Estimating the Long-Term Hydrological Budget over Heterogeneous Surfaces

Abstract

Estimates of the hydrological budget in the Walnut River Watershed (WRW; ?5000 km2) of southern Kansas were made with a parameterized subgrid-scale surface (PASS) model for the period 1996?2002. With its subgrid-scale distribution scheme, the PASS model couples surface meteorological observations with satellite remote sensing data to update root-zone available moisture and to simulate surface evapotranspiration rates at high resolution over extended areas. The PASS model is observationally driven, making use of extensive parameterizations of surface properties and processes. Heterogeneities in surface conditions are spatially resolved to an extent determined primarily by the satellite data pixel size. The purpose of modeling the spatial and interannual variability of water budget components at the regional scale is to evaluate the PASS model's ability to bridge a large grid cell of a climate model with its subgrid-scale variation. Modeled results indicate that annual total evapotranspiration at the WRW is about 66%?88% of annual precipitation?reasonable values for southeastern Kansas?and that it varies spatially and temporally. Seasonal distribution of precipitation plays an important role in evapotranspiration estimates. Comparison of modeled runoff with stream gauge measurements demonstrated close agreement and verified the accuracy of modeled evapotranspiration at the regional scale. In situ measurements of energy fluxes compare favorably with the modeled values for corresponding grid cells, and measured surface soil moisture corresponds with modeled root-zone available moisture in terms of temporal variability despite very heterogeneous surface conditions. With its ability to couple remote sensing data with surface meteorology data and its computational efficiency, PASS is easily used for modeling surface hydrological components over an extended region and in real time. Thus, it can fill a gap in evaluations of climate model output using limited field observations.