Evaluation and Comparison of Observed and Modeled Precipitation Forcing to a Distributed Hydrological Model to Simulate a Regional Monsoon Flood Event

This study evaluates the distributed simulations of hydrological responses (i.e. streamflow, soil moisture patterns, groundwater levels and evapotranspiration) produced by the TIN-based Real-time Integrated Basin Simulator (tRIBS) for a monsoon flood event in Upper Rio Puerco (1168 km2 area), central New Mexico, during September 2003. Observed precipitation from the NEXRAD Stage III product and a rain gauge network, and modeled precipitation produced by the Fifth-Generation NCAR / Penn State Mesoscale Model (MM5) are separately used as inputs to force tRIBS for the period prior to and including the major monsoon storm leading to the flood event. Comparisons of the simulated hydrological responses from the various precipitation products are made using streamflow and soil moisture statistics. Prior to the simulations, MM5 precipitation is compared with that from NEXRAD estimates using a variety of spatiotemporal statistics. The results demonstrate large discrepancies in terms of rainfall intensity and spatiotemporal characteristics of the rainfall field, primarily due to delays in convective initiation and organization at the large scale in the numerical model. Simulation results for streamflow forced by the observed and modeled rainfall are also compared with those observed through USGS streamgauge observations. The results provide some insights for improving the use of atmospheric model output in hydrological forecasts for monsoon periods in large semiarid basins. For instance, calibrating tRIBS with observed precipitation can provide a good estimate of initial soil moisture due to previous storms prior to using atmospheric forecasts of the major storm event. This preliminary work demonstrates the strengths and weaknesses of using atmospheric output in distributed hydrologic models and motivates further study in the two-way coupling of atmospheric and hydrological models.