| Popis: |
Although most operational seasonal forecasting systems are based on dynamicalmodels, empirical forecasting systems, built on statistical relationships betweenpresent and future at seasonal time horizons conditions of the climate system, providea feasible and realistic alternative and a source of supplementary information. Here, anew empirical model based on partial least squares regression is presented. Originallydesigned as a flexible tool, the system is able to automatically select predictors from an initial pool and explore spatial fields looking for additional predictors.the model can be run with many configurations including different predictands,resolutions, leads and aggregation times. The model benefits from specific predictorsfor the Mediterranean region unveiled in the frame of the MEDSCOPE project. Wepresent here 2 sets of results: the first one from a configuration producing probabilistic forecasts of seasonal(3 month averages) temperature and precipitation over the Mediterranean area, their verification and comparisonagainst a selection of state-of-the-art seasonal forecast systems based on dynamicalmodels in a hindcast period (1994-2015). The model is able to produce spatiallycoherent anomaly patterns, and reach levels of skill comparable to those based ondynamical models. To explore the potential of the model for producing skilful forecasts over reduced areas, a second set of results are calculated using higher resolution predictands over Iberia, again comparing its skill with that of a set of state of the art models. Examples of the model usage for evaluating the impact on skill ofcertain predictor helping in the search and understanding of new sources ofpredictability are also shown. |
