Postprocessing Suspiciously High Localised Precipitation in <scp>ERA5</scp> for Improved Hydrological Simulations

ABSTRACT Hydrological simulations critically depend on the quality of precipitation inputs, particularly, for representing extreme events. The ERA5 reanalysis dataset is a widely used product and a key input to the Global Flood Awareness System (GloFAS) of the Copernicus Emergency Management Service (CEMS). Despite its general reliability, ERA5 exhibits a known issue: isolated grid cells with unrealistically high precipitation values, often termed ‘rainbombs’. These anomalies can inflate river discharge estimates in both hindcasts and forecasts, increasing the risk of false flood alerts. In this study, we present a simple yet effective method to reduce the magnitudes of such high‐intensity localised events by leveraging data from neighbouring ERA5 grid points and statistical benchmarks from unaffected datasets. The postprocessing was applied to daily data spanning from 1979 until 2024. Over 7% of ERA5 grid points exhibited at least one rainbomb, with 722 locations, primarily in tropical regions and complex terrain such as the Andes, East African highlands and Papua Island, experiencing more than 100 events. The hydrological impact of these anomalies was evaluated for two basins with frequent occurrences: the Tana basin in Kenya and the Guayas basin in Ecuador. Results show that the postprocessed dataset not only reduces unrealistic discharge peaks from individual events but also significantly alters return period thresholds, yielding more realistic hydrological indicators. While the approach effectively targets spatial inconsistencies and it is flexible enough to be applied to different datasets, some uncertainty remains regarding the treatment of highly localised extreme precipitation events. This modified precipitation dataset will be integrated into the next major release of GloFAS (version 5), scheduled for operational deployment in 2026.