Assessing Multi‐Source Precipitation Estimates in Nepal: A Benchmark for Sub‐Seasonal Model Assessment

Abstract Sub‐seasonal to seasonal (S2S) prediction provides an extended range of lead time for decision‐makers across multiple sectors. S2S forecast is crucial for a country that are susceptible to hydro‐meteorological disasters like Nepal. However, S2S forecast requires assessment with reliable datasets before its application. Since, Nepal lacks a dense station network, multi‐source precipitation estimates (MPEs) are the obvious alternatives. Therefore, using classical evaluation metrics and extreme precipitation indices, this study assessed eleven high‐resolution datasets against 159‐gauge stations over Nepal for the 2001–2020 period. These datasets are classified as gauge‐interpolated (APHRODITE), merged (TPMFD, MSWEP, CLDAS, and CHIRPS), satellite‐based (IMERGV7, IMERGV6, CMORPH, and PERSIANN), and reanalysis (ERA5‐L and HAR). Satellite datasets (except IMERGV7) failed to capture the spatial pattern of mean annual precipitation, while others broadly captured it. Most MPEs struggled to accurately estimate wet season precipitation compared to dry season. Furthermore, increasing estimation error from light to extreme precipitation and decreasing skill metrics from flat to complex terrain, demonstrate the intensity and terrain‐specific limitations of MPEs. In terms of precipitation extremes, APHRODITE exhibits the highest skill, followed by MSWEP, TPMFD, HAR, ERA5‐L, IMERGV7, CLDAS, IMERGV6, CHIRPS, CMORPH, and PERSIANN. IMERGV7 exhibits increased skill in detecting extreme precipitation and precipitation over orography against IMERGV6. APHRODITE and TPMFD datasets performed consistently well in all scales, including weekly anomaly, with an average anomaly correlation of 0.84 and 0.66 respectively. However, since APHRODITE is not widely available, TPMFD can serve as a benchmark dataset for evaluating high‐resolution S2S forecasts within the study region.