Role of high-resolution modeling system in prediction of heavy rainfall events over Tamil Nadu and Kerala on different global/regional datasets

The present study explored the numerical studies on forecast capabilities of the Advanced Research version of WRF (ARW) modeling system in the simulation of two heavy rainfall events (HREs) over Tamil Nadu in December 2015 (named as Chennai flood 2015) and Kerala in August 2019 (named as Kerala 2019 flood). For this, a total of 24 numerical simulations are conducted on four different data sets namely: National Centre for Environmental Prediction (NCEP) Final Analysis (FNL), NCEP GFS, Indian Monsoon Data Assimilation and Analysis (IMDAA) and European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) at three different model horizontal resolutions (4 km, 1.33 km and 0.8 km) on two different HREs. The model results from the finer domain are used for validation purposes and compared with available observations from the India Meteorological Department (IMD), the Tropical Rainfall Measuring Mission (TRMM), Automatic Weather Station (AWS), and Doppler Weather Radar (DWR). The Kerala 2019 case is well simulated with the model in most of the datasets, but with higher magnitude of rainfall in IMDAA compared to the observation. The simulated 24 h of accumulated heavy rainfall is well captured with NCEP FNL followed by NCEP GFS using 0.8 km horizontal resolution with a higher probability of detection, good correlation, and less error as compared to the other experiments. The results also focused on the occurrence of cloudbursts (100 mm rain within one hour) during Chennai's heavy rainfall (Ray and Kannan Mausam 73:587-596. https://doi.org/10.54302/mausam.v73i3.214, 2022), and the results suggested that model is able to capture the cloudburst. Forecast of Chennai HRE with ERA5 and IMDAA failed to capture accumulated rainfall as well as cloudburst. Results also focused on the Hovmoller diagram, maximum reflectivity, and vertical velocity. Results concluded that model simulations from three different horizontal resolutions are sensitive on the different datasets and also need to improve the forecast of heavy rainfall events.