Edwards Aquifer Groundwater Model Uncertainty Analysis

Science Center Objects

USGS Texas Water Science Center (TXWSC) is undertaking a 1.5-year study to assess parameter and predictive uncertainty in the Edwards Aquifer Authority MODFLOW Model using both linear and non-linear techniques. The Edwards Aquifer Authority (EAA) uses two models simulating the periods from 2001 to 2015 (verification model) and the drought of record of 1947-1958 (drought of record model).

Currently, EAA uses their numeric groundwater model to evaluate the efficacy of various aquifer management strategies to maintain springflow at Comal and San Marcos springs, as these springs provide critical habit for several threatened and endangered species. While the EAA groundwater model is in agreement with the conceptual understanding of the aquifer system and is well-calibrated to a variety of aquifer condition measurements, uncertainty may exist in the model-simulated springflow. This modeled response to management strategies could be misinterpreted or misused since the disposition of hydrogeologic properties and boundary conditions are not known with certainty. Quantifying and characterizing the uncertainty in the model inputs and propagating this uncertainty to model forecasts, such as the simulated springflow response to a given management strategy, is important to the implementation of the Edwards Aquifer Habitat Conservation Plan (EAHCP) strategy (Rieble and others, 2015) and can provide valuable insights on the reliability of other forecasts made with model.

To better understand the reliability of the Edwards Aquifer Authority groundwater model, we will evaluate the uncertainty of the model inputs such as

  • hydraulic conductivity,
  • storage,
  • recharge,
  • spring drain conductance,
  • hydraulic flow barrier conductance, and
  • Trinity Aquifer inflows.  

The results of this analysis could be readily applied to any model forecasts of interest stemming from the current suite of EAA Edwards aquifer groundwater models.

Quantification of the uncertainty that exists in the EAA groundwater models will require several tasks, each one of which will require close cooperation and coordination with EAA staff.

  • Uncertainty will be conceptualized in a Bayesian framework, where model input uncertainty is first described based on the existing knowledge about the Edwards aquifer. This “prior” uncertainty (statistical) distribution encapsulates and embodies the current state of understanding of the Edwards aquifer system.
  • Following the construction of the prior uncertainty distribution, the “posterior” model input uncertainty (statistical) distribution will be characterized using both linear and non-linear methods. These linear and non-linear methods will also be applied to two model forecasts, which will be determined by EAA staff.