Applicability of ERA5 Reanalysis Precipitation Data in Runoff Modeling in China’s Ili River Basin

Abstract

The widespread utilization of reanalysis-based precipitation (RP) data has significantly facilitated hydrometeorological studies in areas where observations are scarce. However, assessing the applicability of the RP data before being applied in any basin is necessary considering that the inherited errors of the RP data vary with surface conditions, seasonal cycles, and different climatic zones. In this study, the Ili River Basin (IRB) was selected as the study area to evaluate the accuracy and impact of ERA5 reanalysis precipitation data, a type of RP data, on the hydrological cycling in the IRB. Four methods, linear scaling (LS), local intensity scaling (LOCI), power transformation (PT), and distribution mapping (DM), were tested to correct the bias of the ERA5 reanalysis precipitation. We found the following results. (1) The ERA5 precipitation exhibits a significant overestimation of gauge-observed precipitation within the IRB, indicating poor accuracy. Compared to the gauge-observed precipitation, the comprehensive performance of ERA5 precipitation in the dry season is better than that in the entire year, while the one in the entire year is better than that in the rainy season. The main error of the ERA5 precipitation lies in its insufficient ability to distinguish between no rain (0–0.1 mm) and light rain (0.1–4 mm) events, (2) All bias correction methods effectively address the biases in the ERA5 precipitation, although the extent of correction varies. The LS and DM methods outperformed the others regarding the time-series-based indices, while the DM method outperformed the others regarding the exceedance probability, and (3) In evaluating runoff simulation accuracy indices, hydrologic simulations driven by the corrected ERA5 precipitation using the DM method markedly outperform those driven by the ERA5 precipitation alone and even those corrected by the PT and LS methods. Specifically, the DM method exhibits the best correction of the flow duration curve and peak flow, with Nash-Sutcliffe efficiency coefficient (NSE) at 0.83 and relative deviation (RD) at 9.23% in the calibration period, and NSE at 0.85 and RD at 1.29% in the verification period. Corrected ERA5 precipitation data can fill in the gaps left by meteorological stations, effectively addressing the issue of inaccurate peak values in runoff simulations caused by using uncorrected ERA5 precipitation data. Due to the constraints of complex terrain and natural conditions, the hydrometeorological monitoring stations are sparse and unevenly distributed in western China’s Ili River Basin (IRB). The precipitation data obtained from the monitoring sites can hardly meet the needs of hydrological simulations. Furthermore, it challenges us to manage the IRB better when facing extreme weather conditions (e.g., drought and flooding). The ERA5 reanalysis precipitation data with comprehensive coverage and high resolution is a promising alternative to observing precipitation at the basin level. However, verifying the feasibility of using ERA5 precipitation data in IRB is necessary. In this study, we first noticed that the ERA5 precipitation data has a significant deviation and low accuracy in IRB by comparing it with available observed precipitation. Thus, we corrected ERA5 reanalysis precipitation using four different methods: linear scaling (LS), local intensity scaling (LOCI), power transformation (PT), and distribution mapping (DM). We found that the DM method can better capture extreme precipitation events, and the corresponding simulated runoff has the highest agreement with the observed runoff.