|
publications > report > RECOVER southern estuaries performance measures: identification of hydrology - salinity relationships for coastal estuaries and analysis of interim CERP update scenarios
Critical Ecosystem Studies Initiative Final Project Report
RECOVER Southern Estuaries Performance Measures: Identification of Hydrology - Salinity Relationships for Coastal Estuaries and Analysis of Interim CERP Update Scenarios
Contract No. DACW17-02-D-0009 FINAL PROJECT REPORT Principal Investigator: Frank E. Marshall III, PhD, P.E. September, 2005 Environmental Consulting and Technology, Inc. 340 North Causeway New Smyrna Beach, Florida 32169 (386) 427-0694 Table of Contents I. BACKGROUND II. FLORIDA BAY ANALYSIS III. BARNES SOUND AND MANATEE BAY ANALYSIS IV. ICU RUNS AND PERFORMANCE MEASURES V. POST-PROCESSING ACTIVITIES VI. REFERENCES
List of Tables
1 Summary of Information about the monitoring Stations and Salinity Data
List of Figures
1 Study Area
I. BackgroundA. GeneralEnvironmental Consulting & Technology, Inc. (ECT) working as a sub-consultant to Tetra Tech Inc., has been contracted to support the U.S. Department of the Army, Corps of Engineers, Jacksonville District RECOVER branch with the identification of hydrology - salinity relationships needed to create Performance Measures (PMs). These PMs are to be used to evaluate water delivery alternatives for the Comprehensive Everglades Restoration Plan (CERP) by employing hydrologic model output from the South Florida Water Management Model (SFWMM or 2X2 Model) to predict salinities downstream in coastal estuaries. Translating the water level measured by gauges and other hydrologic structures into salinities in coastal basins requires an analysis of statistical relationships between historical hydrology and resulting salinities in the subject south Florida water bodies. Other factors may also be important in quantifying these relationships, such as rainfall, wind and sea level. Once it is determined which gauges and structures are statistically significant for each basin, multivariate linear regression (MLR) models are to be developed for basins in Florida Bay, Barnes Sound / Manatee Bay, and the Shark Slough discharge area north of Cape Sable. These new models are to be combined with the existing MLR models prepared previously for Everglades National Park (ENP) by Cetacean Logic Foundation, Inc. to create a suite of models that describes the temporal and spatial variation in the south Florida area that receives drainage from the Everglades. The Regional Evaluation Team (RET) of RECOVER can then use the suite of models to simulate the salinity of coastal basins from SFWMM model output for various CERP project alternatives, thereby generating a key piece of information (salinity regime) on the effect of projects on the ecology of the embayments and open-water basins of Florida Bay. The Southern Estuaries Sub-team (Sub-team) of RET has been assigned with the responsibility of developing salinity performance measures that utilize the statistical models to evaluate alternatives for the Interim CERP Update. The Sub-team recommended the work described in this report. The work described in this report builds on the previous salinity model work that has been done in some of the coastal basins of Florida Bay by ENP. As research sponsored by the Critical Ecosystem Studies Initiative (CESI), multivariate linear regression (MLR) models were investigated for use in simulating salinity (Marshall, 2003a; Marshall, 2004) and were found to be capable of providing reasonable daily estimates of salinity in the near shore embayments that were investigated. The techniques were successfully applied for the first time in the evaluation of alternative water delivery schemes as part of the Interim Operations Plan (IOP) Congressional report activities (Marshall, 2003b, 2005). For this project, four tasks were completed, each resulting in the preparation of a draft task report that was distributed electronically. These task reports were incorporated into this single-volume project report. The four tasks that were completed were: (1) the Florida Bay analysis, Significant additional information was added to the task reports to complete the compilation of the four task reports into this Project Report. B. Hydrology and Other Factors Affecting SalinityFreshwater flowing into ENP across Tamiami Trail moves south and southwest towards Florida Bay and the southwest Gulf of Mexico coast, as shown on the study area map, Figure 1 and the more detailed Figure 2. Most of the freshwater flows through Shark River Slough into the Gulf, but some freshwater passes through Taylor Slough, and is distributed into northeast Florida Bay and Barnes Sound / Manatee Bay / Card Sound, augmented by flows from the C-111 Canal system and seepage from stored surficial aquifer groundwater beneath the topographic high of the Atlantic Coastal Ridge on which metropolitan Miami is built. The contribution of groundwater and the effect on salinity variation in Florida Bay and Biscayne Bay is not fully understood.
Rainfall in south Florida typically exhibits distinctive wet and dry season patterns but is variable spatially. The wet season pattern is driven by tropical weather and sea breeze interactions, with frequent storms that exhibit wide spatial distributions. Rainfall data between gauges that are only tens of miles apart often show widely varying rainfall volumes for a single period of time during the wet season (May or June through October or November). In the dry season (November or December through April or May), rain is usually delivered by frontal systems, with wider spatial distribution of single events, and less frequent events compared to the wet season. Evaporation is also seasonal, but with a slightly different pattern than rainfall. This difference in timing has a significant effect on the timing and quantity of fresh water delivered and therefore the salinity. Wind also has a seasonal pattern similar to rainfall. During the wet season the wind direction is predominately from the south and southeast, and during the dry season northerly wind is most common. Additionally, the sea surface elevation and the elevation of the water surface in Florida Bay have their own “seasonal” behavior, with the highest average water surface elevations in the Fall and a secondary high in the Spring. When a hurricane or other significant weather event occurs, wind and storm surge effects can alter the sea surface elevation dramatically over short periods. According to coastal aquifer theories, the salinity in the interface zone between strictly fresh (0-5 psu) and marine waters (33-35 psu) in Florida Bay is influenced by the freshwater head in the upland watershed (measured as the elevation of water in wells in the Everglades) competing with the elevation of water in Florida Bay and the Atlantic Ocean (Pandit, et al; 1991). The result, when combined with the physical features of the embayments, basins, and bays of Florida Bay (including the shallow banks) and wind effects is a complex situation for salinity variability. To-date, only statistical models have been capable of successfully simulating this variability on a daily basis. C. Data for Model Development and SimulationsIn choosing the data that are to be included in the analysis and ultimately the MLR salinity models, the end use of the models has to be considered. The SFWMM or 2X2 model has produced estimates of stage (water level) and flow of freshwater through the Everglades for a number of CERP scenarios. Daily values are available for 31- and 36-year overlapping periods. The 36-year (1965-2000) runs are of interest for this study. The effects of the specified water deliveries on the water levels in the Everglades are expressed in the output of each of the CERP 2X2 runs. The MLR salinity models were developed to use the 2X2 model output in conjunction with available long-term data for wind and sea surface water level to produce estimates of daily salinity for the 36-year period in Florida Bay, the southwest Gulf coast, and Barnes Sound and Manatee Bay. The simulated salinity time series can be analyzed for potential ecological impacts, either positive or negative, of the particular water management alternative. The independent variable data must be available for most, if not all of the 36-year period in order to populate the models and obtain estimates of salinity to be of use for the ICU evaluations. This requirement eliminates the direct use of evaporation for models or simulations since there are no observed daily data for the 36-year period, nor are there any evaporation models that are capable of producing reliable daily estimates. Flows through control structures are not as useful statistically for model development and simulation purposes compared to stages (water levels) in the Everglades. Although rainfall is an important hydrologic parameter for seasonal salinity variation, rainfall at monitoring stations in the Everglades are not highly correlated with salinity at the daily level. Instead, the stochastic effect of rainfall falling on the Everglades and the upstream watershed is integrated by the coastal aquifer system and expressed adequately in stage data. Wind speed and direction and sea level are highly correlated with salinity at the daily level, and are available for the 36-year period. For the above reasons, the data that were used to develop the MLR salinity models and generate the simulations include the following:
For model development, observed stage data are used. For simulations, the 2X2 Model output data are used for stage after adjustments are made. Model development and simulations use the same data bases for wind and sea level although the period of the simulation is longer. Continuous salinity data extend back to 1988 at several locations in northeast Florida Bay. However, some of the stations have only been operational since 1996. For the previously developed CESI / IOP models the period of data used for model development begins on March 24, 1994. For the new models in this study, the longest period of data available was used. The period of record for all stations extends through October 31, 2002. Most series contained some missing values. No attempts were made to fill in data gaps or to eliminate outliers in either independent or dependent variable data sets, as the number of daily values for the shortest time series for the observed data exceeded 2000 values. The models were developed from observed data that have been collected at 15 to 60 minute increments and averaged to daily values. Salinity data are taken from the ENP Marine Monitoring Network (MMN) data base, Table 1. Details about these data can be found in Everglades National Park (1997a and 1997b), and Smith (1997, 1998, 1999, and 2001). A map showing the ENP MMN stations and the locations of the water level monitoring stations used for this study is presented as Figure 2. Wind data were obtained directly from the National Weather Service (Southeast Regional Climate Center) for Key West and Miami stations, and sea surface level data collected at Key West were obtained from the National Ocean Service website (Table 2). Wind data from Key West and Miami were used as these locations had the longest continuous records for wind and were considered to be representative of the regional wind patterns. Sea surface elevation data from Key West were considered to be representative of the average effect of oceanic water level influences, and, to some extent, the average water level patterns within Florida Bay. The stage data are ENP Physical Monitoring Network Everglades water levels. A limited number of continuous water level (stage) monitoring stations in the Everglades began recording data in the 1950’s (see Table 2), but most stage records date from the 1990’s.
D. Model DevelopmentThe framework for the relationship between estuarine and coastal shelf salinity is a coastal aquifer system physical model with a dynamic balance between fresh and salt water bodies and a salinity transition zone from upstream freshwater (salinity = 0) to sea water (salinity = 35 psu; Pandit et al, 1991). In most of the coastal aquifer examples in the literature, the focus is the water table aquifer, with the primary concern being the location of the transition zone as a water supply issue of saltwater intrusion. For salinity modeling in an estuary the focus is the salinity in the interface transition zone. The well-known Ghyben-Herzberg principle describes the location of this interface as function of the height of the freshwater surface in the watershed relative to the height of the sea surface above a common datum, and the relative density of the water masses. When the sea surface level is high enough relative to the freshwater level (such as a normal dry season), the higher density salt water moves the interface landward, increasing the salinity at a fixed station in the transition zone. When the freshwater level is high enough relative to the elevation of the sea surface to overcome the increased density of the seawater (such as a typical wet season in the Everglades), the salinity will decrease at a fixed station in the transition zone. In a shallow estuary like Florida Bay the wind can cause the interface to translocate and also to mix. Therefore, it is reasonable to expect that there would be a correlation between salinity levels and these three factors, which is confirmed by a correlation matrix of the observed data (not presented) at the 95% level of significance, sometimes for lagged values on the order of days. However, each of these forcing factors (fresh water elevation, wind, sea surface elevation) has a different pattern of variability over time. Figure 3 presents a plot of water levels at Craighead Pond (CP) and at P33, and sea surface elevation at Key West for the period March 1971-March 1983. Though the values of CP and P33 cannot be compared directly with the values of Key West sea surface elevation, the variability patterns in each time series can be evaluated. It can be seen that the variability in the sea surface elevation (Kwwatlev) is more uniform from year-to-year than the year-to-year variability in CP and P33, reflecting the regularity of the harmonic components of the sea surface elevation. While the values of CP and P33 are, in general, lower in the dry season (November – May) and higher in the wet season (June – October), it can be seen that the variability in the minimum value reached for a dry period is greater than the variability in the maximum value reached, reflecting the importance of wet season rainfall, or rather the paucity of it. For example, in Figure 3 the minimum water level reached at CP for 1974 and 1975 is much lower than the values reached for 1976 and 1977. A similar situation can be seen for P33 for 1974 compared to the following years. However, for 1975, CP behaves similar to 1974 (both expressing drought conditions), while at P33 the water level is relatively high in the dry season in 1975.
In addition to the variability of watershed and sea / estuary water levels is the variability in wind speed and direction, used in the salinity models as vector quantities. Figure 4 presents the daily time series for 2000 for the u-component of the wind vector, and Figure 5 presents the same information for the v-component. Although vector data are sometimes difficult to interpret, it can be seen that the patterns of daily variability are very similar at the two stations for the v-component. For the u-component, the positive and negative values for the Key West data are larger, and the variability pattern during the wet season (May – September) is different than the rest of the year.
A step-wise multivariate linear regression process was used to determine the most appropriate linear combination of independent variables for each salinity model. In addition to the independent variables in Table 2, the list of potential variables for model inclusion also included several hydraulic gradient variables computed using the stage variables in Table 2. To begin the model development procedure, all independent variables were subjected to a cross-correlation analysis with daily salinity using SARIMA techniques to determine which of the variables were correlated with salinity, to check for lagged relationships, and to evaluate the level of correlation. Lags up to 50 days were initially reviewed, though it was found that lagged correlations never exceeded six days. Then the observed data of the significant correlated variables (current and lagged values) were input to a SAS© PROC REG routine that uses a step-wise regression process to identify the most statistically significant parameters for a multivariate linear regression equation. To ensure that only the most highly significant parameters were selected by this process, the significance level for parameter inclusion in the model was set at 99.9%, a very high level. Parameter inclusion in a model was also manually controlled by eliminating any seemingly correlated variables that acted contrary to known physical relationships (such as an increasing stage in the Everglades indicating an increase in salinity) which can occur when there are cross-correlation effects. These parameters were eliminated, and the step-wise process re-run iteratively. For some of the open-water salinity monitoring stations that are away from the direct influence of the freshwater in the watershed it was found that salinity models were improved when Everglades stage was replaced by the salinity at the near shore stations of Little Madeira Bay and Terrapin Bay. The following CESI / IOP salinity models were developed from Little Madeira Bay and Terrapin Bay salinity, wind vectors, and Key West sea surface elevation instead of the stage elevation in the Everglades:
During model verification it was determined that the salinity estimates produced by these models were more closely simulating the observed values compared to models prepared using watershed water levels. The details on model development can be found in Marshall (2003b, 2004, and 2005). However, it means that simulation is a two-step process with the simulation of salinity at Little Madeira Bay and Terrapin Bay required before salinity at the open-water stations can be simulated.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
U.S. Department of the Interior, U.S. Geological Survey
This page is: http://sofia.usgs.gov/publications/reports/hydrosal_relation/index.html
Comments and suggestions? Contact: Heather Henkel - Webmaster
Last updated: 18 November, 2008 @ 04:17 PM (TJE)
