Abstract Numerical model simulations are sensitive to the choice of physical parameterizations, particularly those related to convection processes, especially over complex terrains like the western Himalayan region. This study evaluates the sensitivity of the Weather Research and Forecasting model to three convective parameterization schemes (BMJ, Betts–Miller–Janjić; KF, Kain–Fritsch; GF, Grell–Freitas) and their multi-physics ensemble mean (ENSM) through long-term (2001-2016) winter (December-March) simulations. Model outputs are validated against multi-source datasets (IMD, IMERG, ERA5, and IMDAA), with extreme precipitation events (EPEs) identified as those exceeding the 95th percentile threshold. Climatology analysis reveals systematic precipitation biases across all single schemes, significantly reduced in the ENSM. Improvements in skill scores observed for ENSM highlight the reliability of ensemble approach as an alternative to account for model errors. Underestimations of low-level clouds and hydrometeors improve in ENSM. BMJ and KF (GF) schemes exhibit overestimated (underestimated) magnitudes and location-wise discrepancies for baroclinic instability and moisture transport. These issues are mitigated in ENSM, resulting in better alignment with reference datasets. Additionally, ENSM and KF’s storm tracks align well with reanalyses, while individual schemes show a minor northward placement. For EPEs, ENSM highlights Jammu-Kashmir, Himachal, and Uttarakhand as zones of maximum heavy precipitation and frequency. ENSM realistically represents the western disturbance (WD) induced circulation characteristics, southward-displaced subtropical jet, shifts in WD-associated vorticity maxima, and potential vorticity intrusions during extremes. Consistent with reference datasets, ENSM demonstrates the crucial role of vertical advection and dynamical controls in moisture distribution during EPEs. Our analysis indicates that a suitable WRF cumulus configuration includes either the multi-physics ENSM or the KF scheme, while the GF scheme yields the poorest results. The findings underscore the importance of evaluating both precipitation characteristics and underlying physical mechanisms to identify sources of systematic biases in model simulations.
Read full abstract