PhaSt: stochastic phase-diffusion model for ensemble rainfall nowcasting

摘要

Hydrometeorological hazard management often requires the development of reliable statistical rainfall nowcasting systems. Ideally, such procedures should be capable of generating stochastic ensemble forecasts of precipitation intensities on scales of the order of a few kilometres, up to a few hours in advance. Ensemble rainfall nowcasting allows for characterizing the uncertainty associated with nowcasting procedures by providing a probabilistic forecast of the future evolution of an event. Here we discuss an ensemble rainfall nowcasting technique, named PhaSt (Phase Stochastic), based on the extrapolation of radar observations by a diffusive process in Fourier space. The procedure generates stochastic ensembles of precipitation intensity forecast fields where individual ensemble members can be considered as different possible realizations of the same precipitation event. The model is tested on a data set of rainfall events measured by the C-POL radar of Mt Settepani (Liguria, Italy) and its performance verified in terms of standard probabilistic scores.