A Multiplicative-Exponential function to correct precipitation for distributed hydrological modeling in Poorly-gauged basins

Abstract

Given the greater spatial and temporal variability in precipitation and the absence of rain gauges at mountainous regions, it is challenging to accurately estimate precipitation, consequently impeding the predictive capability of hydrological models. Here, we developed a nonlinear precipitation correction method in the SWAT (Soil and Water Assessment Tool) model to correct gridded (remote sensing or reanalysis) precipitation data for better distributed hydrological modeling. The precipitation correction module is featured by a multiplicative-exponential equation with two parameters, a multiplicative coefficient (PCP_COE) and an exponent (PCP_EXP). We proposed three configurations to determine PCP_COE and PCP_EXP, i.e., curve fitting between gridded and gauge precipitation data pairs (Configuration C1) or using gridded precipitation from multi-grids within an elevation range of the gauge (Configuration C2), and the calibration of the SWAT model against observed streamflow (Configuration C3). We applied the proposed method to a typical poorly-gauged alpine inland basin, the Xiangride River Basin of Northwest China, with only one rain gauge and streamflow station. Results show that Configuration C1 outperformed C2 and C3 by significantly improving streamflow modeling in both model calibration and validation, particularly when the GPM (Global Precipitation Measurement) precipitation datasets were used. We also argue that Configuration C3 is promising in ungauged basins with only streamflow available, since this configuration does not require gauge precipitation. This study puts forward a feasible method to derive more accurate spatiotemporal precipitation estimate based on gridded precipitation products, hence improving distributed hydrological modeling in poorly-gauged and ungauged basins.