How sensitive are probabilistic precipitation forecasts to the choice of calibration algorithms and the ensemble generation method? Part I: Sensitivity to calibration methods

Abstract

Different techniques for obtaining probabilistic quantitative precipitation forecasts (PQPFs) over South America are tested during the 2002-2003 warm season. They have been applied to a regional ensemble system which uses the breeding technique to generate initial and boundary conditions perturbations. This comparison involves seven algorithms and also includes experiments to select an adequate size for the training period. Results show that the sensitivity to different calibration strategies is small with the exception of the rank histogram algorithm. The inclusion of the ensemble spread or the use of different ensemble members for the computation of probabilities shows almost no improvement with respect to probabilistic forecasts computed using the ensemble mean. This is basically due to the strong relationship between precipitation error and its amount. © 2011 Royal Meteorological Society.