Point–process statistical analysis for the ECMWF Ensemble Prediction System

The possibility of applying mathematical tools of point–process statistics to the ECMWF Ensemble Prediction System (EPS) is exploited in this work, in order to provide a different way to reduce ensemble information. The first two empirical orthogonal functions enable to represent 5–day ensemble forecasts as point processes in a plane. These planar representations are hence compared to a sample of Gaussian random point patterns, obtained by a Montecarlo method. The estimations of the nearest–neighbour distribution function and of the reduced second order momentum function for point processes relative to the ensemble predictions are in good agreement with the corresponding estimations of Gaussian random point processes.