Modeling Spatiotemporal Precipitation: Effects of Density, Interpolation, and Land Use Distribution

Abstract

Characterization of precipitation is critical in quantifying distributed catchment-wide discharge. The gauge network is a key driver in hydrologic modeling to characterize discharge. The accuracy of precipitation is dependent on the location of stations, the density of the network, and the interpolation scheme. Our study examines 16 weather stations in a 64 km2 catchment. We develop a weighted, distributed approach for gap-filling the observed meteorological dataset. We analyze five interpolation methods (Thiessen, IDW, nearest neighbor, spline, and ordinary Kriging) at five gauge densities. We utilize precipitation in a SWAT model to estimate discharge in lumped parameter simulations and in a distributed approach at the multiple densities (1, 16, 50, 142, and 300 stations). Gauge density has a substantial impact on distributed discharge and the optimal gauge density is between 50 and 142 stations. Our results also indicate that the IDW interpolation scheme was optimum, although the Kriging and Thiessen polygon methods produced similar results. To further examine variability in discharge, we characterized the land use and soil distribution throughout each of the subbasins. The optimal rain gauge position and distribution of the gauges drastically influence catchment-wide runoff. We found that it is best to locate the gauges near less permeable locations.