Updated numerical model with uncertainty assessment of 1950-56 drought conditions on brackish-water movement within the Edwards aquifer, San Antonio, Texas

Linzy K. Brakefield
Jeremy T. White
Natalie A. Houston


In 2010, the U.S. Geological Survey, in cooperation with the San Antonio Water System, began a study to assess the brackish-water movement within the Edwards aquifer (more specifically the potential for brackish-water encroachment into wells near the interface between the freshwater and brackish-water transition zones, referred to in this report as the transition-zone interface) and effects on spring discharge at Comal and San Marcos Springs under drought conditions using a numerical model. The quantitative targets of this study are to predict the effects of higher-than-average groundwater withdrawals from wells and drought-of-record rainfall conditions of 1950–56 on (1) dissolved-solids concentration changes at production wells near the transition-zone interface, (2) total spring discharge at Comal and San Marcos Springs, and (3) the groundwater head (head) at Bexar County index well J-17. The predictions of interest, and the parameters implemented into the model, were evaluated to quantify their uncertainty so the results of the predictions could be presented in terms of a 95-percent credible interval. The model area covers the San Antonio and Barton Springs segments of the Edwards aquifer; the history-matching effort was focused on the San Antonio segment. A previously developed diffuse-flow model of the Edwards aquifer, which forms the basis for the model in this assessment, is primarily based on a conceptualization in which flow in the aquifer is predominately through a network of numerous small fractures and openings. Primary updates to this model include an extension of the active area downdip, a conversion to an 8-layer SEAWAT variable-density flow and transport model to simulate dissolved-solids concentration effects on water density, history matching to 1999–2009 conditions, and parameter estimation in a highly parameterized context using automated methods in PEST (a model-independent Parameter ESTimation code). In addition to the best-fit parameter values derived from history matching, the uncertainty