surface water used as drinking
water sources. That discussion addresses the methods that were used to collect and analyze those data. It
also addresses the statistical models that EPA developed to use in conjunction with those measurement
data to derive a plausible range of estimates of the actual (but unknown) national distribution of
Cryptosporidium occurrence in the source waters used by public water supplies. Modeling is necessary
because the national occurrence of Cryptosporidium cannot be fully revealed by the available
measurement data alone due to a variety of limitations and uncertainties inherent in those measurements.
Chapter 4 of this document presents both descriptive summaries of the pathogen occurrence
measurement data collected by EPA, and the results of the statistical modeling performed by EPA to
characterize national occurrence based upon the measurement data.
The purpose of this Appendix is to provide additional technical detail on EPA's approach to the
statistical modeling discussed in Chapter 3 and that was used to derive the occurrence information
presented in Chapter 4.
This appendix also presents several evaluations done by EPA to examine the validity and
implications of some of the key assumptions made in the modeling that was performed.
This appendix is organized into the following major sections:
B. 1 Model Overview
B.2 Model Structure
B.3 Model Inputs
B.4 Model Fitting and Outputs
B.5 Model Evaluations
B.6 Reduced-Form Model
As indicated in Chapter 3, EPA initially developed the full form of the microbial occurrence
model, which is described in sections B.I through B.6, for filtered plants which comprise the majority of
all surface water systems in the U.S. EPA also developed a reduced form of the model for unfiltered
plants primarily to better accommodate the more limited input data available for those plants compared to
the filtered plants. For consistency sake in implementing these models for evaluating the impacts of
regulatory alternatives, EPA chose to also develop a reduced form of the model for filtered plants as well.
The reduced-form model that was used for the economic impact analysis is described in Section B.6.
B.1 Model Overview
There are several related objectives of the Cryptosporidium occurrence modeling performed by
EPA. Key among those objectives is to provide a scientifically defensible characterization of the national
distribution of this pathogen in surface waters that are used as a source by public drinking water systems.
This information is critical for understanding the current risks of endemic cryptosporidiosis among those
served by surface water systems, and to estimate and compare the costs and benefits of reducing that risk
from the implementation of treatment changes to comply with several regulatory alternatives being
considered by EPA for the LT2ESWTR. In addition to that overarching objective, the occurrence
modeling effort also identifies important relationships between pathogen occurrence and other specific
Occurrence and Exposure Assessment
for the LT2ESWTR B-l December 2005
-------
characteristics of the source waters examined to help guide the development of regulatory and treatment
alternatives to most effectively eliminate or minimize the risks posed by Cryptosporidium for public
water supplies.
There are two main facets of the Cryptosporidium occurrence modeling. The first focuses on
modeling the expected average concentration of Cryptosporidium at individual locations reflecting the
observed data, uncertainties and limitations of the sampling and analysis procedures, and possible
influence of other characteristics of the source water being considered. The second facet of the modeling
focuses on characterizing the national distribution of the average levels of Cryptosporidium in source
waters used by public water systems based upon the aggregation of modeled levels at specific locations
obtained in the first stage of the modeling.
The modeling performed to characterize average Cryptosporidium at individual locations
involves the basic assumption that measurements taken at a individual locations will each follow a
Poisson distribution specific to that site. As discussed below, there are a number of general and specific
technical issues that must be addressed in deriving the specific Poisson distribution for each location.
The modeling performed to characterize the national distribution of average Cryptosporidium
concentrations involves the assumption that the distribution of those individual location averages can be
characterized by a lognormal distribution.
Another important consideration in modeling both the individual location averages and the
national distribution of those individual location averages is the recognition of the many limitations and
uncertainties in the underlying measurement data, as well as those resulting from other modeling
assumptions used by EPA. Therefore, another key objective of the modeling approach used by EPA was
to capture and reflect that uncertainty in the model outputs. As described more fully in the sections that
follow, the results of this modeling are not limited to a "best estimate" of occurrence, but rather these
results are presented as sets of plausible occurrence distributions that are consistent with the underlying
observations.
B.I.I Overview of Modeling Occurrence at a Single Location
As discussed by Haas et al. (1999), the most appropriate probability distribution for
characterizing the occurrence of microorganisms in source water at a particular location is the Poisson
distribution. The Poisson distribution is a fundamental probability distribution that is applied when the
average number of occurrences of a discrete event is the result of a large collection of situations in which
that event could occur, and a very small probability for it to occur in any one specific situation. It is used
extensively to address problems that arise in the counting of relatively rare and independent events
occurring in some unit interval of time, length, area, or volume (Sachs 1984).
Some examples of discrete, independent events occurring in some unit interval that may be
appropriately described by the Poisson distribution are radioactive disintegration (time interval), material
irregularities in a wire or surface (length or area interval), and raisins in raisin bread (volume interval).
The Poisson distribution is commonly expressed mathematically as:
p(x=
Occurrence and Exposure Assessment
for the LT2ESWTR
B-2
December 2005
-------
This equation describes the probability that in some randomly selected interval (or volume or
area) the observed number of events X will equal some specified value x, given that the known average
(or expected) number of events is equal to the value X. The only parameter of the Poisson distribution is
X, the average (or expected) number of events in that unit interval.
The occurrence of microorganisms in a unit volume of water is also a type of phenomenon that
can be appropriately described by the Poisson distribution. EPA has used the Cryptosporidium
monitoring data obtained from the Information Collection Rule (ICR) monitoring program, the ICR
Supplemental Surveys, and the Poisson distribution assumption to derive estimates of the average
Cryptosporidium concentrations in the source waters used by the plants included in those surveys. More
specifically, EPA has used these monitoring data and model assumptions to derive a range of plausible
estimates of the underlying source water occurrence of Cryptosporidium that are consistent with the
observations and that also reflect the limitations and uncertainties inherent in collecting and analyzing
those data.
A basic type of question about Cryptosporidium occurrence in the source water used by an
individual plant that the Poisson distribution is suited to answer would be the following:
If the actual average concentration of Cryptosporidium in the source water is 1 oocysts
per liter, what is the probability that an analyst examining a randomly selected one liter
volume of water will observe exactly one oocyst? Or two oocysts? Or zero oocyts?
If, as stated above, the known underlying concentration of Cryptosporidium is 1 oocyst per liter,
then one would compute from the Poisson distribution that the probability of observing exactly I oocyst
in a randomly collected 1 -liter sample is the following:
IV1
P(X = 1|1) = —— = e~l = 0.368
That is, given that the actual average Cryptosporidium concentration is 1 oocyst per liter, it would
be expected that one will actually observe exactly 1 oocyst in a randomly selected 1-liter sample only
about 37 percent of the time.
Similarly, the expectation of observing zero oocysts in a liter sample given a 1 oocyst per liter
average is the following:
P(X = 0|1) = —— = e~l = 0.368
That is, under these circumstances it is just as likely to observe zero oocysts in a random 1-liter
sample as it is to observe 1 oocyst in that sample, even when the known underlying concentration is 1
oocyst per liter. Note that under these assumptions, there is an approximately 18 percent chance of
observing exactly 2 oocysts, and approximately 6 percent chance of observing 3 oocysts, leaving about a
2 percent chance of observing 4 or more oocysts in any randomly selected 1-liter sample.
Occurrence and Exposure Assessment
for the LT2ESWTR B-3 December 2005
-------
In the above example, the assumed actual concentration of Cryptosporidium of 1 oocyst per liter
is the upper end of the range expected to be encountered in source waters. A more typical source water
concentration would be 0.1 oocyst per liter. In the ICR, the median sample volume analyzed was
approximately 3 liters. With a concentration of 0.1 oocycst per liter and a 3-liter sample, the expected
count (X) is 0.3 (=0.1 * 3). Using the Poisson distribution, the probability of seeing zero oocysts in a
randomly drawn 3-liter sample is the following:
P(X = 0|0.3) =
0.3° - e~°3
0!
= e
-03
= 0.741
That is, even though Cryptosporidium is present at a concentration of 0.1 oocyst per liter, just
over 74 percent of all randomly selected 3-liter samples are expected to result in observations
(measurements) of zero. The expectation of a observing substantial number of zero count measurements
even when Cryptosporidium is present in the source water is an important factor to keep in mind when
considering how the Poisson model is actually used in the occurrence modeling.
The use of the Poisson distribution as shown in the above examples allows one to calculate the
probability of observing a particular number of events (in this case, counts of Cryptosporidium oocysts
present in some unit volume of water) when one already knows the model's parameter X (in this case, the
expected count of Cryptosporidium in that unit volume of source water—which is to say, the average
concentration of Cryptosporidium in association with some specified sample volume). However, what is
actually needed by EPA are estimates of the average Cryptosporidium concentrations in the sampled
source waters that have resulted in the observed occurrence data at various locations studied in the ICR
and ICRSS. Therefore, the modeling effort undertaken by EPA in this first facet of the overall modeling
process is focused on estimating the Poisson distribution X parameters—or more specifically, the
underlying concentrations reflected in those parameters—based upon the observed measurement data
from the ICR and ICRSS.
The process of estimating the X parameter for a Poisson distribution can often be a challenge
because one of the characteristics that gives rise to the Poisson distribution is that the events being
characterized are relatively rare. Ideally, one would be able to make a large number of reliable and
representative observations so that a sufficient number of the rare events are observed in order to estimate
the value of the X parameter with a high degree of confidence. In many cases, however, there are
limitations on the number of observations that can be made and uncertainties inherent in the collection
and measurement of the data that are used to derive the model parameter.
Several such difficulties occur in the case of Cryptosporidium measurements in the ICR and
ICRSS data. Key among the difficulties encountered are the relatively small (and nonuniform) sample
volumes collected in those studies, the limited number of total samples taken, and measurement
difficulties that result in less than 100 percent recovery (counting) of all of the oocysts that are actually
present in a sample. These difficulties, and the efforts taken by EPA to overcome them, are discussed in
the next several sections of this appendix.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-4
December 2005
-------
B.I.2 Overview of Modeling National Occurrence
Because the 1CR and ICRSS do not provide data for Cryptosporidium occurrence at all times for
all source waters used by public water systems, it is necessary to consider the occurrence data and
modeling results for the individual locations included in those surveys as representative samples.
EPA has used the basic assumption that the national distribution of plant-mean Cryptosporidium
levels can be modeled as a lognormal distribution. The lognormal distribution is another fundamental
probability distribution that is used commonly and effectively to characterize environmental contaminant
occurrence. The basic characteristic of a lognormal distribution is that the logarithms of the values being
evaluated (in this case, the plant-mean concentrations of Cryptosporidium in source waters) are normally
distributed. One property of the lognormal distribution that makes it particularly well-suited to describing
phenomena like environmental contaminant occurrence data is that it is bounded by zero on the low end
and it reflects a "right-skewed" distribution—that is, it has a tail in the upper end—that is consistent with
having a small proportion of values with relatively high values. The lognormal distribution has two
parameters, the log mean (usually referred to as mu or fi) and the log standard deviation (usually referred
to as sigma or a).
In addition to a set of estimates of the plant-mean Cryptosporidium levels at the various sample
locations from the first aspect of the modeling, the model also derives the parameters of a lognormal
distribution that is consistent with those modeled plant-mean values. In the overall modeling framework,
this process is repeated a large number of times to capture the uncertainty in both the estimates of the
plant-means at individual locations and the uncertainty associated with estimating the lognormal
distribution parameters. However, this process does not capture the following types of uncertainty: model
uncertainty, uncertainty in the analytical method, and uncertainty associated with assuming that small
systems are like medium and large systems. Note, as with all modeling efforts of this type, the scope of
the uncertainty analysis is constrained by the specific distributional assumptions adopted in performing
the hierarchical modeling, and therefore results obtained from the analysis represent a lower bound on the
overall uncertainty.
B.I.3 Basis for Modeling
There are two features of the underlying data that suggest that modeling is an appropriate
approach to estimating national occurrence. These are small sample volumes and low recovery rates,
both of which operate to produce low counts of Cryptosporidium oocysts.
The first of these, small sample volumes, was touched on somewhat in the general description
given above of the difficulties in parameterizing a Poisson distribution model. As noted there, the median
sample volume size in the ICR was only 3 liters, and if the "true" underlying average concentration in the
source water is 0.1 oocyst per liter, it is expected from the Poisson distribution that no Cryptosporidium
oocysts would be observed in approximately 74 percent of the samples.
The second aspect of the measurement process is that the recovery rate for the methods was less
than 100 percent. As discussed in Chapter 3, the mean recovery for the ICR method was 11.6 percent and
for the ICRSS method was 43 percent. Therefore, this suggests that most of the oocysts present in the
source water samples collected in these surveys many not have been counted in the assays performed.
A simple simulation analysis was performed to show the potential combined effect of both a
relatively low sample volumes and the low recovery rate in the ICR across a range of possible
"true:"Cryptosporidium concentrations in a source water. Based on analyses of spiked samples using the
Occurrence and Exposure Assessment
for the LT2ESWTR B-5 December 2005
-------
ICR methods, the recovery rate for the Cryptosporidium was modeled as a beta distribution with
parameters a= 1.44, p= 11,20 (mean = 0.114 or 11.4 percent recovery) (Messner, 2000). Note that the beta
distribution is the natural distribution for describing a continuous random variable with a value between
zero and one. This implies that, on average, an oocyst actually present in a water sample would only be
counted as such about 11 percent of the time. Note that the mean of the beta distribution, 11.4, differs
slightly from the mean of 11.6 based on the spiked study.
Exhibit B.I a shows the distribution of recovery rates as the density function of a beta distribution
having the parameters a=l.44, p=l 1.20.
Exhibit B.1a
Recovery Rate Distribution Described by a Beta Density Function
with Parameters a=1.44, (3=11.20 (average = 0.114)
0.2
0.4
0.6
0.8
Recovery Rate
Exhibit B.lb depicts the combined effect of small volume assayed and low, variable recovery
across a range of possible "true" source water concentrations of Cryptosporidium. These possible "true"
average Cryptosporidium concentrations in the sampled water are shown on the x-axis of Exhibit B.I.
The probability of observing zero oocysts in a 3-liter sample drawn from a source water having a specific
underyling actual concentration is shown on the y-axis.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-6
December 2005
-------
Exhibit B.1b
Expected Probability of Observing a Zero Count as a Function of Actual Oocyst
Concentration, When 3 Liter Sample is Assayed and Recovery is Beta (1.44,11.2)
0.01
0.1 1
concentration, oocysts per liter
10
100
As indicated by the two examples featured by the vertical and horizontal lines on the graph, a 3-
liter sample drawn from a source water having O.I oocysts per liter, and a recovery rate varying about a
mean of 1 $.4%, is expected to yield a zero count with probability 0.97. Even when the actual
concentration is an order of magnitude higher at 1 oocyst per liter, the Exhibit shows that a 3-liter sample
will yield a zero with probability 0.73.
An alternative view of this is provided in Exhibit B.lc. This graph displays the probability of
observing all zeroes in 18 samples of 3 L each given the 'true' concentration shown on the x-axis and an
average recovery rate of 11.4%. For example, if the concentration is 0.01 oocysts per L, then the
probability of observing zero oocysts in a single sample is l-exp[(-(3)(0.01)(0.114)] = 0.0034. The
probability of observing zeroes in all 18 samples is (1-0.0034)" = 0.94.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-7
December 2005
-------
Exhibit B.1c Probability of Observing All Zeroes in 18 Samples of 3 L Each for the
Given Oocyst Concentration and Assuming 11.4% Average Recovery Rate
1.00 n
„ 0.90
1
> 0.80
0.70
CO
e
I
<
o>
.5
0.60
0.50
0.40
I
0.30
0.20
0.10
\
0.00
0.001
0.01
0.1 1
Concentration, oocysts per liter
10
100
As noted previously, EPA developed the occurrence model to accommodate a number of
limitations and uncertainties, including the effect of small sample volumes and low recovery rates, in
order to characterize a range of plausible actual average concentrations of Cryptosporidiwn in sampled
source waters that are consistent with the relatively low incidence of oocysts observed in those surveys.
B.2 Model Structure
There are three levels to the occurrence model; these levels are depicted in Exhibit B.2. At the
lowest level are the modeled features of individual measurements. These include the observed
measurement results (microbial counts, volumes assayed, turbidity values, and source water type) and the
unobserved true concentrations, measurement recoveries, and residuals (e^ = difference between
model-predicted and true concentration). These low-level variables are indexed by both i (location) and j
(month). The middle level of the model includes effects for locations, months, and source water types.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-8
December 2005
-------
Each mid-level variable has only one index. The highest level of parameters includes an intercept term
(P0 = overall mean of log-concentrations), a turbidity effect, and four precisions that define how
lower-level effects are distributed. These high-level parameters are global and require no indices.
Exhibit B.2 Components of the Three Primary Levels of the Occurrence Model
TOP LEVEL
Precisions = <|>MSW)
• Locational Effects = £., ~ Normal(mean 0, precision ^loc)
Month Effects = EJ ~ Normal(mean 0, precision <|>motlU')
BOTTOM LEVEL
• Unobserved Components
• Microbial Concentration = Cj,
• Measurement System Recovery = r^
• Residual = Eij = InfC^) - model-predicted Concentration^
• Observed Components
• Volume Assayed = V,j
• Turbidity = turfy (and standardized value t^}
• Source Water Type = MSW^
Microbial Counts = ¥„ ~ Poisson(Cij * r(j * V^)
Occurrence and Exposure Assessment
for the LT2ESWTR B-9 December 2005
-------
B.2.1 Expected Counts
The observed counts, Yy, are modeled as Poisson random variables:
Where: i = 1 to number of sample locations (350),
j = 1 to number of sample periods (18)
This part of the model can be described as a two-dimensional grid of sample locations (rows) and
sample periods (columns), with a different expected count, \^ for each cell in the grid. Each expected
count is modeled as a function of four parameters:
Xjj = (Concentration^ * Volume^ * Recovery^) + FP
The three subscripted parameters vary with each individual test result. For a given sample tested from
location i and time period j:
Source Water Concentration (Concentration^) — the actual (but unknown and therefore modeled)
underlying concentration of microbes in the source from which the sample was drawn.
Samcle Volume Analyzed (Volume^ — the sample volume analyzed.
Recovery Rate (Recovery^)— the predicted ratio of microbes detected to microbes present in the
sample (as a percent).
The fourth parameter is the same for all samples:
False Positives (FP) — a model adjustment for the possibility of false positives, or detections of
microbes that are not actually present in the sample (e.g., algae or other constituents in the water
may be mistaken for a Cryptosporidium oocyst).
As noted previously, the Concentration tj parameters are unknown and estimating these from the
data is the major focus of the model. The Volume^ values, on the other hand, are known. These are the
sample volumes analyzed, along with sample counts, in the ICR and ICRSS. Because sample volume
analyzed varied widely, and larger samples will contain more microbes, on average, for a given true
concentration, these sample volumes are important predictors of the expected counts.
The Recovery^, values are not known, but are simulated based on results from test method
evaluations. For each test method employed in the ICR and ICRSS, recovery was evaluated by testing
"spiked" samples with known concentrations. These experiments allowed for direct estimation of the
number of microbes detected versus the number actually present in the sample. From these experiments,
a recovery rate distribution was estimated for each analytical method to capture the typical variation in
recovery rates, and the modeled Recovery^ values are drawn randomly from these distributions.
The FP adjustment is derived from: 1) an assumed false positive rate,fp, for a given test method,
and 2) the average number of microbes detected (positives) per sample in the data set. For example, iffp
is assumed to be 1 percent, and a given data set shows a total of 1 00 microbes detected in 1 000 samples,
FP would be set equal to 0.001 (0.0 1 false positives per positive x 0.1 positives per sample = 0.001 false
positives per sample).
Occurrence and Exposure Assessment
for (he LT2ESWTR
B-10
December 2005
-------
Based on expert understanding of the analytical method results, reasonable false positive rates
were tested during model development. The value used is shown in section B.3.3.
B.2.3 Modeling Contributions to Concentration
The modeling of expected counts provides a basic probability structure that links the primary data
(sample volumes and laboratory counts of oocysts) to the values of primary interest — possible true source
water microbial concentrations. In this next step, these possible true source water concentrations are
further broken down as follows:
Concentration^ = R * exp(p0 + p^ + p^MSW^ + e, + Sj + Eio)
Where:
• R = 0 for Zi fraction of concentrations, and 1 for (1-Zj) fraction of concentrations
• Z, = Assumed true zero probability.
* P0 = an intercept term that reflects overall log concentration across all i locations and j sample
periods. Other parameters model deviations from this overall mean.
• P, = Regression parameter for turbidity.
• tg = Loglo of observed turbidity value (in nephelometric turbidity units (NTU), at location (i)
and sample period (j). Measured turbidity values are standardized (re-scaled by adding a
constant to have overall mean zero and standard deviation of one) before input to the model.
This preserves eAp0 as the natural log of the overal! median concentration, and also allows for
easy interpretation of the magnitude of Pi estimates (relative to the other model parameter
estimates, which are all on the microbe concentration scale, not the turbidity scale).
* MSWy = Type of source water (mixed surface water) - 1 ) surface flowing stream, 2) surface
reservoir/lake, 3) ground water under the influence of surface water, 4) mixed surface and
groundwater.
• p2 s = The MSW fixed effects. The P's allow each MSW class to have a different
concentration.
• EI = The location random effect that allows each location to have a different concentration
(it's a "random" rather than "fixed" effect because we are more interested in how these
location effects are distributed than in any particular estimated et).
EJ = Monthly random effect that allows each sample period to have a different concentration.
An important distinction is that EJ and EJ are crossed, not nested, effects, which means that the
EJ measure monthly effects common to all locations and not just within a particular location.
• E,J = Residual term that embodies all other variation and uncertainty.
The true zero parameter accounts for the possibility that a particular water source is entirely free
of a particular microbe. Since the exponential term in this equation for concentration is always greater
than zero, the exponential term is multiplied by a 0/1 random variable, R, that is governed by a "true zero
Occurrence and Exposure Assessment
for the LT2ESWTR B-11 December 2005
-------
probability" model parameter. This parameter can be easily varied to explore model sensitivity to
changes in the assumed "true zero" probability rate.
Note that when R is set equal to one (the usual case) the natural log of concentration is modeled
as a linear function:
^(Concentration,,) = p0 + p.ty + P26MSWi0 + £; + % + ^
Prior distributions are required for the unknown P parameters and for the variance (precision)
parameters on the e distributions since Bayesian techniques are used to estimate their true values. In these
models, broad uncertainty is expressed by using widely dispersed prior distributions that allow the
modeled results to rely largely on the data to drive the parameter estimates.
B.3 Model Inputs
Given the complexity of the occurrence model, it is easy to lose track of exactly where the data
inputs end and the model assumptions begin. To reinforce these distinctions, this section summarizes the
inputs to the Cryptosporidium occurrence models discussed. Much of this information has been discussed
earlier in this appendix and also in Chapters 3 and 4 of the document. Rather than address it all in detail
again, the goal here is to primarily list all the inputs concisely, in one place, and in a logical framework
that clarifies how each contributes to the overall modeling.
B.3.1 Survey Data: Counts, Sample Volumes, Turbidity, and Source Water Type
There are six inputs that come directly from the ICR and ICR Supplemental Surveys. The following
comprise the raw data inputs:
1) Microbial counts
2) Associated sample volumes
3) Associated turbidity measurement
4) MSW categorization (e.g., flowing river/stream, reservoir/lake)
5) Sample location
6) Sample month
B.3.2 Simulated Test Method Recovery Rate
This is a simulated, random input to each model. Recovery values are sampled from the
following probability distributions:
ICR: beta distribution with parameters a=1.44, P=l 1.20 (mean = O.I 14 or 11.4 percent)
ICRSS: beta distribution with parameters a=2, p=3 (mean = 0.400 or 40 percent)
The beta distribution generates values between zero and one. Here it used to characterize a range
of recovery rates from zero (no oocysts ever detected, regardless of how many are in the test sample) to
one (all oocysts in the sample are detected). Based on spiked sample evaluations, these beta distribution
parameters were chosen to closely approximate, based on the best available estimates, the true range of
recovery rates in actual Cryptosporidium testing (including sample to sample variation in this true rate).
Occurrence and Exposure Assessment
for the LT2ESWTR
B-12
December 2005
-------
For both the ICR and ICRSS, there are slight differences in the measured and modeled means (11.6 vs.
11.4 percent for the ICR and 43 vs. 40 percent for the ICRSS).
B.3.3 Tuneable Model Inputs: False Positives, True Zero
There are two "tuneable" inputs to the occurrence models: 1) the false positive rate, and 2) the
true proportion of systems with source water that is completely free of the target microbe. These model
parameters could be easily changed in the model development process to both reflect expert opinion and
to assess the impact of changing the parameter on overall model results.
False positive rates were based largely on expert opinion. In both ICR and ICRSS modeling, a
false positive rate of 0.01 was assumed for total Cryptosporidium counts (the category of count
summarized in Chapter 4 modeled distribution curves for plant-mean concentration).
In developing the model, several values for true zero were tested, ranging from 0 to 50 percent.
Experts believed true zero concentrations rarely occur and based on initial model run results, chose 0.001
percent for the input to the model.
B.3.4 Prior Distributions for Parameters
As discussed elsewhere, parameters in Bayesian models are random variables characterized by
probability distributions. Initially, the researcher chooses a probability distribution for each estimated
model parameter based on previously available information. These are referred to prior distributions or,
simply, "priors." In the case of multi-parameter models, a joint prior distribution captures expected
correlations among these parameters.
Once prior distributions are defined, the method of Bayesian inference uses data to update them.
The result is a joint "posterior" probability distribution for all the model parameters, one that combines
information from the prior distribution and the data to describe the likely range of true parameter values
and relationships among these values.
This Bayesian framework allows for expert opinion, independent of the data, to impact parameter
estimates by way of the prior distributions. It is also possible, however, to choose prior distributions that
have little or no influence on results. This latter approach, which is driven almost entirely by the data,
was adopted in this modeling effort. Broad prior distributions were chosen to reflect considerable
uncertainty about parameter values at the outset of the surveys.
Given this use of these "minimally informative" prior distributions, it is important to emphasize
that these priors are not really an "input" to the model in the same sense as the ICR data, the simulated
recovery values, and the tuneable model parameters discussed above. Instead, these priors are more
accurately thought of as a flexible structure on top of which parameter estimates are built.
B.3.4.1 Prior Distributions for Cryptosporidium Modeling
The next two sections document the prior distributions used in Cryptosporidium modeling. Note
that these parameter values are on the log-scale for concentration. So, for example, the prior distribution
for P0, the overall mean concentration in the model, is centered at zero, or 10°= 1 oocyst/lOOL in terms of
actual concentration.
Occurrence and Exposure Assessment
for the LT2ESWTR B-tt December 2005
-------
Another potentially confusing concept is the specification of variances or "spread" parameters for
the prior distributions in the model. There are two key points to keep in mind. First, because it makes the
Bayesian math easier, these spread parameters are defined in terms of the distribution's precision, which
is the inverse of the variance:
precision = I/a2
The most intuitive measure of spread, the standard deviation, is related to precision as follows:
standard deviation = o = 1/precision"2
Note that this inverse proportion means that larger precision values correspond to smaller
standard deviations, and vice versa.
Second, in the model the four precision parameters—pre^ through prec4—are "meta-parameters"
that are not describing any real-world variation in concentration. Instead, they characterize uncertainty in
estimated parameter values. For example, prec4, the spread parameter for the prior distribution of P0, the
overall mean concentration, does not characterize the spread of plant mean concentrations around P0 but,
instead, uncertainty in our estimate of P0's true value.
Following each prior definition listed below, the range in parentheses captures, roughly, the 1st
and 99"1 percentiles of the distribution. The corresponding 1" and 99lh percentiles of the standard
deviation computed from the precision as shown above are provided in the brackets. These ranges show
that these prior distributions, for the most part, define a very broad range of possible parameter values,
and that the prior probability is roughly equal across these ranges. Because there is so little information
in these prior distributions, the resulting parameter estimates are driven almost entirely by the data.
This model for log concentration is defined in section B.2.3. Prior distributions for the model
parameters are defined below (see previous section for explanation):
P0 ~ Normal(u = 0, precision = prec4), (10'' to 10*32), Overall mean concentration
p, ~ NormalQi = 0, precision = prec,,), (1002 to 1CT32), Slope for standardized turbidity
P26 ~ Normal(u, = 0, precision = prec4), (10°2 to 1(T3 2), MSW class effects
E, ~ NormaKp. = 0, precision = prec,), (10"'2 to 10+l 2), for i = 1 to number of plants
GJ ~ Normal(u, = 0, precision = prec2), (10''2 to 10+l 2), for j = 1 to number of months
Eg - Normal(u = 0, precision = prec3), (10'5 2 to 10"5 2), for all i j
prec, ~ Gamma(a = 2, T = 0.2), (0.7, 33); [0.17, 1.2]
prec2 - Gamma(a = 2, i - 0.2), (0.7, 33); [0.17, 1.2]
prec3 ~ Gamma(a = 2, t = 4), (0.04, 1.6); [0.79, 5.0]
prec4 - Gamma(a = 2, T = 2), (0.08, 3.3); [0.55, 3.5]
B.3.4.2 Comparison of Prior and Posterior Distributions
Exhibit B.3 provides a comparison of prior distributions used in the modeling with the resulting
posterior distributions.
In this comparison, we expect to see an extreme contrast, with the posterior being much narrower
than the original prior. When this happens, it suggests that model parameter estimates are insensitive to
Occurrence and Exposure Assessment
for the LT2ESWTR B-14 December 2005
-------
our choice of prior distribution (that is, they are based largely on the data). When this is not the
case—when the posterior resembles the prior—we question whether our choice of prior has had undue
influence on resulting parameter estimates. In Exhibit B.3, some prior distributions are so broad, relative
to the resulting posterior distributions, that we see only a small portion of the prior distribution in the plot
(for example the flat line at the bottom of the posterior distribution for e,).
Exhibit B.3 Comparison of Prior Distributions with Resulting Posterior
Distributions
Mean of Posterior
Posterior
Prior
1.4 1.6 1.8 2.0 2.2
Std Dev for ei (mu=1.88}
Prior
0.1 0.2 0.3 0.4 0.5 0.6
Std Dev for ej (mu=0.29)
Iki.
1.2 1.4 1.6
Std Dev for eij (mu= 1.51)
1234
Std Dev for beta[1: 5 ] (mu=1.52)
Occurrence and Exposure Assessment
for the LT2ESWTR
B-15
December 2005
-------
Exhibit B.3 (Continued)
Comparison of Prior Distributions with Resulting Posterior Distributions
Mean of Posterior
Posterior
Prior
0.100
RL/FS(mu=0.18)
0.010 0.100 1.000
UF/FS (mu=0.1)
0.001 0.010 0.100 1.000
GI/FS (mu=0.04)
0.100 1.000
MIX/FS (mu=0.25)
Concentration Change with 10-fold Turbidity Increase (mu=1.79)
Occurrence and Exposure Assessment
for the LT2ESWTR
B-16
December 2005
-------
B.4 Model Fitting and Outputs
The hierarchical Bayesian model was fit to ICR and ICR Supplemental Survey data using an
iterative technique known as Markov Chain Monte Carlo (MCMC). This computationally intensive
method was carried out using WinBUGS, a software platform developed jointly by the UK Medical
Research Council, Biostatistics Unit and the Imperial College School of Medicine at St. Mary's, London.
WinBUGS is documented at: http://www.mrc-bsu.cam.ac.uk/bugs/welcome.shtml.
At each step in the bayesian model using MCMC, a single value is drawn from the current
distribution for each parameter, with each draw, in turn, conditional on the current distributions for all the
other parameters. So the result of a single step is a complete set of parameter values, and, in the limit,
these sets of sampled parameter values converge, in distribution, to the target posterior distribution.
There is typically an initial "burn-in" period in which the model fitting takes place, and the
sampled values from these iterations are eventually discarded. In our Cryptosporidium modeling, this
burn-in period was always 400 iterations.
After the burn-in period, when the algorithm has converged to a stable distribution, sampled
values are retained, and these empirical distributions (parameter values sampled from the posterior
distribution) are used to derive parameter estimates. In Cryptosporidium modeling, 5,000 iterations were
run following the burn-in period and, from these values, every tenth set was saved (to avoid auto-
correlation between values) for a total sample size of 500 per parameter.
B.4.1 Plant-Mean Distributions
Exhibits 4.10,4.11, and 4.14 summarize modeled Cryptosporidium plant-means from each of the
three primary occurrence data sets. To obtain these estimates, the average plant-mean concentration is
computed at each sampled model iteration, from the set of 12 concentrations drawn from the bayesian
model using MCMC for each plant in that iteration. The result is 500 distributions of the average plant-
mean concentrations. Observed values from the surveys were used to obtain the simulated sets of 12
samples for each plant that were then used to calculate the means for the plants. The estimates of the
means of each plant were then used to compute the parameters of the lognormal distribution of means
characterizing plant to plant variability. Uncertainty in the true distribution is reflected by the set of 500
such distributions that are generated. In the exhibits, then, each table row gives the overall mean, median,
and 90th percentile from these 500 distributions.
B.4.2 Plant-Mean Distribution Curves
Chapter 4 and Appendix E present cumulative distribution curves for plant-mean concentrations
(e.g., Exhibit 4.9). These are obtained from the model, again, through the bayesian model using MCMC
sampling algorithm. At each iteration, a mean concentration is computed as described above in B.4.1 for
each plant based on the current sample-set of parameter estimates. This collection of sampled plant
means is then compared to 41 reference concentrations, one at a time, and the proportion of plant-means
falling below each of these reference concentration is computed. The result is a 41-point cumulative
distribution curve from each bayesian model using MCMC iteration, or a total of 500 such curves.
Each of the cumulative distribution plots in Chapter 4 and Appendix E summarizes one such set
of 500 cumulative distribution curves. The center curve in each plot connects the median (middle) value,
Occurrence and Exposure Assessment
for the LT2ESWTR
B-17
December 2005
-------
across the 500 sampled values, at each of the 41 reference concentrations. In the same way, the dotted-
line curves connect the 5"1 and 95th percentiles.
B.5 Model Evaluations
In each analysis, a single data set (ICR, ICRSS Large Systems, or ICRSS Medium Systems) was
used to estimate occurrence of either Cryptosporidium or Giardia in drinking water sources. Each
analysis produced a large sample of occurrence distributions. Each individual occurrence distribution
provides one plausible picture of the variability of average concentration among the nation's drinking
water sources. A large "sample" of such distributions reveals uncertainty, due to limited data in the
"true" distribution of variability. The following subsections discuss mixing and autocorrelation, internal
and external model checks, and checks for bias in annual estimates due to seasonality.
B.5.1 Mixing and Autocorrelation
Because the bayesian model using MCMC fitting is an iterative process, sampled parameter
values from nearby iterations are often correlated. When present, this auto-correlation can result in
parameter distributions that systematically under-estimate the variance of the target posterior distribution.
To avoid this problem, it is standard practice in the bayesian model using MCMC sampling to skip
iterations between samples. In the modeling documented here, samples were always drawn from every
tenth iteration, since lag plots consistently showed little evidence of auto-correlation at this spacing.
B.5.2 Internal Model Check
Exhibits 4.9, 4.12, and 4.13 summarize the fit of modeled plant-mean distributions to observed
sample distributions. In each exhibit, one for each of the three primary data sets, the dashed line shows
the distribution of observed plant-mean concentrations. As discussed in Section B.4.2, the solid line and
dotted lines, together, summarize a collection of 500 modeled occurrence distributions.
In the lower half of each distribution, the effect of limited sample volumes is clear. Modeling
predicts smooth distributions through these very low concentrations, while the observed distribution
curve is constrained by "detection limits", and never drops below the overall proportion of zero-count
locations. In the upper half of each distribution, though, the observed data curves generally fall within the
90 percent modeled limits suggesting a good fit, model to data.
B.5.3 External Model Check
To investigate the predictive value of the Cryptosporidium modeling, the following external
model check was carried out:
1) Fit the model (Section B.3) to the first 12 months of ICR Cryptosporidium data only.
2) Use the fitted model from Step 1 along with the input values from months 13 to 18 (sample
volume, turbidity, etc) to predict oocyst counts for months 13 to 18.
3) Compare the predicted counts for months 13 to 18 to the actual, observed counts for these
months.
Occurrence and Exposure Assessment
for the LT2ESWTR B-18 December 2005
-------
Results are summarized in Exhibit B.4 in the form of cumulative count distributions. On both
plots, the circles represent the actual sample counts over the last 6 months of the ICR. The various lines
capture different statistics for the modeled counts. Agreement is good.
Exhibit B.4 Results from External Model Check
•t t\ -
I.U
0.8-
0.6-
8 0.4-
1 °2"
£ o.o-
Proportion <
£ 1.00-
JS
| 0.98-
O 0.96 "
0.94-
0.92
nsn -
C
r
o Observed Counts
Model: Mean Counts
5 10 15 20 25 30 35 40
. _, ^. ^ «-*_ -a- o- ---r
I» .-•''
r; o Observed Counts
•~ "' -.-. MrtrlaL- RtK OjC+ilti
, Model: Median
i , Model, yo /otne
10 15 20 25 30
Total Count for Months 13 -18
35
40
Occurrence and Exposure Assessment
for the LT2ESWTR
B-19
December 2005
-------
B.5.4 Check for Bias in Annual Estimates Due to Seasonally
Microbial concentrations are thought to be related to the frequency and intensity of rainfall,
especially for water systems that are fed by flowing rivers and streams. In attempting to estimate annual
occurrence from an 18-month survey, such effects introduce the potential for seasonal bias. This is
because, for each location in the survey, we capture one complete annual cycle plus one half-year block.
Unless the half-year block fairly represents the typical full-year cycle at a given location (is not
disproportionately from any season), we run the risk of over or under-estimating annual occurrence at this
location. In going from individual location estimates to a national distribution of plant means, such errors
would have to "cancel" each other perfectly to avoid bias.
In Section 4.6, we present some evidence of seasonal trends in Cryptosporidium occurrence. As a
result, we are not able to rule out the possibility of such a bias, due to seasonal effects, in our estimates of
annual average Cryptosporidium occurrence rates from the ICR. Note that this potential problem is only
relevant to the 18-month ICR Survey. The ICR Supplemental Survey was carried out over a 12-month
period, covering one annual cycle.
This section summarizes attempts to measure how big such a bias might be and whether it could
have a significant impact at the next level, where Cryptosporidium occurrence models serve as input to
the LT2ESWTR Economic Analysis. The basic approach is to construct alternative, unbiased estimates
of the national plant-mean distribution based on 12-month intervals. Because they are based on less data,
these alternative estimates are in some ways inferior to our primary 18-month estimate, but they are free
from potential bias due to seasonal effects. We then compare these alternate distributions, based on 12
consecutive months, to our primary distribution, based on 18 consecutive months, to assess the likely
magnitude of any such bias.
Within the 18-month ICR monitoring period, there are seven overlapping 12-month intervals:
July 1997 to June 1998, August 1997 to July 1998, ..., and January 1998 to December 1998. Estimates
of occurrence based on any one of these intervals will be free of any bias from seasonal effects, since each
captures one complete cycle of seasons.
There are two ways to obtain these 12-month plant mean distributions. In the first approach,
model parameters are estimated using all 18 months of data. Since the model includes a set of parameters
that measures each monthly effect, it is possible to construct plant-mean estimates by month from these
18-month parameter estimates, and then group these monthly means into consecutive 12-month
collections. The second method is to simply subset the data into 12- month periods and model each
period separately. Since there are pros and cons to each approach, the model check was carried out both
ways.
In both cases, the comparison of 18-month and 12-month plant-mean curves will be confounded,
to some extent, by the differences in number of months sampled (n=12 means will vary more than n=l 8
means, all other things being equal). Also, the second approach might show slightly more spread in
estimated plant means than the first due to smaller-data-set parameter estimates.
Exhibit B.5 shows the results from the first method, and Exhibit B.5 results from the second.
Both show little difference between the 12-month and 18-month distribution curves. Although
differences are clearly greater in the second approach (Exhibit B.6), they are small enough to be caused
by the sample-size effects discussed above.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-20
December 2005
-------
Exhibit B.5 Annual CDFs Constructed From Parameter Estimates Based on AH
Data
100%
90%
18 Months
1st 12 Mos
* 2nd 12 Mos
3rdi2mos
4th 12 mo*
5th12mos
6th12mos
7th12mos
10%
0% i
1.00E-04
1.00E-03
1.00E-01 100E+00 100E+01
Plint-Mran Concentration (oocy«ts/100L)
1.00E+02
1.00E+03
1.00E+04
Occurrence and Exposure Assessment
for the LT2ESWTR
B-21
December 2005
-------
Exhibit B.6 Annual CDFs Based on 12-Month Data Sets
100%
ISMarths
1st12Mos
2nd 12 MX
v 3rd12Mcs
* 4th 12 W*
• 5H112MB
O
1.00E04 1.COEO3 1.0QE-Q2 1.QOE01 1.00&00 1.00B01 1.00&C2 1.00EKB 1.00&04
Rant-Mean Concentration (oocystsMOOL)
B.6 Reduced-Form Model
The occurrence model described in the preceding sections of this appendix was used to develop
the information for filtered systems in Chapters 3 and 4. EPA has also developed and implemented a
reduced form of the model (also referred to as the "simplified" model), which was used to provide the
information in Chapters 3 and 4 on unfiltered systems. The output of the reduced model was also used as
the direct input to the cost and benefit analyses for both filtered and unfiltered systems in the economic
analyses of LT2ESWTR regulatory options.
The reduced form of the model was developed because of limitations observed in the national
occurrence distributions for unfiltered systems generated by the full form of the model. While those
distributions appeared reasonable across most of the range, it appeared that the upper tails were
overstating the possible occurrence of average Cryptosporidium levels in source water used by unfiltered
Occurrence and Exposure Assessment
for the LT2ESWTR
B-22
December 2005
-------
systems. EPA considered a number of alternatives, such as truncating or modifying the upper tail of the
unfiltered system distributions, including some modifications to the model to reflect some particular
aspects related to the data available for the unfiltered systems.
Unfiltered systems are locations that, at the time of the ICR survey, met strict source water purity
standards that excluded them .from the regulatory requirement to filter. In light of this prior knowledge,
it makes sense to estimate occurrence independently for this class of system. The ICRSS also included a
few unfiltered plants, but too pew to model separately, leaving ICR data as the only useful source for
unfiltered system occurrence estimates.
I
The full model descri
terms, a second set of paramei
sample month, and, finally, ai
plants in the ICR survey (n=3
ed in the previous section includes a large number of parameters: six p*
ers representing every location in the survey, a third representing every
even larger collection of residual terms. The large number of filtered
8) provides enough data to comfortably estimate all these independent
parameters. However, the data from the much smaller sample of unfiltered plants (n=12) is too sparse.
The number of modeled pararieters begins to approach the number of independent data points,
diminishing the usefulness of parameters as representations of more general patterns.
Moreover, measured
variation among them. This niakes
On the standardized scale,
0.08 for filtered plants.
For the economic ana
filtered systems' and unfiltere
1 iirbidity values for these unfiltered iocations are all very low, with little
sense in light of the regulatory requirements for avoiding filtration.
the average plant-mean turbidity for ICR unfiltered plants was -1.5, versus
ysis of the LT2ESWTR, EPA used the reduced form model to predict both
! systems' occurrence distributions. While the simple model was initially
developed for unfiltered systeips, it was also able to produce the data needed as input to the risk
assessment model. Note, the r
distributions of plant-mean to
distributions). Each of the 1,0'
sk assessment model for the LT2ESWTR uses 1,000 log-normal
eflect both variability and uncertainty in Cryptosporidium national
occurrence (plant to plant varii foility in each distribution, uncertainty from the set of 1,000 of these
)0 distributions of plant means represents a plausible national distribution
of plant means based on underlying data. EPA compared the occurrence estimates of the full model and
simple model when considering which to use in the economic analysis, and determined there was no
significant difference between the two.
The following provides additional detail about the reduced-form model and the comparison
between the full model and reduced-form model.
B.6.1 Expected Counts in Reduced-Form Model
At this level, the reduced-form model is the same as the full model. The observed counts, Yy, are
modeled as Poisson random variables and the expected counts are built from concentration, volume, and
recovery:
Yy ~ Poisson(Xg)
Xg = Concentration^ * Volume^ * Recovery^
The only difference here is the lack of an assumed false positive contribution (FP) to the count means. In
earlier work with the full model, the impact of this false positive term was negligible over the range of
likely values, so it was dropped from this simpler model.
Occurrence and Exposure Assessment
for the LT2ESWTR B-23 December 2005
-------
B.6.2 Reduced-Form Model Estimates of the Distribution of Plant-Mean Concentrations
In the full model, estimated concentrations are broken down into a number of underlying effects
and parameters are estimated to model the impact of each general effect on concentration. In the
reduced-form model, the focus shifts to modeling the distribution of estimated concentrations, both within
a particular location (over time) and from location-to-location.
The reduced model is similar to the full model described in B.2.3:
Concentration^ = exp(pFill + P,* Unfiltered?^ + E, + e^),
Where:
• pFilt is an intercept term, the median occurrence level among filtered plants
• Pi is a fixed effect for plants that do not filter. A negative value for this parameter would predict
lower median occurrence in the source waters of plants that filter.
• Unfiltered?jj = 1 if plant i does not use filtration during month j and 0, otherwise.
• E| = random effect for location. This allows different source waters to have different occurrence
levels
• EIJ = residual term that embodies other variation
The reduced model is simpler because it predicts different occurrence levels for only two types of
water (filtered and unfiltered) and includes no effects for months. There are only two precision terms,
prec, describing normally distributed z-, and prec2 describing normally distributed e^.
Thus, each filtered plant will have geometric mean defined by (5^, and the plant-specific effect, GJ.
Concentrations at filtered plant i would be lognormally distributed so that the natural log concentration is
normally distributed with mean pFill + e, and variance l/prec2.
The random effects EJ describe variability from location-to-location among plants that filter and
also among unfiltered plants. These are normally distributed about zero with variance 1/prec,.
$i ~ N(0, prec,)
Within a plant (either filtered or unfiltered), random effects E^ describe how concentration varies
over time. Within a plant, these effects are normally distributed with variance l/prec2.
E^ ~ N(0, prec2)
For unfiltered plants, the model equations are nearly identical. Unfiltered plants will have
medians defined by punfill = pFill + P,. In the discussion that follows, priors, likelihoods, and estimates for
unfiltered systems are expressed in terms of punfi|, rather than ppill + P,.
Occurrence and Exposure Assessment
for the LT2ESWTR
B-24
December 2005
-------
B.6.3 Reduced-Form Model Prior Distributions
In the reduced-form model, it is necessary to use prior distributions for the fixed effects (pFm and
Pundit) and precisions (prec, and prec2). As in the full model, these prior distributions were selected to be
as broad as possible, reflecting genuine uncertainty about these true values and allowing the data to drive
their posterior distributions.
Prior distributions for the model parameters are defined below:
2 = 10,000)
pFill~N(0,o-2 = 10,000)
precl - Gamma(ct= 2, T= 0.2), which has 98% probability mass in [0.7, 33]
prec2 ~ Gamma(oc= 2, t = 2), which has 98% probability mass in [0.08, 3,3]
In the initial modeling work, separate precision terms were assigned to filtered and unfiltered
plants. Although that setup was theoretically reasonable, the small number of unfiltered plants failed to
support estimating their precision parameters. There was insufficient evidence to reject the notion that
unfiltered and filtered plants have common precision parameters (i.e., the hypotheses that the two kinds of
systems have equal between-plant and within-plant variances could not be rejected). As a result,
parameter prec, describes between-plant variability, while parameter prec2 describes within-plant
variability for both filtered and unfiltered plants.
Accordingly, the model uses all of the data (both filtered and unfiltered) to estimate these two
precision parameters. The overall group medians (PFiU and PUlimi), however, are estimated independently.
B.6.4 Comparison of Full Model to Reduced-Form Model
To ensure that the reduced-form model did not differ from the full model in a manner that would
affect the estimated plant-mean distributions used in subsequent economic analyses, EPA conducted a
comparative analysis. To capture uncertainty and variability of the occurrence estimates, the risk model
uses 1 ,000 plant-mean occurrence distributions. These distributions comprise the individual plant-means
(350 plant-means for the ICR data set and 40 for each the ICRSSL and ICRSSM data sets). Exhibit B.7
shows the individual plant-means predicted by the full model versus the simple model and indicates how
closely the two models correlate, with respect to plant-mean estimates.
The plant-means fall on the line where the full model and simple model predicted the same
values. There is a small difference around the 0.01 oocyst/L level, where the simple model predicts
slightly lower concentrations than the full model does.
Occurrence and Exposure Assessment
for the LT2ESWTR B-25 December 2005
-------
Exhibit B.7 Comparison of Full and Simple Occurrence Models
o
o
8
to
5-
o
,0,
q
o
•§
"5
u_
0.000001 0.0001 0.01
100 10000
Simple Model Plant Mean (oocysts/L)
Occurrence and Exposure Assessment
for the LT2ESWTR
B-26
December 2005
-------
Appendix C. Boxplots of Observed ICR Data
Observed data are graphically presented using the following distributions:
• Boxplots of monthly distributions (18 months of data)
• Boxplots of annual cumulative distributions (7 running annuals from the 6 quarters of data)
• Annual cumulative distributions (7 running annuals from the 6 quarters of data)
Exhibit C.I presents an example of the boxplot diagrams used to present the data in Appendix C.
The boxplot identifies the mean, median, minimum point, maximum point, and 10lh, 25th, 75th, and 90th
percentiles. The data points located below the 5* percentile and above the 95lh percentile are plotted
individually.
At the bottom of each boxplot diagram for the Information Collection Rule (ICR) data, the
number of samples taken (N), the number of non-detects (N-Dct), and the standard deviation (Std) are
given for each month or year of sampling. If no standard deviation is shown, this means the standard
deviation is too large to display (>999). This happens only for coliform bacteria. ICR Sampling began in
July 1997 (noted as JL1 in the monthly boxplots) and ended in December 1998 (DC2). The annual
boxplots contain all the monthly data for a given 12-month period (e.g., July 1997-June 1998 (J-J),
August 1997-July 1998 (A-J)).
Exhibit C.1 Boxplot Diagram
4 Above 95th percentile
•« 90th percentile
75th percentile
Mean
Median
25th percentile
10th percentile
Below 5th percentile
Occurrence and Exposure Assessment
for the LT2ESWTR
C-J
December 2005
-------
Exhibit C.2 Exhibit List
Exhibit
C-3
C-4
C-5
C-6
C-7
C-8
C-9
C-10
C-11
C-12
C-13
Pathogen
Cryptosporidium Total
Cryptosporidium Non-Empty
Cryptosporidium - With Internal
Structure
Giardia Total
Giardia - Non-Empty
Giardia - Internal Structure
Giardia - Greater than One Internal Structure
Viruses
Total Coliform
Fecal Conforms
E. Coli
For each protozoan and for coliform bacteria, data are separated by source water type and by
filtration status. There was insufficient data to generate boxplots for viruses in unfiltered systems.
Occurrence and Exposure Assessment
for the LT2ESWTR
C-2
December 2005
-------
Exhibit C-3:
Concentration of Total Cryptosporidium Oocysts in Plant Influent,
ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
u
I
100 •
10 •
0.1 •
0 •
Monthly (18)
iiiijli'illiiihii
1 S , » . • o • ! • • • * i *
• O * ft A ° •
H-EM XM 110 !
ttfc 2 t
Annuals (7)
• •
1
t
' «
> e
) <
1 <
> <
<
> (
»
>
f f * * t t •
Annual Cumulatives (7)
At AQ1 B»» OCt «V1 DCi *H F»t UMl «P1 MTI JEt Al tOt ft OOt KW DGt **df>
O-H
i i a e
N MM am MOV »10 3B11 3S1fl UMtt
•Mf I 0 0 0 0 0 D
(a
o>
•55
>»
CO
CO
CD
.o
U.
100 -
10 -
0.1 -
•
o o
•
•
li:
A1 AUI «fl OCt NVi TCI jMi F»I kMt V< UV« Jfi j« MB •« OC8 WO OC8
1H 117 1t> 117 til 1t« 114 1H 1» tat t!4 IK ttl 117 1W lit IV* ItO
n to* wi io» » I'M aa » i* til IBS tn i>o m» MB IM KB iw
JJ AJ 5 A
N-O D-N
t«M 140C
0)
0)
CO
I
o>
w
0)
££
10 •
0.1 -
I • . . . .
. f. * t • •. •" : • •:
! o ••.*••• : •'•••":
** *5_ •**! # *
I * • I •
*.. *•*•• 0 -
* A • * . «
" o o
o ooo
JLI Ml l»t OCI NVf DCf JHI («l MM M>1 UY1 JEI JIJ ACB 8« Oca NiU DC2
* *
I
I I t • * * *
t « t t t • •
J-J A-J A-A Oft
a» 33M ZB9 XEZI
Cumulative Parcer t
-------
Exhibit C-4:
Concentration of Non-Empty Cryptosporidiurn Oocysts in Plant Influent,
ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 • December 1998
Monthly (18)
Annuals (7)
Annual Cumulatives (7)
10 -
(0
E 1 •
< 0.1 -
o -
tf> 10 •
o
*-rf
CO , -
1 1
c? 5*
35 ^
0> °'1 "
5
_o
o •
w 10 •
E
S
S,
CO 1 -
o> **
1 i
3 |
'o 0.1 •
£
0
DC
o •
*
•
* * * * •
• ; * • s * • .
****** • • 1 • *
• * t ***I»*"i»*'"«
! * . !•!«.. !•:•!*•
;;• J. I ' ' • i ' t J : .
1 ' i • : • i • « ! * : : • I "
•»!.•::••:•• •:•
. Jo'oe**"0 * 5*8
° 0 0 ° ° a *
Monti JL1 Ml) SP1 QC1 NV1 OCI JR1 F*l MAI APt W* J£l JIT AOJ S« Od NV2 PCS
w Hij»B7a»iJi7jnj30iai3a«»sai!Bi»a»aai«sajM
N-OQt tm 311 M9 111 1M JtO »f IQb l(!i 1« WW »1 IIS «W 11« X* 30* 2M
»
: •..•*.;•
* • • • •
*,•••• • ••»
. ! * * . • i * • " * i .
:...!:!! : • : : *
1 • . • * • •
s . • o : :
g
*°8 o0?-5 ° °
0 * o o
MC-* Jt1 Ml »i OC1 NVl OCI JRi FB1 Mm W*1 MYl J£l JU AOZ SfZ 002 HVZ EX3
N 114 117 ItS ITT 111 1»A 114 tfS 190 13D Hi 1l» 11> 117 11» IH 1>B H»
HDot » til MS ttQ KB ICff K6 t« 'i« 114 1M ;M 112 lOt W* in t» io>
I .
*
8 a *
• •
• • • • * *
• * • • •
• * I * •
• . • l . ' ' • : : '
• • • » • •
* * * 0 • 8
o o • «
„ ° e • .
o * o o o o
Uuntt JI1 Ml S»> OCI Kvt OCI On Fftl HRI M"l MYl J£l JU AOK SR OC2 HW PC«
H -a 1- l« l« 1.1 1IB IH .« n» IM l« IN »l IIS IH 1«2 1M IT.
-
'
'
.
-
t
'
-
.
_
«
1
1
Ifenf* J-
N 19
N-Oa «f
•4
H M
t-t>A ^
(
1
UXi i
N K
N-Dct X
M*
1
»
J *
DC H
1$ 37
»
3
<
.1 A
V1 t:
7J IJ
!
i
I
i
i
i
i i
•
>
^ «
QS a?
«7 Z
1
1
1 i
• 1
-J »
«• »
to JT
1
1
» <
^1 ft
IK t]
M) 12
! i
i
1 i
i t
j
1
i
i
»
» <
>j »
O4 23
!SB 29
1
!
A &
M M
>* »7
, '
A O
H n
M U
1
1
>
V
»
A O
aa M
41 29
• * •
i
t t t
• • i
S (*O 0-N J-P
i« Mil M>« *M2
* OT M »«
» * * 0
> « 0 0
S NO ON 4O
• 1401 1*05 »W
OS 11W 1KB 1JBJ
! ! i
1 1 1 1
il 1
j
* * * *
S N-O 4>W JD
Hfl ZWI 22M 22»
0!>» W7a««6W
Cumulative Percent
-------
Exhibit C-5:
Concentration of Cryptosporidium Oocysts with Internal Structures in Plant Influent
ICR Plant-Month Data for 18 Months by Surface Water Category
(All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
Monthly (18)
Annuals (7) Annual Cumuiatives (7)
10 -
All Systems
Oocysts/L
P -L.
_»
i i
0 "
(0 10
E
OT , -
1 ^
CO «
£ 1
55 ° 01
tz
o -
to 10 •
s
CO* 1 -
J i
1 1
"o 0.1 -
0)
* I
o •
*
• m *
• . • . ! •
• • • . • • 8 - .
• . •••••••: • •
i ••••;••:•;• •
• • • • i • • • f •
« • • . • .
o . . o ° • •
• • o o J
0« °0S o» » 8°0
UnX JU mi V< OCl KV1 Del JUl F«l MRI «•! MTt JEI JU M> m (Ml WB Od
N-Oct »• HZ 94 XB SI? »1? »i? Wt Uft «S !!• M« >» 11* >1» lit )M 2W
• »
« * . * « • * « . ' *
: • • " • : ', *
• • O •
* • 0 •
• » «
0 • •.• .• . •'„
o o o 9
MM* JLI «ai SPI oci Nvt oct Jt» rat M«» API mi JEI oca NW cxa
N 114 117 >1S ttr ill H* 1U MS 1M 1» 1f4 n« )(• 1t7 1« lift til 1«
.
*
1 ' : '::'
» •
•
° * *
o m • °
Monti A.1 Mki Wl OCt KV1 DCi JRt Pai URt APt MVt Jtl VU AOB SW OC2 NV5 OCa
N t« 1UB 1SM 1M Ittl 1» 1M '» IW 1B' IB 1M W 1M 1#* l« «M I7fl
'
.
•
• :
i 5 ; i i i i
i t t i i i i
illlll!
II i • t i i
i i i i i i
fen* -Kl AJ 8-A O-4 N-Q (Wt >Q
tp* MM M»T W46 Wtt MM SMJ Ml*
: : : 11 11
J • i 1 j 1 l
MINN
I'lM!!
•::::;;
:..;?::
^ ,«* iMI 1W 1«* IN. «** 1W
• * * *
* » • *
i ! I i i i i
: : . tit
*
2 2 0 9 8 8 8
Monti J-J iW ft-* t>S H-O D-N J
-------
Exhibit C-6:
Concentration of Total Giardia Cysts in Plant Influent, ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
V)
100 -
10
§ 1 1
:». & 1
CO
0.1
0 -
Monthly (18)
Annuals (7)
Annual Cumulatives (7)
iii
MWI*> JL! Mil *Tt DC1 «V) OG1 J^l FBI MR1 API MV1
1 t ! 3 1 I 1 t Z ' »
N 3901 9MB 3WM M04 M0T Wf» MUM
2*1 ZM> 2M 344 2)7
i
03
a>
1
u_
100 •
10 •
o.i
o -
X.1 M11 SPt OC1 KVI DCI JR1 FBI MM1 API UV1 jEt Aj 402 «P3 OC1 NV2 OC2
111 t17 MS 117 Mt Ttf 114 IIS 1» 120 1U IU 111 117 110 1t« US 110
Mltri7r»»KMt>7!KII«MI7!7H>r7N
A-J S.A OS 1*0 0-tt J-t>
,. 10 -
CO
jci ja *oa «J»2 oca NVZ ix»
i& 17t 174 177 m 1JI 1b£ IT4 177 tU
.
-
.
<
•Mi* .t
Sl^tM M
5 (
1 1
j A
«4 »
) (
-J 6-
?1 JO
> (
A 0
77 30
> c
a N
n »
) <
a D>
H 20
> <
N J
70 20
>
0
"
1
/
/
/
/
r
Cumulative Percent
-------
Exhibit C-7:
Concentration of Non-Empty Giardia Cysts in Plant Influent,
ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
to
0>
t5
>s
CO
100
10 -
0.1 -
o -
Monthly (18)
• I * •
• iiMiihisIi
ilh1'1*11 I •
t«l ' i I
• • t
o a
Annuals (7)
Annual Cumulattves (7)
Mw* JL1 MM SP1 OC1 «Vi PCI Jftt m tHn A*l -UY1 rtt J12 Mi ft CCt Wq DCJ Ugnti
M *O frA H lit US lib
N-OET «H9>tfttM7ST7uwMmmi«tMti MT»
Vfek !( t i 1 1 1 I I I 1 1 1 1 1 1 1 I
iii
hhnti JW A-Jj ».A CVS MO 0-N j.D
N <»1 IMS UK 1>M 1*» 1«B tlM
M-M WW KT77 <«M 1««| WW tCtt tQfD
«M> 1 I I 1 i t 1
(0
Q>
CO
^
10 -
_
1
0.1 -
-------
Exhibit C-8:
Concentration of Giardia Cysts with Internal Structures in Plant Influent,
1CR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
Monthly (18)
Annuals (7)
Annual Cumulatives (7)
10 -
V)
E 1 -
?« *4 »* *» s-n in »i »n aw 111 n« »t« n* SM MB
• » » • • •
::•' .i,: :--::-;:
» . . : : ' , • • •:•.»!!
* :
o „" 0 ° • °» °°,0
o°ofo o ° *
O o
o
M*<* Jt1 *OS **>! DC1 NV1 DCt Jf)t F«i MW W»1 hpri JR1 JU K» «P* OC2 KV3 OC*
N tK 117 118 1t7 HI lit 111 118 1& 100 114 I1* Ml M7 11» it* 1^9 110
M-o«t Mm it* «« toa w> 191 iw iw 109 it* m ID* in KM 111 i«a ia> M
* * • » • •
• » .
* 9 :*..*»
* * * * * s .
. . . : . « i
• * * i
• * • 9
*
" j
T ^C-*T-*"T*^*"r!!«r«
(••OC1 1» tR m fW 1» IM ItS IM TV) IM t»1 IM IM 1» »»1 in IW '74
-
•
-
1
•
• * a •
1
1
•
» 8 • J
1 • 1 * 1 .
— »*•«««««
MM ». ,« ™ ™ «» «. »»
... 1
Mont) JNJ *J a A 3-« HO 0-N 40
H t»1 t>K 1X3 ISM t«t IMC tIM
H-CC. ,a» .« .«r ,». ,a» ,m .»
1 t 1 t • • •
• • I • •
till |
till 1
i M M ii
• • • •
• * «
0 O 0 O O O O
* • t * * * t
U**l J-J ArJ Q.A O-B MO D* JO
N^d ^. »s a« m* a« 22s, »7
o wao»«o»«TOso»iflo
Cumulative Percent
-------
Exhibit C-9:
Concentration of Giardia Cysts with >1 Internal Structure in Plant Influent,
ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 * December 1998
Monthly (18)
Annuals (7) Annual Cumulatives (7)
10 •
In »
< °-1
o •
w 10 "
OT , -
E rf
a 1
® 5
? "•
1
u- 0 .
ervoir/Lake Systems
Cysts/L
2-3
0)
0>
DC
0
• « • * i i ; * • * i 1 1 i : i
t • • t i i : ' • . I i |si
• 9 * * m '
• , •;•:••• «
• • •
• • • • t
• • . * • . " • • o » ; • •
°o» , " o 9 0 ° • • ° o
fl • O
MM) JLi *31 SPi DCl HVl DCl Mi FBI MAt ffl MVl Jfl JU Ml |P| OPt NVB DCS
CM^t aw *!• WO OT JO* 107 W7 Hi J'l B1 S1« Hi Itt *'» •» S1I M7 3M
• • • ! . * •
• t *
« : • i ' • : ' . ' • I " : •
•
°oo' - °" "-"o0-.
• o o"o,a o
...«:i...: 1.
UbnV, JLI Ml Sfl Ut1 W DC! *U FBI MR' W1 1W1 JE1 JU *W «P> OC8 NVJ tK3
h 114 it? n* t*7 HI »»f lu tic ijd lan H4 iti m IIT 111 i« tit no
M-Ort ID* WS no 114 iflS 142 1(M IM tlO 11' 10* HI 119 lift 11J tdi 1M 1C1
• •••*•* *
# *
• _ *
: » I
o o 0 . •
* * * ? » tt?9**fl»4?29
• •
hHM E»l 1S9 1W IK !K TWO i«r Itl IK 1* HI IK !«• 1H 1M t«» W 17*
'
•
.
•
-
•••Ill
1 1 1 1 1
•
1 1
: i
! ! t J :
llgnti W AJ M ^» l»0 W. >0
•KM xrav xn« mi Z?M am vn ITS*
•••iii
i ! '
H tM4 tM» tWti tjftf 1*3* l«o» 11W
W4M ittB IMft tioi r»« 1*M 13CQ <»1
* • •
• • • •
i i ; ! ; ! !
* • • • * *
t i ; ; ; ! !
* i t i $ t i
• * *
o o o o o o o
Man* J-J JW A* o* HO D-K .«)
M ZB$ 23M ntt 1319 3320 »» nM
H-tM 9901 XZM 2300 12** 2B7 JQM SOM
<
0 10 J* » 40 flO 30 T6 W W 106
Q ion»««t>wnMwiae
Cumulative Percent
-------
Exhibit C-10:
Concentration of Viruses in Plant Influent, ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions,
July 1997 - December 1998
100 -
10 •
CO
(1)
^ 1 1
to a
0.1 •
Monthly (18)
* . » *
. ? °
• , »
Annuals (7)
i i i i 1 i J
00°°
III II
JU AQ1 *P1 OCl Mtfl DCI Ml FBI MA1 AP* HV1 JC1 JU AIM 8W OM MV* OC3
MMfli >J A^ 6-* OS N-0 0-N J-0
H 2144 2W4 2MB 2»» 3|1« »^« £m
K-Oei lira im 177* 17W 1?H 1744 T?4*
si* o e » A a 4 t
Annual Cumulatives (7)
w 10 -
£
t
£ «•
c
S
11 n -
Q
4
•
*
;.-::
. : •» . ,
1 .
t:l: ,
ftttMi Jtl Ml W»i OCl W
•1 1W 1W WW IPS X
1
1
4
i
»
»
1
n DC
a «
. * .
t
• •
• * •
. • <
i
» 1 1 i 1
1 1 1 H 1
U U U il u
1 JH1 FV1 *W1 *»1 MVl Jf
a tr i«i in 195 w «
1
; • i
!
'
1 JU AC
* W t
4
; *
i
12 SI
X) 1(
,
1
1
) <
,
•a o
8 «
t
> i
' 1
»
t
1
r» *(
« •
<
»
^
1
t
<
a CM
• i
i
i
r
y
»
.
-
,
kto4> J-
N <9
SMb
i (
•f A
U 13
)
• : : i
Mil
INI
Si i i i
tilt
1 1 11
0 0 U 0
j *A as *0 OH J-D
99 I£M 12TC 1210 UD? 12M
0 t 1 } 1
/
/
/
/
rl
co 10 •
a>
CO* 1 -
1 »
"o 0.1 •
IT
0
D *
9 •
• * *
*
*
• *
» 4 * •
*°«*0i " • • 0°*
o° I0i° ° o
uonfc JLI Mt «*; oci NVI DCI jm rtn Mm API MYI jet JL» AS? s« oc« «•«? oca
«w >rMitn77»«7.n>r>4>irr»»<.»«
•
•
!!!!!!!
• o o 8 8 • *
Ukiiti JJ A-J 9-A 0-8 NO OW J-t)
*"" *" *" "" "* W '" *"
/
Cumulative Pwcent
-------
Exhibit C-11:
Concentration of Total Coliform in Plant Influent, ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
CO
&
V)
CO*
<
Monthly (18)
100000C
10000O
10000-
1000-
. 100 •
10 -
1 -
\
I
ii
I"
Ii
jt
I
I
. . •
ill
it J It
III
I I I
«
1
jj
1
1
*
t <
1 1
ii
i
; •
1 «
1!
> 1
k
1 I
> •
t
,
<
1
«01 ITI OCl IW1 PCI JH1 TO1 URt »l>1 MYI JE1 Al «ll WI OCJ tva VOt
Annuals (7)
o-a 1*0 O-M
4M *1* *» 431 4*« «M «6t
nau OM« »m IM*? WDM
Annual Cumulatives (7)
w
£
CO
>»
CO
nj
&
co
o>
u_
1000000
100000-
10000-
i
3 1000-
i> 100 -
a
5 10 -
1
o -
i *
*•
I I J I • • J i
htonffi Al *Ql S?l DCt NVt Dei jRl Fat MM *P1 H¥i JCI JIX Mtf BH| OR? Hva OCX
N 121 120 IZ1 tZl 113 tir 11!> 11* '1? 12D I1» t2Q 111 114 IK «« 111 115
s s :
• »••**•
illiiil
MINI
>J AN! »-« O-B IMS CM4 J-O
UIB I4K 1«10 1«M 1«B 1407 I«M
o IB an jo
JO W 00 MO
w 1000008
§ 100000-
CO* ^
0 o 1000-
J3 > 100 •
^ § 10 -
go
1 -
to
<
1
M
J
!
n
i
a
i
1]
•*
i
i i
1 !
•e
IS
i
<
*:
, «
» l
;]
71
1
i «
i
t
\
•
•
'
•
•
*
Hxptl »
SMi 11
j
« 2
IK
) £
IK
) 2
Bt
>
17
1 *
OM
>
•
i
28
a
)M
^
t
A
i
j
MM* JLl «t «Pl OCI NVl DCl JM! »1 MM* API W»1 JCt Jl3 AO? Srt OO «VJ t
H 112 IB IM TM 1»? 901 «? 107 »i IM 1tt 30O IM «3 1W W 1«C '
N4v( M?*2£M2«M2tMW2*9finVM7iS]M
5** BQ» MM »'t «M K1» B« 9CM 14191 1WM 2
Rkx« JJ A-J S-» 0-5 1*0 i»-N J.O
N 2M «U OM 235* 2MZ »M »M
N-DC4 337 3M J« Mil JSZ J» 3V1
50 flO ftj M 90
Cumulative Percont
-------
Exhibit C-12:
Concentration of Fecal Coliform in Plant Influent, ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
Monthly (18)
Annuals (7)
Annual Cumulatives (7)
100000-
10000-
« _, 1000-
+2 o 100 •
w ^
C/3 "| 10 •
o -
1
1
1
t
!
'
•
1
i 1
> 1
i i
4
;
1
i
» i
ij
•
t
i!
4
4
1
^
»
I
|
1
1
4
•
1
i
1
i
i
i
i
1
•
( 4
1
!
1
.'
«
'
t
i
-
-
-
•
1
4
ft
» 4
1
i
1
<
1
'
X|
X
J
Mar* JU *S1 Wl OGI WV1 get JWI «l Mm API urt Jtl JL2 AOl m OC2 MV2 DO
•t IM »» xa »« aoo an JH »« MI aw aw AM »ii iw aM jw «M aoc
*W» 773 «U M6 tlft MK) ZTt It* X* Mil 4*0 TlT «4S >flZ
<0
CD
I
o>
c
_o
u.
100000-
10000-
i 1000 •
| 100 •
I 10 -
)
I
1
Month JU AOi SP1 OC1 «Vi DC1 Jftl fit UR1 MM MV1 JEl AS AO] SI? 0C3 NV3 DC3
t • I
N-OcI 11 7 t 11 7 IS 5 Iff r ft •
Sl» 139f 1414 HI 1M3 MB 974 »!
t t C to H ifl
*sJ tU O-S M-O D-N
Ml HO HI 4* «4Q
at* 2194 rju 917
3P « tO n TO H 90
100000-
E
0) 10000-
cl - 100°-
5 8 100 •
Jj i1
^J w 10 •
•pr .S
£ i •
(0
0
cc
o •
i
t
!
1
1
I
*
ii
i
»
4
t
1
\ <
«
• ,
;i
»
i;
t
1
» 4
(
;'
• .
1 1
i
i
i
<
«
<
!
i
l
I
»
1
1
i
i
1
i
ii
•
4
>i
1
!
1
4
4
i
i
t
1
!.
<
f
•
i
i
-
•
•
-
*
"
4
4
1
1
»
II
1*4
ft
III
r
1
9 i
II
1 (
)
II
[
9
1
J
X
^-^
/
OCi NVi DCl Ofll *
12T 12) 12S 121
Month J-J A-J V* O-S K-O I>*r J-D
N 14M 14M I4M 1«M 1*9* 1«* UM
l«t 738 2U 113 100 «• SO? »' H5 * t« » l'g lift 441 »7 110
Cumulative Percent
-------
1
Exhibit C-13:
Concentration of E. coli in Plant Influent, ICR Plant-Month Data for 18 Months
By Surface Water Category (All, Flowing Streams, Reservoir Lakes)
Monthly & Annual Boxplots, Annual Cumulative Distributions
July 1997 - December 1998
B
Monthly (18)
1000000
100000-
10000-
1000 •
- 100 -
10 -
1 -
o •
• •
• *
\}
I
i
4
» ,
; i
i
i
1
1
1
i
!
J
•
I
4
*
1
J
1
1
i
•
» •
I *
It
1
1
4
1
i
1
Annuals (7)
Annual Cumulatives (7)
• i
Monti A1 AQI API OC! KVt DCi JRl
Q-* N-Q O-H J-0
aft 2M MM 2386
«TO »» m* »
CO
at
CO
o>
.£
o
d
1000008
100000-
10000-
S 1000-
| 100 -
§ 10 •
1
• •
• • • *
..: 2: •,
t 1 1 • 1
: s •
. s • •
JU A01 9^1 OCt 4V1 OC1 JR1 P11 Mftt API WV] Jt« AV AW Wt OO MVt PCI
n M M «4 «e « ei ««6»«*««w**«€ii3
1»13»I 71 4»»tO«« / » « 10 » »
MW »id nn Mt wr n? wi wr na rzz «M
M m 7»7 Tit 777 77» 7T» 777
H-D«4
I
CO
O
O
(0
(D
cc
100000-
10000-
_, 1000 •
§ 100 •
g 10 -
1
i •
ju.i mi *ri oei NVI DCI
UAT API MYI
i» 137 >n
ju MR &» oca NW oca
MB m 127 i» i» «M
Hill
o o
i n A i
B-H J-t>
lit* IHJ
J-
Hfkt HI «• «t fT* WV
W» 1«4 Ml »«• Mft KW4
Cumulative Percerr:
-------
Appendix D. Graphs of Observed Supplemental Survey Data
D.1 Exhibit List
Exhibit
D.2
D.3
D.4
D.5
D.6
Title
Monthly Mean Cryptosporidium Concentration
by Source Water Type
Monthly Mean Giardia Concentration by
Source Water Type
Monthly Mean Total Coliform Concentration by
Source Water Type
Monthly Mean Fecal Coliform Concentration by
Source Water Type
Monthly Mean E. coli Concentration by
Source Water Type
Occurrence and Exposure Assessment
for the LT2ESWTR
D-l
December 2005
-------
Exhibit D.2 ICR Supplemental Surveys Monthly Mean Cryptosporidium
Concentration by Source Water Type
Total, Non-empty, and Internal CryptatportOum Concmtnten
By Month tor Flowing Stream Sy*twm
Mean Total, Non-Empty, and Internal CryjKiMporWurn Concentration
By Month for ftcmvoMUM SyMem
Occurrence and Exposure Assessment
for the LT2ESWTR
D-2
December 2005
-------
Exhibit D.3 ICR Supplemental Surveys Monthly Mean Giardia Concentration by Source
Water Type
MIM Totil, Nonempty, InUml.ind Inttrrwl >1 OUnM Conontntlon
By Month lot All Sucfict Wit«r«
Mem Total, Non-Empty, Internal, ind Intemil >1
GfeRfe ConccnlraNon by Month for Flowing Stream System*
I O-40
M«an Total, Non-Empty. Internal, tnd feitamal >1 Gtarota Concentration by Month
for RaaervoMLak* Syttem*
QTotal
• Non-Empty
• Utemal
Dlntemal>l
Occurrence and Exposure Assessment
for the LT2ESWTR
D-3
December 2005
-------
Exhibit D.4 ICR Supplemental Surveys Monthly Mean Total Coliform Concentration by
Source Water Type
i
1
7000,
6000
I5™
| 4000
1 3000
Loco
1000-
o.
M»n ToUl Colifwni Cone.n1r.tion by Month for All Sou re. WkMr*
1
1.
.
.
1
1
*
1
II
• 1 1
III
Mean Total Conform Concentration by Month for Flowing Stream Syatems
!i
7000
flood
sow
i *°°
3! -
2000
1000
t>
Mean Total Conform Concentration by Month for R*Mivolrfl_ik« Systems
••••••._
Anuwy February Itoch Apt! **r JUM Jdf AuguM 9»pl«n«»r Octatei Mwtmter DtCtAMf
Occurrence and Exposure Assessment
for the LT2ESWTR
D-4
December 2005
-------
Exhibit D.5 ICR Supplemental Surveys Monthly Mean Fecal Coliform Concentration by
Source Water Type
1900 T
2000
li"
ii~
500
0
MMiiFtc«lCe»armCOTM«n«i1^^H(>in
• - . 1 M
•
I M » ' ' * 1 | 1 | |
Item F«nl CoMorm Concentration by Month for Flowing Strum Systems
U
11.
l..lllllllll
JXU»I> (*fth AF! My
2500
3000
| * 1SOO
11"
i
500
0
Mean Fecal Co Worm Concentration by Month for ReaervolrfLike Syttomc
f I I ^ i l t i j | I i
Occurrence and Exposure Assessment
for the LT2ESWTR
D-5
December 2005
-------
Exhibit D.6 ICR Supplemental Surveys Monthly Mean E. col! Concentration by Source
Water Type
I 3°°
ui
J 200
M*in £. co# Oonc*ntr«0on by Monffi for All Sown:* W«t»*•
October Novwrtwr O*c«ntw
Mean E. cot Concentration by Month for Flowing Stream Syttem*
t
Mill..in
HOT E.cMconc«it»gci!t£.co*flM*il.)
o 8 g g § § § g g
Mean £. co// Concentration by Month for Reservoir/Lake System*
_••_-._-__-•-.
JwMwry tobrury Mirch Apr! U*y Jurat July AuDuM S»p«*nt»tf Otttb»f Nav»Mb*r DK«ftl»r
Occurrence and Exposure Assessment
for the LT2ESWTR
D-6
December 2005
-------
Appendix E. Bayesian Analysis Cumulative Distribution Functions
Exhibit E.1 Table of Graphs
Exhibit Number
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Data Source
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
ICR
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Source Water
All
All
All
Flowing Stream
Flowing Stream
Flowing Stream
Reservoir/Lake
RL
RL
All
All
All
Flowing Stream
Flowing Stream
Flowing Stream
RL
RL
RL
All
All
All
All
All
All
Pathogen
Crypto-Total
Crypto-Non-Empty
Crypto-lnternal Structure
Crypto-Total
Crypto-Non-Empty
Crypto-lnternal Structure
Crypto-Total
Crypto-Non-Empty
Crypto-lnternal Structure
Giardia-Total
Giardia-Non-Empty
Giardia-lnternal Structure
Giardia-Total
Giardia-Non-Empty
Giardia-lntemal Structure
Giardia-Total
Giardia-Non-Empty
Giardia-lnternal Structure
Crypto-Total
Crypto-Total
Crypto-Non-Empty
Crypto-Non-Empty
Crypto-lntemal Structure
Crypto-lnternal Structure
Occurrence and Exposure Assessment
for the LT2ESWTR
E-l
December 2005
-------
Exhibit Number
26
27
28
29
30
31
32
33
34
35
36
37
Data Source
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplemental Survey
- Large Plants
Supplemental Survey
- Medium Plants
Supplementa! Survey
- Large Plants
Supplemental Survey
- Medium Plants
Source Water
Flowing Stream
Flowing Stream
Flowing Stream
Flowing Stream
Flowing Stream
Flowing Stream
Reservoir/Lake
Reservoir/Lake
Reservoir/Lake
Reservoir/Lake
Reservoir/Lake
Reservoir/Lake
Pathogen
Crypto-Total
Crypto-Total
Crypto-Non-Empty
Crypto-Non-Empty
Crypto-lnternal Structure
Crypto-lnternal Structure
Crypto-Total
Crypto-Total
Crypto-Non-Empty
Crypto-Non-Empty
Crypto-lnternal Structure
Crypto-lnternal Structure
Occurrence and Exposure Assessment
for the LT2ESWTR
E-2
December 2005
-------
Exhibit E-2 Cumulate Dfctflbullon o< Tool Crfftosportdiunn Oocy»B in Al PUnts
19-005 00001 0001
001
01
10
Concentration (ooeysts/L)
ICR Modeled Data
Exhibit E-3 Cumutatiw Distribution of Non-Empty Cwtosporidkjm Oocysts in All Plants
1e-005 0.0001 0.001
Concentration (oocysta/l)
ICR Modeled Data
ExtuM E-4 Cumul«tiva CMttfltxition of CiypK>»pondium Oocyitt «flh Imemik stiuoturm in All Plantt
• 314 (ol 3SO) >*w r«M of z»n>
1B-OOS 00001
0001
0.1
Concentration (oocyats/L)
ICR Modeled Data
Occurrence and Exposure Assessment
for the LT2ESWTR
E-3
December 2005
-------
ExtilUt E-t CumuMKa Distrtbullon or Total Cri&BSfOMm OocySs in RawMg Stream Plants
IB-COS 0.0001 o.ooi o.oi
Concentration (oocysts/L)
ICR Modeled Data
EMM E-6 CumuMvB attribution of Non-Emmy CrypKKporidkjni Oocysts In Flowing Stream Plants
16-005 0.0001 0.001 0.01
Concentration (oocysts/L)
ICR Modeled Data
Exhibit £.7 Cumutltiv* Distribution of Cryptotpontfum Oecyti* nn«h Inunv*) Stmauru *i FkNMng str*jm Ptanti
10 _____ " ..........
O.B
0.7
0.6
O.S
0.4
0.3
0.2
0.1
/
' i.144 (gl 1«3) nw r*M at torn
19-005 00001 0.001
0.01
0.1
Concwintion (.
ICR Model*! Data
100
Occurrence and Exposure Assessment
for the LT2ESWTR
E-4
December 2005
-------
E*M E-8 Cumulate™ DBtrtMinn olTMal Crjpocporidlum OocyMi in ResanoWLake Plants
1e-005 0.0001 0001 0.01
0.1
Concentration (ooeysls/L)
ICR Modeled Data
Etfiftit E-» Cumulate Diartbulwr of Non-Empty CrytSOKKXWHim Oocysu in RL Planes
1e-005 00001 0001 0.01
10
100
Concentration (oocysts/L)
ICR Modeled Data
1.0
09
09
f °7
| 06
j OS
1 04
3 03
0.2
0.1
0.0
10 Cumutetiv* OrabitiuMn of Cl>pnm»nd«m Oocyttl wilh lnt»m«j Smc
-------
Exhibit E-11 CumJrfiv« D»t*irt«inl Total GauciaCjats In AlPtanH
16-005 0.0001 0.001 0.01 0.1 1
Conoentnlion (cy»&/L)
10 100
Exhibit E-12 CumMve D»mbutoii of Non-Empty Onnta Cf»t» in All Plans
0.001 0.01 0.1
Conointubon {cyitit.)
100
E>HM E-tl CumJilv. OiWulix of Glmli CyiH wDK Internal SlnKtirei In Ml Pl«in
le-005 0.0001 0.001 0.01 01 1
Concentration (cyttiA.)
Occurrence and Exposure Assessment
for the LT2ESWTR
E-6
December 2005
-------
ErMM E-14 CutraJ»«ve 0»trtbutton of Total Slnd>> Cytta in Fkming Steam p
1o-005 00001
0001 001 01
Con«ntr»t»>n (cy»WL)
10 100
ErhWI E-15 CumulMvt DbBibuton of Mon-Empty Givdii Cylli m Rmkig Str.im PlnB
1e-005 00001 0001 0.01 01 1
Concentration (cytts/L)
Exhibit £-16 Canton DutnbuHon of GuiHa Cyn win Inlemd Structures in Ftowitg Steam Ptarm
1«-005 0.0001 0001 0.01 0.1 1
ConcentntJon (cyits/1.)
Occurrence and Exposure Assessment
for the LT2ESWTR
E-7
December 2005
-------
Exhibit E-17 CumUMw Dfctnbuton of Total Gianfi Cttts X Reswvoi/Uke Plans
1e-005 0.0001 0.001 001 0.1 1
Concentration (cylK/L)
EiliM E-1 > CumJatv* DMiibiilkn >T Non-Envty Ciwib CytH in RL HMt
H-005 00001 0.001 001 0.1 1 10
Concentration (cyltt/L)
ExNM E-19 CunlitvB D»tr*iu«
-------
Exhibit £-20 Cumulative Distribution of Total CiyptosporkHum Oocysts in Aft Plants
t
16-005 0.0001 0.001
0.01
0.1
10
Concentration (oocysts/L)
ICRSS (Large Plant:) Modeled Data
Bdlibil E-21 Cumulative Distribution of TMll Cryplosp«t(Hum Oocysls in All Plants
1e-005 0.0001
0001
Concentration (oocysts/L)
ICRSS (Medium Plants) Modeled Data
100
Exhibit £-22 Cumutabv* DwtributKin of Non-Empty Grypttiporidlum OocysM in AN PltnU
1.0
0.9
0.8
0.7
0.6
OS
04
i
3 0.3
0.2
0.1
' <-6[ot«0)mwntMDra
H-OOS 0.0001 0001
Coficantration (oocysti/L)
ICRSS (Large Plants) Moddlod Data
Occurrence and Exposure Assessment
for the LT2ESWTR
E-9
December 2005
-------
E*ib« E-23 CumuMM Distribution 0) Non-Empty Cwtosporidium Oocysts to All Plants
1^005 0.0001 0.001 001 0.1 1
Concentration (oocysts/L)
ICRSS (Medium Plaits) Modeled Data
100
ErfiW E-24 Cumulative Dfelrltutton of CwKnpcnfiini OocyAi xtti immul struflures in All Plants
18-005 00001 0.001
0.01
Concentration (oocy«ts/U)
ICRSS (Laroe Plants) Modeled Data
«-19 (erf 40) rtw rttt* at an
1*00$ 0.0001 0001
Conconfritlon (oocyiU/L)
ICRSS (Medium Plinli) Modeled Data
Occurrence and Exposure Assessment
for the LT2ESWTR
E-IQ
December 2005
-------
Exhibit E-26 Cumulative Distribution of Tolal Cryptosporidium Oocysts in Flowing Stream Plains
10-
0»
Off
f 0.7
1 fle-
et
| 0.5
1 0.4-
o O.J
0.2
O.t
00
EmpncalCDF. 24 pw *MB. 14 h» 40%
1e-005 00001
•« at «r« ' ' ' "
0.001 0.01 0.1 1
Concentration (oocysts/L)
ICRSS (Large Plants) Modeled Data
10
100
Exhibit E-27 Cumulative Distribution of Total Crypojpotrtum OocjsB in Plowing Stnum Plants
0.9-
I 0.7
| 08-
a OS
1 04-
0.2-
O.t
o.o-
4p* «tt.M|4or«%ncoMt(Y
^
1B-005 00001 0.001
001
0.1
10
100
Concentration (oocysts/L)
ICRSS (Medium Plants) Modeled Data
« DiMtbuten of N«n-£mpty C
00c)r«t» in Ftowing S
I
s
Q,
|
o
1.0 j.-
0.9
o.e
0.7 '
0.6 1'
0.5 /,'
!
0.4 t.j
0.3
0.2
01 I«w»0» Hmau «••»»%»<»»> /
/
oo ^ .
1e-OOS 00001 0001 0.01 0.1 t 10 100
Concentration (oocysts/L)
ICRSS (Large Plants) Modeled Data
Occurrence and Exposure Assessment
for the LT2ESWTR
E-I1
December 2005
-------
EjJiitX E-M CumuUlrve DotrtbuUon of Non-Empty Cryptospfxiduin Oocysu in Hmiing Stream Plants
09
oe
I °?
I oe
Q.
5 0&
•§ 0.4
o 03-
02-
or
oo-
1e-005 00001 0.001
0.01
10
100
Concentration (oocysts/L)
ICRSS (Medium Plants) Modeled Date
ExttiM E-30 CumulaUw DelribuUm of Cryptosponcfcjm Oocysts wWi Inteiml Sftudms in RoviinD Steam Ranis
1.0-
0.9
0.8-
16-005 0.0001 0.001
0.01
100
Concentration (oocysts/L)
ICRSS (Urge Plants) Modeled Data
EnfiiM E-3t CumulMvt D«tnbu1iOfl af CryptiMiniMluin OACyM* with Inttnul Stuctun» n Flowing SITMm Ptanl*
1.0 • ^
0.9 „ v
a.t '.-'"'
f °-7
| O.B
6t //
j 0.5
1 0.4
U 0.3 EBMO* „„.».„,„»,«,»»,
0.2
/
0.1 ...'
00 «.<(o(i7)r»wl»»»of»«l
10-005 0.0001 0.001 001 0.1 1 10
Concentration (oocysts/L)
ICRSS (Madlum Plmu) Hodeloa DM
Occurrence and Exposure Assessment
for the LT2ESWTR
E-12
December 2005
-------
Extilbi E-32 CunuMK* Distnbiition of Total Cn/ptosporttunn Oocysls in RtsenoULOa Plants
16-005 00001 0001
001
10
100
Concentration (oocysts/L)
ICRSS (Large Plants) Modeled Data
Exhibit 6-33 Cumulative Dislr&rtion of Total Crypospomfcim Oocysts ft ReservmrfUke Plants
le-005 00001 0.001
0.1
Concentration (oooysts/L)
ICRSS (Medium Plants) Modeled Data
Exhibit E-34 CtxnuMvt DirintHrton ol NDn-empty Crypte^poiidium OocyMI in RtMrvoirJUk* Plcntl
1
1
«
1
0
1 0
09 „"-"''
f
0.8 /
0.7 /
0.8 /.-'
0.5 ;/
0.4 , -/
0.3 ,' /
02 E.OK.OX ;.„». M«.«^^-.,, , ^
0.1
/
\»-OOS 00001 0001 001 01 1 10 100
Concentration (oocyata/L)
ICRSS (Langs Plants) Modeled Data
Occurrence and Exposure Assessment
for the LT2ESWTR
K-13
December 2005
-------
Erfiib* E-35 Cumulative Distribution ot Nov&nKy CiyptospwHium OocyMs >n RtsenaKLalia PtonU
10-005 0.0001
0.001
0.01
0.1
Concentration (oocysls/L)
ICRSS (Medium Plants) Modeled Data
EjJiM E-» Cumulative Distribution of CryiUwpondiuin Oocysb "«i Internal SVuclures h Res«VDii«Ult« Plaits
1e-OO5 0.0001
0.001
001
Concentration (oocystsA.)
ICRSS (Large Plants) Modeled Data
Exhibit E-37 Cimtulfttfv* Dntributton of Cf>ptMp«1dkim Oocyttt wtti InMrrul Sfructurtt in RtMfvoMLih* P
10 _^";'
/.-1""
0.8 / •
07 ,-.
06
5 0.5
1 °4
o 03
02
01
/
Concanlmtion joocylUA.)
ICRSS (M*dium Plmts) MocKied Date
Occurrence and Exposure Assessment
for the LT2ESWTR
E-14
December 2005
------- |