Statistical comparison and hydrological utility evaluation of ERA5-Land and IMERG precipitation products on the Tibetan Plateau

Abstract

In the context of global warming, precipitation information over the Tibetan Plateau (TP), known as the water tower of Asia, is crucial for both atmospheric and hydrological cycle research. As important alternatives of gauge observations, the satellite-based precipitation estimations and atmospheric reanalysis precipitation prodcuts have undergone rapid development over the past few decades. However, their performance and usabilities are not effectively investigated over the TP, particularly for the latestly released products. Hence, this study presents a comprehensive evaluation of a state-of-the-art reanalysis precipitation (ERA5-Land) and compares it with two IMERG products (i.e., IMERG-E and IMERG-L) from 2001 to 2018, with a special focus on their hydrological utility. Results indicate that ERA5-Land presents a severe overestimation of precipitation amount and precipitation frequency but well reflects their spatial–temporal patterns over the entire TP. The IMERG products, particularly for IMERG-F, outperform ERA5-Land in estimating precipitation amount. In terms of hydrological utility in the upper Yellow River basin, both ERA5-Land and IMERG-E show poor simulation performance using gauge-based model parameters, which is closely related to their overestimation/underestimation of precipitation amount. In contrast, IMERG-F forced simulation performs much better due to its low systematic error. When the hydrological model is recalibrated based on individual precipitation input, the simulation performance of ERA5-Land is comparable to IMERG-F, slightly inferior to gauge observation. Therefore, it is recommended that the selection of appropriate precipitation products need to accord their own research purposes. IMERG-F provides a more accurate estimation of precipitation amount, while, employing input-specific calibrated model parameters, ERA5-Land can provide long term hydrological simulations.