Managing Water Quality in the Face of Uncertainty: A Robust Decision Making Demonstration for EPA’s National Water Program

Culvert with a drop, with flowing water, photo by Sickter6/CC BY-SA 3.0

Providing safe and reliable drinking water supplies and managing water quality and ecosystem health present a challenge of decision making under conditions of deep uncertainty.

Maintaining safe and reliable supplies depends on timely investments in water treatment, storage, and delivery infrastructure, and the success of these investment decisions relies on predictions about water availability, system needs, and infrastructure performance many decades into the future.


RAND has undertaken a project for the U.S. Environmental Protection Agency (EPA) to determine the utility of Robust Decision Making (RDM) methods for evaluating the agency's needs and priorities under the National Water Program (NWP). Two case studies on RDM applications on NWP activities have been completed—one in the Patuxent River in Maryland and the other in the North Farm Creek tributary of the Illinois River. These case studies show how analytic RDM methods can be used to identify future vulnerabilities in total maximum daily load (TMDL) implementation plans and suggest appropriate responses. A third case study that focuses on the Los Angeles River in Southern California is currently underway.

The Los Angeles water quality case study will also examine the potential vulnerability to climate change of TMDL implementation plans for its featured Los Angeles’ San Fernando Valley and potential responses. However, in contrast to the previous two case studies, this one will explore a region with very different hydrology. This case study will also expand on the treatment in the other two case studies of the role adaptive management in enhancing TMDL implementation plans and examine how coordination among jurisdictions managing water in the same geographic area can more effectively address water quality goals.

Research Questions

  1. How might Robust Decision Making help the EPA and its partners to improve TMDL implementation planning in the face of imperfect models, climate change, and other uncertainties?
  2. What assumptions about future hydrology, land use, or water quality best management practice effectiveness most often lead proposed watershed implementation plans to exceed TMDL water quality standards for the Patuxent River in Maryland or the North Farm Creek tributary of the Illinois River respectively?
  3. What types of additional investments might help reduce the vulnerabilities identified, and when might planning processes need to be reconsidered?

Key Findings

Climate Change and Other Uncertainties Could Have a Significant Impact on the Success of Water Quality Plans

In both pilot regions studied, TMDL watershed implementation plans expected to meet water quality standards if future climate resembles the past do not meet these standards over a wide range of plausible future climate conditions. An increase in the amount of paved or impervious area cover due to future population growth, as well as worse-than-anticipated performance from stormwater best management practices (BMPs), could also lead to missing future standards.

Rainfall-Runoff Simulation Models Can Provide Useful Information for TMDL Planning Under Deep Uncertainty

Used within a RDM framework, these models can help decisionmakers explore the performance of TMDL plans across many plausible paths into the future.

Currently Available Simulation Models Are Suitable to Support RDM Analyses, but Much Could Be Done to Improve Their Utility

Treatment of BMP performance in the rainfall-runoff models could be improved, with better accounting for uncertainty related to BMP effectiveness. Models could be streamlined to better integrate with simulations of a more complete range of biophysical and socioeconomic processes and to facilitate scanning over many possible futures.


  • To improve and maintain high water quality standards in changing, often difficult-to-predict conditions, EPA and its partners will need to employ iterative risk management and rely increasingly on robust and flexible implementation plans.
  • Future analysis could help make TMDL plans more robust by considering a wider range of potential uncertainties and a richer set of response options. The treatment of adaptive TMDL implementation plans could be considerably expanded.
  • Improvements to the analytic tools used in this study could significantly improve the effectiveness of the decision support available to water quality planners. Packaging such tools in more user-friendly and potentially web-accessible toolkits could help make these methods widely available to decisionmakers at the local, state, and regional levels.