Verificação da previsão do tempo em São Paulo com o modelo operacional WRF / Review of weather in São Paulo with the WRF Operational Model.

AUTOR(ES)

Fabiani Denise Bender

FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

01/11/2012

RESUMO

Forecasts of daily maximum and minimum temperatures and rainfall performed by the operational numerical weather prediction WRF (Weather Research Forecasting) model in the São Paulo are evaluated. Initial and boundary conditions provided by the 00UTC Global Forecast System (GFS) Model and WRF run for 72 hours, with two nested grids (with horizontal grid spacing of 50 km, D1, and 16.6 km, D2). The study was made for April 2010 to March 2011 period. Daily maximum and minimum temperatures comparisons were made, between predicted and observed data of the surface weather stations of Registro, São Paulo, Paranapanema, Campinas, Presidente Prudente and Votuporanga (CIIAGRO Data), through the mean error (ME) and root mean square error(RMSE), for the 36, 60 and 72 hours forecasts. The daily accumulated rainfall is evaluated using MODE with respect to the MERGE product, for the 36 hours forecast, with threshold of 0.3 mm over the spatial domain covering the State of São Paulo and neighborhoods. First, an analysis was made comparing grid pairs of predicted and observed fields, through the traditional statistical verification indexes: accuracy (PA), critical success index (ICS), bias (VIES), probability of detection (PD) and false alarm ratio (RFA). Subsequently, we analyzed the precipitation field with respect to area ratio (AR), distance from the centroids (DC), ratio of the 50th percentile (RP50) and ratio of the 90th percentile (RP90). The WRF, with D2 nested grid, had better performance compared to the grid of lower space resolution (higher horizontal grid spacing, D1) for both, daily temperatures (maximum and minimum) and the accumulated rainfall forecasts. The temperature forecast presented a damped pattern, with underestimated maximum and overestimated minimum values. Rainfall was overall overestimated spatially and in intensity for the three models throughout the analized period. The forecasted 50th percentile is generally higher than that observed, however, the 90th percentile is closer to observations. The results also indicate that the bias of the models varies annually. The best performances for both rainfall and temperature were obtained for the summer season, with the D2 showing slightly better results. However, the models had the biggest errors during the winter and autumn seasons. These errors were due to underestimation of maximum temperatures and overestimation in area and intensity of precipitation.

ASSUNTO(S)

estado de são paulo. modelo wrf operational numerical weather forecasting previsão numérica do tempo operacional state of são paulo. verificação verification wrf model

Documentos Relacionados

• Previsão de descargas atmosféricas usando o modelo de mesoescala WRF
• Prognóstico das variáveis meteorológicas e da evapotranspiração de referência com o modelo de previsão do tempo GFS/NCEP
• Previsão do vento utilizando o modelo atmosférico WRF para o estado de Alagoas
• The impact of eta model nesting on the weather forecasts.
• O impacto do aninhamento no modelo Eta nas previsões de tempo.