Adaptive Implementation of TMDLs Kenneth H. Reckhow Duke University

1. Adaptive Implementation of TMDLs Kenneth H. Reckhow Duke University

2. ASSESSING THE TMDL APPROACH TO WATER QUALITY MANAGEMENT Committee to Assess the Scientific Basis of the Total Maximum Daily Load Approach to Water Pollution Reduction Water Science and Technology Board Division on Earth and Life Studies National Research Council National Academy Press Washington, D.C. 2001

3. Contents (1) INTRODUCTION The Return to Ambient-Based Water Quality Management (2) CONCEPTUAL FOUNDATIONS FOR WATER QUALITY MANAGEMENT Ambient water quality standards, and decision uncertainty (3) WATERBODY ASSESSMENT Listing and delisting (4) MODELING TO SUPPORT THE TMDL PROCESS Prediction uncertainty (5) ADAPTIVE IMPLEMENTATION FOR IMPAIRED WATERS

4. Key Recommendations – Modeling and Implementation • Uncertainty must be explicitly acknowledged both in the models selected to develop TMDLs and in the results generated by those models. • The TMDL program currently accounts for the uncertainty embedded in the modeling exercise by applying a margin of safety (MOS); EPA should end the practice of arbitrary selection of the MOS and instead require uncertainty analysis as the basis for MOS determination. • TMDL plans should employ iterative, adaptive implementation and revision. • In order to carry out adaptive implementation, EPA needs to foster the use of strategies that combine monitoring and modeling and expedite TMDL development.

5. Model development is likely to proceed along the conventional lines: • Advances in process models will likely lead to increasingly elaborate mechanistic descriptions, with improvements expected. • More/better observational data, and advances in statistical techniques, will likely lead to gains in empirical model forecast accuracy. However, it is hard to believe that either of these will result in dramatic improvements (perhaps mechanistic/statistical hybrid models have more promise). The purpose of this presentation is to discuss another approach - using implemented actions on the real system as learning experiments to augment/improve model forecasts.

6. Water Quality Forecasting The problem with water quality forecasting is that we’re not terribly good at it. Result: Prediction uncertainty is high

8. Three Distinctly Different Models were Applied • CE-QUAL-W2 (NEEM; 2-dimensional) • EFDC-WASP (3-dimensional) • A Probability Network Model (Neu-BERN)

10. Yet, we need predictions to guide decision making, so what should we do? Adaptive Implementation We can “learn while doing;” that is, we can observe how the real system (the actual waterbody) responds, and then use that information to augment the prediction for the modeled system.

11. Why does this make sense? • If water quality model forecasts are quite uncertain, then chances are high that adjustments will be necessary (i.e., we’ll likely get it wrong on the first try). • If we know that we’re likely to be wrong with the initial water quality forecasts, then it makes sense to develop and implement management actions that: (1) help us learn about system response, and (2) are flexible to allow the necessary adjustments.

12. How might we conduct adaptive implementation? • Step 1: To define the allowable pollutant load, a water quality model is applied; the forecast from this model provides the initial estimate of how the waterbody will respond to the pollutant load reductions required. • Step 2: After the load reduction is implemented (i.e., nonpoint & point source pollution controls in place), a properly-designed monitoring & research program is established; this program can be focused on assessment of particular pollutant controls and/or on overall waterbody compliance with standards. • Step 3: The pre-implementation model forecast (from step 1) is combined with the post-implementation monitoring (from step 2); this provides the best overall estimate of success of the initial TMDL implementation.

13. Posterior (integrating modeling and monitoring) Sample (monitoring Data) Prior (model forecast) Adaptive Implementation: Bayesian Analysis Water Quality Criterion Concentration

14. What do we do if the initial post-implementation assessment still indicates noncompliance? • Improve the model using the new post-implementation data. • Continue with another round of pollutant load reductions, guided by the improved model.

15. A Recommendation for Improvements in the TMDL Implementation Process • Allow two forms of implementation: • Standard (or conventional) • Adaptive Standard implementation (SI) of a TMDL should occur when the level of certainty regarding causes, remedies, and water body condition is high, or when the costs of making an error in the face of uncertainty are deemed acceptable. Adaptive Implementation (AI) should occur where uncertainty is substantial and the costs of error are deemed significant.

16. What is meant by “error” and the “costs of error”? Standard implementation in the face of uncertainty creates the real possibility that strict adherence over time to the original implementation plan will cause resources to be spent on the controls at sources and locations that will not secure desired water quality outcomes. This scenario can be avoided using an adaptive implementation approach.

17. Three Possible Pathways for Adaptive Implementation • Implementing the TMDL where there is no question about the endpoint to be attained by the TMDL. • Implementing the TMDL where the applicability of the original water quality standards (designated use and/or water quality criteria) is questionable. • Implementing watershed management, where the impairment is the manifestation of stressors lying outside the requirements of the TMDL regulations (habitat disturbance, hydrologic modification, geomorphic alteration).

18. How can we make adaptive implementation work? • An organization is needed with the commitment of resources to the learning process. • There needs to be an initial implementation plan, a funding strategy to support the commitment of follow-on monitoring and modeling, and support for continuing stakeholder involvement that will achieve agreement on modifications to the implementation plan over time. • When there are point sources of pollutants, attention may need to be paid to possible accommodations for AI in the NPDES permitting process given that AI may result in modification to the TMDL or the WLA over time.

19. In conclusion, adaptive implementation begins with installation of certain controls that serve to move the watershed in the direction of reducing pollutant loads, while also providing information on their effectiveness in influencing water quality at different geographic and time scales. • With the new knowledge, the original watershed and water quality analyses and models can be revised. • This will allow updating of the estimates of current and future pollutant loads and the resulting water quality in the impaired water body as a result of revised control strategies (based on the revised model).