High accuracy of precipitation reanalyses resulted in good river discharge simulations in a semi-arid basin

Abstract

Hydrological modeling requires accurate climate data with high spatial resolution, what is often unavailable in a large fraction of the world. As an alternative to scarce meteorological observations, global reanalysis emerge as potential substitutes. Herewith, we compared CRU, NCEP CFSR, APHRODITE, and Asfezari reanalyses precipitation data with observed station data and SWAT applications using each of these reanalises to simulate the hydrology of the Maharlu Lake basin in Iran. In addition, we assessed the accuracy of these reanalysis in representing the local precipitation. Further, we employed observed and reanalysis precipitation in SWAT to simulate river discharge and evaluated the performance of the models. Monthly observed precipitation was best represented by CRU and NCEP CFSR (Nash-Sutcliffe Efficiency of 0.91), followed by APHRODITE (0.86) and Asfezari (0.80). In general, local precipitation was well represented by reanalysis products. According statistical metrics of the interpolation of CRU and APHRODITE were good and NCEP CFSR and Asfezari were satisfactory (Moriasi et al., 2007). Our assessment supported the use of reanalysis precipitation as a valid alternative to observations for hydrologic modelling in all regions especially in semi-arid basins.