Nonpoint source pollution control, incomplete information and learning: An entropy approach

Since the passage of the Clean Water Act, most pollution control measures have focused on the problem of reducing point source pollution. However, water resources in the United States are predominantly polluted by nonpoint sources (NPS) which, by definition, are more difficult to measure. For this reason, there is a need to study NPS pollution control.; NPS pollution control is primarily an information (or uncertainty) problem. By definition, the dominant characteristic of NPS pollution is that information on the linkage between source and load is incomplete, creating uncertainty about the efficient treatment level. Therefore, a realistic analysis of NPS pollution control requires that the role of information and learning be explicitly specified and addressed.; The analysis starts with the construction of a theoretical NPS pollution control model, depicting the behavior of a budget-constrained manager who minimizes the pollution-related damage. With this model, we examine the tradeoff between the level of information acquisition on pollution loading and treatment of the pollution loading sources, and derive optimal data collection and treatment strategies.; Next, a sequential entropy filter (SEF) is developed and applied to the problem of estimating the NPS pollution loading from among various sources. The SEF accommodates ill posed NPS pollution data where the number of sources exceeds the number of observations. We show how the SEF nests traditional Bayesian updating approaches, and provide Monte Carlo simulations to illustrate the ability of the SEF to reconstruct pollution loading estimates from known pollution loading processes.; Finally, the theoretical and methodological results are incorporated into a sediment control model for Redwood National Park. In this application the SEF is applied to stream flow and ambient sediment loading data for Redwood Creek, which flows into and through Redwood National Park. This estimation is followed by a simulated policy analysis of the sediment control management program in Redwood National Park. We compare a uniform treatment policy where no data is collected versus high and low intensity data collection policies. The results indicate that the high intensity data collection sufficiently reduces uncertainty so that sediment related damages decrease below damages under the uniform treatment policy.