USGS Releases New Groundwater Model for Edwards Aquifer

Last July, the United States Geologic Service (“USGS”) released a new groundwater model for the Edwards Aquifer that represents a significant improvement over prior versions by:

  • addressing the interaction between freshwater and brackish groundwater zones and
  • introducing uncertainty analysis about the model’s predictions.

While the document serves as a technical report on groundwater model development, policy makers and water decision-makers can benefit from understanding the high-level themes.

In 2010, USGS in cooperation with the San Antonio Water System began a study of brackish groundwater movement in the Edwards Aquifer, including the potential for brackish water encroachment into freshwater wells near the freshwater-brackish water transition zone, and effects on spring discharge at San Marcos Springs and Comal Springs under drought conditions (the spring flows protected under the federal Endangered Species Act that became the basis for the Texas Legislature passing the Edwards Aquifer Authority Act to regulate groundwater pumping in the Edwards Aquifer).

Using a developed numerical groundwater model, the objectives of the study are to predict the effects of higher than average pumping under rainfall drought conditions on changes in total dissolved solids (salinity) at production wells near the transition zone, spring discharge at San Marcos and Comal Springs, and the elevation of well J-17 in Bexar County (the elevation of J-17 determines the amount of pumping allowed under Edwards Aquifer groundwater permits for the San Antonio Pool).  The predictions were presented in terms of a quantified uncertainty reflecting the potential variability in the values of the model’s parameters (see below).

Freshwater/Brackish Groundwater Interaction

The study concludes that there is a minimal increase in total dissolved solids concentrations for wells near the transition zone from higher than average pumping during the drought conditions of the 1950s.  The upper bound estimated increase was less than 200 milligrams per liter for 19 of the 25 wells (or 76% of the wells) and above 200 milligrams per liter but less than 400 milligrams per liter for the remaining 6 wells.  The use of upper bound estimates means that there is a 95% probability that the actual increase in dissolved solid concentration will be less than predicted levels.  The minimal increases in predicted dissolved solid concentration indicate a small potential for movement of the transition zone between freshwater and brackish groundwater.

Uncertainty Analysis

 The study’s use of uncertainty analysis is a major contribution to the understanding of the Edwards Aquifer.  While the specifics are complex (including excursions into Bayesian statistical decision-making), the concepts are simple.  As with other groundwater model studies, USGS uses “best fit” parameters (initial dissolved solid concentrations, hydraulic connectivity, specific storage, etc.) to match the model’s predictions to a calibration period; for the USGS study, 1999-2009.

Unlike most groundwater modeling efforts, the uncertainty analysis recognizes that “best fit” parameters are selected to minimize the discrepancy between the model’s predictions and observed conditions (spring discharges, concentrations of dissolved solids, well elevations and the like).  There is an underlying variability in the parameter estimates.  The study’s uncertainty analysis identifies the variability in the model’s predictions that reflect the underlying variability in parameter estimates.

The width of the 95% confidence interval (that is, 95% of the predictions fall between the lower and upper bound defining the confidence interval) measures the reliability of a model’s predictions.  USGS notes that its predictions for concentration of dissolved solids has a narrow confidence interval—indicating that the model’s predictions are reliable.

USGS finds that their model’s predictions for spring discharges and elevation of well J-17 are not reliable.  The upper bounds of the 95% confidence intervals for spring discharges and well elevation are an order of magnitude greater than the model’s predictions.  USGS concludes that this demonstrates that the model’s predictions of these factors are not reliable.

The unreliability of predictions reflects the fact that the parameters controlling the predictions are not well-identified by the “matching” of model predictions and observations during the calibration period.  Further, there may be greater complexity in the physical processes governing spring discharges and well elevations than accommodated by the groundwater model.

In the end, the USGS model can reliably predict the movement of dissolved solid concentrations and cannot reliably predict spring discharges and J-17 well elevations.


The USGS findings have significant implications for policy makers and water project operators.

Brackish Groundwater Legislation.  Development of brackish groundwater legislation received significant attention during the 2014 Interim Session in Texas.  I attended a hearing in Austin where lawyers dominated the presentations.  They demonstrated a remarkable (lack of) understanding about hydrogeology. My personal favorite was about the “commingling” of freshwater and brackish groundwater (how can commingled waters have significantly different salinities?).  At least for the hydrogeological conditions for the portion of the Edwards Aquifer studied by USGS, there seems to be significant separation between the various strata yielding freshwater and brackish groundwater to suggest that concerns about the impact of development of brackish groundwater on freshwater supplies may be overstated.  More scientists and fewer lawyers should be informing the Legislature about brackish groundwater issues.

Adaptive Management.  The USGS study does a great service by stressing the underlying uncertainty in predictions.  Water resource management is not a venture without uncertainty.  Therefore, projects operations should be based on a monitoring program that tracks how actual experience in the field compares with model predictions.  Triggers should be set for reconsidering project operations or design when actual experiences exceed defined thresholds.  In effect, a model’s predictions define a “working hypothesis” that is tested by actual experience.  Once the experience on the ground deviates sufficiently from predictions (taking into account the predictions underlying uncertainty), it would be time to recognize that a project is based on a deficient model.

Embrace Uncertainty.  Policy makers, decision-makers and investors understandably dislike uncertainty. However, they must face reality.  Water resource management is fraught with uncertainty.  Hydrologic conditions are volatile. Our understanding of natural resources is imprecise. Rather than ignoring uncertainty, responsible decision-making must confront it.

To that end, the USGS study makes an excellent contribution to better understanding the Edwards Aquifer by identifying what we can predict reliably (impact of pumping and drought on dissolved solid concentrations) and what we cannot predict reliably (spring flows and J-17 well elevations).

Final Comment

Since the enlightenment of the 17th century, western culture has embraced the role of reason in human affairs.  Investing in and improving our scientific understanding of water resources will improve water resource management, including the identification and management of risk.  As a developer of brackish groundwater projects, the San Antonio Water System made a sound investment in advancing the state of knowledge about the Edwards Aquifer.  While improvements in our understanding of the resource are still ahead, identifying what we reliably know today and what we do not know reliably today helps identify where the next investments in science should be made.


Written by Rodney T. Smith, Ph.D.