An approach to seasonal forecasting of summer rainfall in Buenos Aires, Argentina

Abstract

This paper analyzes some summer rainfall characteristics in Buenos Aires and developes a seasonal prediction scheme. Buenos Aires is located along the coast of the Río de la Plata in Argentina. The outstanding rainfall feature is the presence of an annual cycle with maximum precipitation in summer. The analysis of the annual rainfall evolution since 1908 showed a positive trend of 2.1 mm/year and 1.8 mm/year for the period between December and February, representative of the summer season. The Observatorio Central Buenos Aires/station, located in the downtown, registered a mean annual accumulated rainfall of 1070 mm with a standard deviation of 239 mm, during the period 1908-2007. The mean accumulated precipitation during January, February and March was 305 mm with a standard deviation of 125 mm. The wet and dry periods were identified and the dry periods tended to be longer during 1908-1957 meanwhile wet periods resulted longer and more intense in 1958-2007. Accumulated rainfall between December and February was related to some mean meteorological variables between September and November, with the aim to develop a statistical prediction scheme. Careful selection of predictors, based largely on physical reasoning, was done and they were used in a regression model, following a forward stepwise methodology. The analysis shows that the most important source of predictability comes from the cyclonic activity in the Atlantic Ocean and the flow from Brazilian forest. The observed and forecast rainfall series were significantly correlated (0.59) and nearly the 35% of summer rainfall variance was predicted by the proposed method. A semi-quantitative validation was done by using terciles of the observed and forecast distributions. The skill of the forecast got a good result although there is still an important portion of the variance that cannot be explained by this model and therefore, the method might be improved in future research.