Resumo:
The Paraiba do Sul River Basin (PSRB) is frequently affected by extreme rainfall events, especially by weather systems during its rainy season, leading to financial damage and losses of human lives. The Numerical Weather Prediction (NWP) is a tool capable of supporting the forecast of natural disasters. However, its limitations and capabilities require understanding and addressing. In those terms, this work aims to investigate the sensitivity of the Weather Research and Forecasting (WRF) model on forecasting an extreme rainfall event that affected the Basin in January 2016, causing financial and human losses. Eight high resolution (3-km of maximum horizontal grid spacing) numerical simulations were performed by combining four cloud microphysics (CMP) and two cumulus convection schemes (CC). The simulations were compared against observed rainfall data (weather gauges), satellite-based precipitation estimates, and weather radar data. Results showed that WRF was able to represent the large-scale processes (South Atlantic Convergence Zone), with secondary errors caused by the positioning and intensity of the horizontal moisture convergence fields. Statistically, the simulations captured the correlations between forecasted and observed rain at the gauges, with values above 0.9. However, WRF showed strong dispersion between the experiments (above 50%), with strong tendency of underestimation of accumulated precipitation (by up to 80% at some sites). The vertical fields of radar reflectivity showed satisfactory results, but WRF tended to overestimate cloud depths by up to 2 km. Also, the simulations showed the early suppression of convective activity after the first two days of integration, leading to low rainfall amounts. However, WRF was able to simulate convectively active environments. On the vertical distribution of hydrometeors, the role of CC was indifferent in comparison with the sensitivity from the hydrometeor classes to CMP choices. The sophistication of CMP schemes was crucial to the horizontal and vertical fields. The prognostic of hail in CMP Milbrandt 2-moment and the consideration of advanced warm-rain processes in WDM6 contributed to a more adherent representation of cloud processes below the melting layer, as well as ice particles higher up in the atmosphere. Such representations impact hydrometeors distributions as well as further processes regarding up-/down-drafts. Overall, complexities associated with orography and land-use at the Basin watershed suggest the application of higher horizontal resolution runs to overcome errors from interactions between convection and terrain. This work is a contribution to the development of operational weather forecasts to the PSRB to mitigate natural disasters associated with extreme precipitation.