Assessment of the WRF Model for Real-Time Prediction of Heavy Rainfall Events over the Twin Cities of East Coast of India

Abstract

The East coast of India, including Bhubaneswar and Cuttack in Odisha, often faces heavy rainfall events (HREs), leading to floods and significant loss of life and property. The present study evaluates the performance of a previously customized WRF model, forced by NCEP-GFS, for its capabilities in HRE forecasting and compares it with the India Meteorological Department's Global Forecast System (IMD-GFS) model in predicting HREs in quasi-operational mode. Their performance is assessed against observed daily rainfall station data, considering 23 HREs that occurred during the 2022 monsoon season. Our findings indicate that the optimum WRF configuration successfully captures both the occurrence of HREs and their magnitudes. Results show that the optimized WRF model effectively captures both the occurrence and intensity of HREs, achieving an overall success rate of 64% compared to 16% for the IMD-GFS at the station level. Concerning various lead times, the WRF (IMD-GFS) exhibited success rates of 45% (8%), 40% (8%), and 46% (4%) for day-1, day-2, and day-3 lead times, respectively. Regarding rainfall magnitude, the WRF model showed a 30% overestimation, while the IMD-GFS delineated a 65% underestimation. Furthermore, the optimized WRF model effectively predicts widespread HREs influenced by large-scale factors. The differences in results between the WRF and IMD-GFS models can mostly be attributed to variations in resolution and model configuration. However, the present study emphasizes the need for dynamically downscaling using high-resolution mesoscale models to accurately predict city-scale HREs in urban regions for its usefulness by stakeholders.