Texas Source-Water Susceptibility Assessment

In many earlier USGS investigations, strong relations between the occurrence of agricultural chemicals in streams and their use, or land-use within the up-stream watershed have been noted. In some investigations (Mueller et al., 1997; Battaglin and Goolsby, 1997; 1998) statistical models are used to describe these relations. One type of model, logistic regression, estimates the probability that a particular chemical will be above or below a specified concentration. Chemical use patterns, streamflow, land-use within the up-stream watershed, physical characteristics of the watershed, climate, soil type, and other factors are the independent variables in these models, which are conditioned using available water-quality data. Once developed for a region and if sufficiently accurate, logistic regression models can be used to model the probability of chemical occurrence at sites where water-quality data are not available. The logistic regression models developed by RPE researchers for Texas are one part of a tool and developed by Randy Ulery and others at the Texas office of the USGS to support the Texas Natural Resources Conservation Commission (TNRCC) Source Water Assessment Program efforts to assess the susceptibility of all Texas public-water supplies to contamination.

In the State of Texas, both surface water (streams, canals, and reservoirs) and ground water are used as sources of public water supply (PWS). Surface water sources of public water supply are susceptible to contamination from point and nonpoint sources. To help protect sources of drinking water and to help develop protective yet cost-effective and risk-mitigated monitoring strategies, the Texas Commission on Environmental Quality (TCEQ) and the U.S. Geological Survey (USGS) developed procedures to assess the susceptibility of PWS source waters to the occurrence of 227 contaminants. One component of the assessments is the determination of susceptibility of surface water sources to nonpoint source contamination. To accomplish this, water quality data at 323 monitoring sites are matched with geographic information system (GIS) derived watershed characteristic data for the watersheds above the sites. Logistic regression models are then developed to estimate the probability that a particular contaminant will exceed a TCEQ-specified threshold concentration. Logistic regression models are developed for 63 of the 227 contaminants. Of the remaining contaminants, 106 are not modeled because monitoring data are available at less than 10 percent of monitoring sites; 29 are not modeled because there are less than 15 percent detections of the contaminant in the monitoring data; 27 are not modeled because of the complete lack of monitoring data; and two are not modeled because threshold values are not specified by TCEQ.

Results of this research are available on in two publications. “Estimating the susceptibility of surface water in Texas to Nonpoint-Source Contamination by Use of Logistic Regression Modeling”, by William Battaglin, Randy Ulery, Thomas Winterstein, and Toby Welborn was published as U.S. Geological Survey Water Resources Investigations Report 03-4205. “Susceptibility of Surface water in Texas to Nonpoint source Pesticide Contamination", by William Battaglin, Randy Ulery, Toby Welborn and Thomas Winterstein was published as a proceeding article for the American Water Resources Association 2003 Spring Specialty Conference, Agricultural Hydrology and Water Quality, May 12-14, 2003 in Kansas City, MO. Maps of monitoring data and modeling results are available for the 63 modeled contaminants and a table of the logistic regression models are also provided on this web site.