Natural Hazard in Nova Friburgo (Brazil): Numerical sensitivity analysis using different parameterization combinations in the WRF model

Abstract

Abstract Forecasting rainfall is essential for warning issues and mitigating natural disasters. For this purpose, the employment of numerical weather models, even with their uncertainties, can generate reliable forecasts and guide decision-makers. The accuracy of a numerical model can be verified using statistical tools, and it is an essential procedure that needs to be made operationally, aiming to increase the forecasts' reliability. Numerical precipitation forecasts for the mountainous region of Rio de Janeiro, Brazil, were performed using the Weather Research Forecasting model, configured with three spatial resolution grids of 9 km, 3 km, and 1 km, and combining different parameterizations for five physical processes: cloud microphysics, cumulus, planetary boundary layer, surface layer, and land surface. The period of interest was January 11th-12th, 2011, when large rainfall accumulations originated the fatal natural hazards in Brazil. Analyses of the spatial distribution of rainfall and its temporal evolution were performed to evaluate the predictions from the quantitative and qualitative approaches. The results showed that the Kessler (cloud microphysics), MYNN3 (planetary boundary layer), Grell-Freitas, Betts-Miller-Janjic (cumulus) parameterizations, and the two highest resolution grids (at times, one was better than the other) had predicted the highest rainfall accumulations. From the initial results, this work reinforces the importance of forecast verification, especially considering different physical parameterizations and spatial resolutions since they can strongly influence the results. Also, corroborate the importance of local numerical forecasts studies aiming to identify the best numerical configurations to forecast heavy rainfall events to alert decision-makers to the possibility of a natural hazard.