Self-Organizing Maps (SOM) can be used to analyze both atmospheric patterns and local observations. In this project, SOMs are used to analyze multi-model ensemble seasonal forecasts from the DEMETER project. SOM is a particular clustering technique, in which the centroids have been topologically ordered. Different events define a Probability Density Function (PDF) on the SOM, and is easy to built the climatological SOM of every particular phenomenon. Therefore, the seasonal ensemble forecasts from the multi-model DEMETER ensemble can be downscaled to local stations and projected on that SOMs, providing a probabilistic local forecast from the resulting PDFs. Moreover, a measure of predictability for the downscaled forecast can be computed in terms of the entropy of the multi-model PDFs.