Assessment of IMD-HRRR Model in Predicting Extreme Rainfall Events over India

Climate vulnerability in India has intensified due to the increasing frequency of hydro-meteorological extreme rainfall events (EREs). While numerical weather prediction (NWP) models have improved for large-scale medium-range forecasting, precipitation forecasting, especially for extreme events, remains less reliable due to simplified representations of topography, land cover, and vegetation in global models. To address this, the India Meteorological Department (IMD) operates the High-Resolution Rapid Refresh (IMD-HRRR) model at 2 km horizontal resolution, assimilating hourly radar and surface observations to provide 12-hour nowcasts updated every 2 hours. This study evaluates the performance of the IMD-HRRR model in predicting extremely heavy rainfall events (≥ 204.4 mm/day) across mainland India in 2023. The IMD gauge-adjusted merged rainfall product GSMaP_ISRO (0.1° × 0.1°, hourly) was utilised to compare model forecasts using traditional metrics (bias, RMSE, correlation, and critical success index) and the Extreme Dependency Score (EDS), which is better suited for rare events. Results indicate that IMD-HRRR generally captures the spatial distribution of extreme rainfall over the Northwest, Northeast, and South Peninsular domains, with peak activity correctly predicted over Meghalaya, Assam, and the Western Ghats. However, systematic dry biases were observed across all domains (-0.61 to -0.93 mm/12 h), with higher root mean square deviation in the Northwest (~ 1.79 mm) and Northeast/South Peninsular regions (~ 3.54 mm). The model exhibits a slight dry bias over the Western Himalayas and Kutch, a wet bias over the Western Ghats, and an accurate representation of the Konkan–Karnataka coastal rainfall belt. EDS analysis confirms the model's ability to capture extreme events. The contiguous rain analysis (CRA)-based object verification shows efficient moisture convergence; however, it indicates shape & displacement errors for higher rainfall thresholds. Overall, IMD-HRRR provides valuable early warning guidance for extreme rainfall, though improvements to orographic and convective parameterisations are needed in mountainous regions.