by V. Rakesh, G. N. Mohapatra & Ajay Bankar
Urban extreme rainfall events (UEREs) are one of the weather hazards where modelers continue to struggle for accurate advance prediction despite recent advances in forecasting techniques. Bangalore, one of the prominent IT hubs in India, witnessed many urban flash floods in the recent past; a number of such events happened on 14 and 15 of August 2017, where overnight heavy rains brought several parts of the city to a standstill. In this study, the skill of sub-kilometer resolution forecasts from Weather Research and Forecasting (WRF) model in simulating these UEREs is examined. This study also compared the model performance for forecasts initialized from different forecast cycles in a day. Our examination for two UERE cases showed that forecast initialized at 0600 UTC yielded best results in terms of forecasting spatial distribution and intensity of rainfall. Our analysis also indicated that increasing model resolution further in convection-permitting resolution (< 3 km) has only marginal impact on spatial distribution of rainfall while model skill in simulating rainfall intensity slightly improved with increasing resolution. As expected, simulation by not deploying a cumulus scheme at cloud-resolving resolution do not make a significant difference in spatial distribution; however, inclusion of cumulus scheme improved quantitative precipitation forecast skill. The inferences from this study are useful for modelers working on improving operational forecast skill of UEREs using mesoscale model.