Performance evaluation of WRF for extreme flood forecasts in a coastal urban environment

Abstract

Increasing incidences of flash floods highlight the need for a reliable flood forecasting system to minimize the losses of lives and property. The most formidable challenge of flood forecasting is the availability of high resolution and accurate precipitation forecast despite having a sophisticated 3-ways flood hydrodynamic model. Global rainfall forecasting products are of coarse resolution, which makes them less reliable for urban flood forecasting. Therefore, high-resolution regional weather forecasting models such as the Weather Research and Forecasting (WRF) model are used to generate fine-scale rainfall estimates. Precipitation forecasting from the WRF model is highly dependent on model configuration, especially cumulus (CU) parameterization and microphysics (MP) schemes. In the present study, three physics schemes that include two urban, four CU and three MP schemes of WRF model are investigated for extreme precipitation estimates. The six events comprised of two of the highest rainfall events of the years 2012, 2013 and 2014 have been selected for investigation over Mumbai, India. The events are simulated using initial and boundary conditions from the ERA-Interim Reanalysis dataset. The simulated rainfall events are evaluated against the observations from 28 automatic weather stations over Mumbai. The analysis suggests that building environment parameterization (BEP) scheme influences the spatial pattern of the rainfall along with the reduction in rainfall bias. Further, CU schemes affect the magnitude of the rainfall while MP schemes have a lesser impact than the former. WRF simulations with BEP urban scheme, Grell-Devenyi 3D CU, and Lin MP scheme performs best (out of selected combinations). Besides, the best performing scheme has been tested with initial and boundary conditions from the global forecasting system (GFS) for the same events; the results have shown improved rainfall estimates than the GFS forecasts.