Water quality problems associated with agricultural nonpoint-source pollution remain significant in the majority of US watersheds. In this dissertation, I present a theoretical model of water quality that captures the main characteristics of agricultural pollution (the unobservability and the interactions between the field-level emissions, the imperfect knowledge of the abatement costs), propose and empirically estimate a simplified proxy model for the complex process that characterizes the fate and the transport of agricultural pollutants, and apply this model in a variety of empirical studies to evaluate alternative policy programs designed to improve water quality. Under a linear approximation of the abatement function, more flexible policies like the performance standard or trading program may outperform a command-and-control program in terms of abatement costs, but they may also result in the non-attainment of the abatement goal. However, the incentive-based policies can overcome, partially or totally, the issue of cost asymmetries, since the regulator does not need to know the farm-level abatement costs.

I propose and estimate an approach for linearizing the abatement function using a system of point coefficients that measure the impact of an abatement action on the overall abatement level. The point coefficients are estimated for nitrogen and phosphorus with consideration that the two pollutants have separate abatement functions.

The empirical assessments of the proposed policies for two agricultural watersheds in Iowa show an overall good performance of the incentives based programs: the deviations from the abatement goals are not significant and sizable cost savings relative to the command-and-control programs are realized. A robustness analysis shows that the results are consistent across different: (a) pollutants (nitrogen and phosphorus), (b) sets of point coefficients (field-specific level, subbasin-specific, or watershed-specific), and (c) the distribution of historical weather. The point approximation procedure is extended to two pollutant markets, where each market uses a separate set of point coefficients. Given that the same abatement actions that have the potential to increase the amount of carbon sequestration in soil, the point-based trading program is extended to allow trading participants to enter a market for carbon, including selling the carbon offsets associated with the abatement actions.