Abstract
We develop dynamical-statistical forecast models in order to predict seasonal rainfall in southern Uruguay during summer and spring. The statistical technique consists of linear regressions between dynamic variables and rainfall observations. The forecasts for September-October-November are initialized in August and the ones for December-January-February in November. The dynamic variables are ICTP-MGCAs outputs, forced with sea surface temperature, predicted by NCEP-CFSv2, as boundary conditions. The observational data is accumulated monthly, values are measured in 10 meteorological stations that are cross-correlated, using one year windows, with the dynamic variables in order to find the best predictor indexes. We conclude that the best predictor index is the meridional wind in the 200 hPa level averaged in an area that includes the Southeast of South America. Northern wind anomalies in this area are associated with positive rainfall anomalies in Southern Uruguay. We found that while forecasts for the south of the country are skillful only in spring, forecasts for the metropolitan area of Montevideo are skilfull in both seasons showing their best performance in summer.
Keywords:
southern Uruguay; seasonal forecast; rainfall