Urban effects on the Extreme Precipitation: Advanced Statistical Analysis and Numerical Weather Prediction Model at Convective Scales

Chennai is the fourth largest metropolitan city in India and being a coastal city, this region receives extreme precipitation events, especially during the North-East monsoon season. Rapid urbanization has led to profound changes in the city's land-use land cover, which has the potential to impact the local microclimate. The 2015 December Chennai flood resulted in the loss of lives and an economy of 2.5 billion $. Hence a reliable forecast system for the city helps for better preparedness. It is well acknowledged that urbanization affects rainfall distribution, but several limitations exist in past studies. The discrepancies between climatological and numerical investigations have not been addressed. To understand the impact of urbanization on the rainfall space-time distribution, it is also essential to choose rainfall events that originated from distinct synoptic conditions. Hence in this study, the Weather Research Forecast (WRF) model and advanced statistical analysis are used to examine the effects of urbanization on rainfall modification over Chennai city. The present investigation considers satellite estimations and observed station data due to the absence of a dense rain-gauge network. From ECMWF reanalysis ERA5 data, large-scale weather predictors are selected to create weather patterns using a fuzzy clustering method, with Chennai as the domain center. Subsequently dividing the observational rainfall data into two equal periods, the changes in the rainfall quantiles (80th to 99.9th) are obtained for each cluster.  Since large-scale circulation patterns are similar, these shifts show the likely impact of urbanization on rainfall.  Extreme value theory combined with regional frequency analysis is implemented to understand the changing rainfall statistics in the past 50 years. Numerous WRF simulations are performed at convection-permitting scales with varied domain and physics configurations for three extreme precipitation events from different synoptic conditions. The study also investigates the ability of scale-aware convective schemes to describe precipitation processes in high-resolution simulations for the study domain. Then, the WRF simulations are conducted with optimal physics combinations with default USGS land cover data (1991-1992) and high-resolution land-use data (2017) with Local Climatic Zone (LCZ) classifications, that represents the present urbanization. The simulations with recent land cover data coupling with Building Energy Parameterization (BEP) in the WRF model significantly improved the rainfall prediction skill in the spatial-temporal domain minimizing the bias. Interestingly, an intense convective rainfall event which was neither detected by the regional meteorological department nor WRF simulations with the default LULC map has been forecasted by representing the current urbanization scenario in the model. Further, this study also explains the dynamic and thermodynamic responses influencing rainfall distribution due to urbanization. The study advances the role of urbanization in a coastal city and provides pathways for better urban planning and design. The improved ensemble forecasts help for better preparedness during extreme rainfall events.