Stochastic downscaling of LAM predictions: an example in the Mediterranean area

Abstract. In the absence of a full deterministic modelling of small-scale rainfall, it is common practice to resort to the use of stochastic downscaling models to generate ensemble rainfall predictions to be used as inputs to rainfall-runoff models. Here we present an application of a novel spatial-temporal downscaling procedure based on a non-linear transformation of a linearly correlated (gaussian) field. This procedure allows for reproducing the scaling properties (if any) of the rainfall pattern and it can be easily linked with meteorological forecasts produced by limited area meteorological models.