Evaluation of NEXRAD precipitation estimates and their potential use for nonpoint source pollution modeling

Watershed models of nonpoint source (NPS) pollution are increasingly being used for watershed planning and management. It is well known, however, that such models are plagued with uncertainty. This uncertainty has many sources: inadequate model structure, poorly defined parameter values, and error in data inputs. The research contained in this dissertation investigates the role of precipitation sampling in models of NPS pollution. In particular, precipitation estimates from the Next Generation Weather Radar (NEXRAD) system are evaluated and compared with gage data for NPS modeling.; First, the radar-only hourly digital precipitation product (HDP) is evaluated for the complex terrain of the northern Appalachians. This evaluation demonstrates how beam blockage, nondetection and underestimation, and isolated ground returns are significant problems for radar-based precipitation estimation in mountainous terrain. As a result, these estimates may only be useful for event-based NPS simulation. This dissertation also presents an evaluation of the NEXRAD multisensor product for the Arkansas-Red Basin River Forecast Center (ABRFC). This research shows that the multisensor algorithms effectively remove range dependent radar biases. An attempt to evaluate the error variance of these estimates is presented, although the results are inconclusive due to inadequate raingage data.; Next, this dissertation presents a modeling framework using the Hydrologic Simulation Program---FORTRAN (HSPF) to compare precipitation inputs in the context of NPS modeling. HSPF was selected because it is functionally equivalent to the BASINS NPSM developed by the EPA for TMDL development. In this framework, HSPF is combined with automated calibration techniques to objectively compare the operational NEXRAD multisensor estimates from the ABRFC and raingage data, to assess the effects of gage sampling on model calibration, and to determine the impact of precipitation spatial resolution. In general, the gage data lead to better calibration results than the NEXRAD product, except when only a few distant gages are available. Gage sampling has a significant impact on the calibration of NPS model hydrology and on the simulation of water quality variables. Precipitation resolution, however, does not have much impact on the calibration of model hydrology, but does affect the simulation of surface dominated NPS contaminants.