Assessing the Impact of BMP Performance Uncertainties in TMDL Allocation

Abstract

Uncertainty originates from the variability in efficacy of the control measures, or Best Management Practices (BMP), used to implement a Total Maximum Daily Load (TMDL) allocation. A sustainable TMDL allocation should take into account this source of uncertainty at the implementation phase. This study characterizes the uncertainties in the pollutant removal efficiencies of BMP’s and BMP treatment trains, followed by development of equally likely BMP performance scenarios, expressed as a percentage of the reduction goals. The Latin Hypercube Sampling technique is utilized to generate BMP performance distributions. The distribution parameters are chosen based on various published pollutant removal efficiency values so that the efficiencies would closely represent the efficiencies exhibited by BMPs in practice. The cumulative efficiencies of BMPs within a BMP treatment train, which are used for high reduction goals, are quantified as well. The watershed model (HSPF 11.0) and Generalized Likelihood Uncertainty Estimation (GLUE) technique has been used in previous studies to characterize mechanistic parameter uncertainty (Yanbing 2004) and loading rate uncertainty (Foraste 2006). A genetic algorithm coupled with response matrix is then used to perform robust optimization of TMDL allocations and predict the reliability of compliance associated with each allocation policy. This study modifies the robust optimization framework to couple BMP performance uncertainty with mechanistic parameter uncertainty or loading rate uncertainty. This methodology is demonstrated for fecal coliform contamination in the Moore’s Creek watershed in Virginia. The results are then compared to previous studies that assumed deterministic BMP performance.