Estudo da sensibilidade do modelo WRF às parametrizações de microfísica de nuvens e à assimilação de dados observados.

Abstract

The quantitative prediction of precipitation and the positioning of the rain clouds is one of the main challenges of numerical modeling of the atmosphere in present days. This work aims to evaluate the performance of the microphysical parameterizations in regional modeling, with emphasis on the role of large- scale information and its influence on the simulations, and the use of observational data from radiosondes as a way to add information to modeling. Initially, two reanalysis (NCEP2 and ERAI) were statistically compared with data from PCDs from the Ceará State. It was found that the ERAI showed similarity to the observations, especially for variables directly linked to convection. Then, the ERAI is used as large scale forcing in simulations with the WRF model. It was observed that the use of detailed microphysics does not necessarily improve the model performance, if in situ data were not used. Finally, two high resolution simulations were performed. The first f orced by reanalysis without modification and other forced by reanalysis using the modified method of objective analysis of the WRF, to include the time series of radiosonde observations collected during the experimental campaign of the CHUVA Project in Fortaleza- CE. The two simulations were compared with data observed by the radiometer to the same place and period of the radiosonde. It was observed that the inclusion of radiosonde observations in to the model leads to a better simulation of a convective system that occurred in April 2011, mostly for the variables related to convection. Using comparative statistical analysis, t his work points that the use of a higher density of valid observational data in the model can improve much more efficiently the model results than the use of a dynamic downscal ing of large- scale data or the use of schemes with detailed microphysics, which in some circumstances may even introduce more errors into the modeled system s.