This study presents an assessment of seasonal forecasts from the regional climate model RegCM3 nested to the CPTEC/COLA global model. The used RegCM3 version is 60 km horizontal resolution over an area that includes large part of South America. RegCM3 and CPTEC/COLA forecasts were evaluated using the rainfall and air temperature analyses from Climate Prediction Center (CPC) and National Centers for Environmental Prediction (NCEP), respectively. From May 2005 to July 2007, 27 seasonal forecast of rainfall and air temperature (except for air temperature from CPTEC/COLA with 26 forecasts) were evaluated over three regions of Brazil: Northeast (NDE), Southeast (SDE) and South (SUL). RegCM3 forecasts were also compared with the climatologic fields of the analyses. The statistical indexes (bias, correlation coefficient, root mean square error and efficiency coefficient) showed that over the three regions (NDE, SDE and SUL) the RegCM3 predicted seasonal rainfall is closer to the observations than the CPTEC/COLA. Moreover, the RegCM3 is a best predictor of the seasonal rainfall than the average of the observations during the period. For air temperature, the RegCM3 forecasts perform better than the CPTEC/COLA over NDE and SUL regions, while the CPTEC/COLA shows better performance ovver SDE. Finally, the RegCM3 seasonal air temperature and rainfall estimated are closer to the observations than the observed climatology. These results suggest a potential utility of RegCM3 to be used for seasonal forecast which will be explored in the future using ensemble forecast.
seasonal forecast; regional climate model; objective evaluation
