ABSTRACT

The problem of water quality management under uncertain emission levels, reaction rates, and pollutant transport is considered. Three performance measures – reliability, resiliency, and vulnerability – are taken into account. A general methodology for finding a cost-effective water quality management program is developed. The approach employs a new stochastic branch and bound method that combines random estimates of the performance for subsets of decisions with iterative refinement of the most promising subsets.

INTRODUCTION

Devising successful and cost-effective water quality management strategies can be difficult because the inputs to, and the behavior of, the system being managed are never entirely predictable. Decision makers do not know what conditions will exist in the future nor how these conditions will affect the impact of their decisions on the environment. Vincens, Rodriguez-Iturbe, and Schaake (1975) classify uncertainty in modeling hydrologic systems into three categories: uncertainty in the model structure (Type I uncertainty); uncertainty in the model parameters (Type II uncertainty); and uncertainty resulting from natural variability (Type III uncertainty). For water quality systems, uncertainty in the pollutant transport model, the model reaction rates, and the natural variability of emission rates and receiving water conditions, such as streamflow, temperature, and background pollutant loadings from unregulated pollution sources, contribute to difficulties in predicting the future behavior of the system (Beck 1987). This chapter develops an approach for identifying water quality management solutions under Type II and Type III uncertainty. It is based on an application of the stochastic branch and bound method of Norkin, Ermoliev, and Ruszczyński (1994) to water quality management, which is modified to account for the performance indicators of reliability, resiliency, and vulnerability.