资源环境科技发展态势分析平台

This study investigates the feasibility and presents an assessment of probabilistic rainy season onset forecasts for Sao Paulo, Brazil, located in a region with a well-defined wet season from mid-austral spring (October) to austral autumn (March/April). The probabilistic forecasts were produced with (1) a simple empirical Cox-regression model using July Nino-3 index as predictor, (2) the dynamical coupled atmosphere-land-surface-ocean-sea-ice model used in the UK Met Office Global Seasonal Forecast System (GloSea5) and (3) a procedure that combines the empirical and dynamical model onset probabilistic forecasts. The probabilistic forecast assessment was performed over the 1996-2009 retrospective forecast period for the event rainy season onset earlier (or later) than the historical (mean) onset date. The three investigated approaches resulted in similar discrimination ability of around 80%, which represents the probability of the probabilistic forecasts correctly distinguishing earlier from a later than mean onsets, suggesting good potential for rainy season onset forecasts for Sao Paulo. The robustness of this assessment for an extended period (longer than 1996-2009) and for a region (20 degrees S, 25 degrees S, 45 degrees W, 55 degrees W) that includes the city of Sao Paulo was checked, reinforcing the validity of the obtained results at both local and regional scales.