Improvement in ensemble-based localized rainfall forecast skills over the Indian northwestern Himalaya states using initial states from a high-resolution land data assimilation system

Abstract

Heavy rainfall events (HREs) during the summer monsoon (2023) have caused severe flooding and significant socio-economic losses, particularly in the mountainous regions of Himachal Pradesh and Uttarakhand. During August 12–16, 2023, these states experienced one of the most historic HREs. Accurately forecasting such events over complex terrain remains challenging due to limitations in land surface initialization and land use and land cover (LULC) representation in numerical weather prediction (NWP) models. In the present study, the Weather Research and Forecasting (WRF) model and the High-Resolution Land Data Assimilation System (HRLDAS) are used to conduct numerical simulations with lead times of up to four days. The study involved 72 simulations using six multi-physics parameterization combinations, evolved land-state data from HRLDAS, and LULC datasets from the Indian Space Research Organisation (ISRO) and the United States Geological Survey (USGS). These simulations were organized into eight control (CNTL) and eight HRLDAS ensembles to evaluate HRE forecast accuracy. Key findings highlight that the ISRO HRLDAS ensembles significantly improved HRE forecast skill (absolute percentage error ~ 32%) compared to CNTL ensembles (> 40%) over Himachal Pradesh, particularly on peak HRE day 2. The improved HRLDAS soil moisture enhanced rainfall performance in low-elevation districts of both states, which also coincides with a high rainfall vulnerability index. Among the physics configurations, microphysics (CMP)-based ensembles produced the most realistic rainfall distributions, followed by cumulus (CPS) and planetary boundary layer (PBL) ensembles. Overall, ISRO HRLDAS ensembles outperformed CNTL ensembles in accurately simulating rainfall location, timing, and intensity on day 2 and day 3. Additionally, ISRO HRLDAS demonstrated improved representation of liquid and frozen hydrometeors critical to rain-formative processes, as demonstrated by the microphysics (LE1) and cumulus (LE2) parameterized ensembles. However, all ensembles underestimated rainfall over high-altitude districts, indicating unresolved sub-grid-scale processes. This study emphasizes the importance of integrating evolved HRLDAS land states and updated LULC in high-resolution numerical models for improved HRE forecasting and disaster mitigation over complex mountainous terrain.