Previsión de precipitación a muy corto plazo combinando técnicas de extrapolación de datos radar y modelos meteorológicos

Rainfall nowcasting based on the extrapolation of most recent radar rainfall fields usually shows better skill than NWP models for short lead times, since it benefits from observations of the initial rainfall state. In contraposition, NWP forecasts are based on modeling the atmospheric state and, thus, reproduce (to some extent) the evolution of the rainfall field, which makes them more skilful for longer lead times. Actual investigations suggest that the combination of extrapolation-based nowcasts with NWP outputs usually results in more skillful very short-term forecasts. In this thesis, an extrapolation technique of radar rainfall fields and an intensity correction to the NWP forecasts to match probability function of radar rainfall fields have been implemented. In addition, different configurations have been developed to combine linearly the rainfall forecasts by both methodologies, using different definitions of weights to blend them, to generate forecasts in the range of 0 to 6 hours. The different configurations have been evaluated in two domains: one using the Spanish radar mosaic observations and the Nonhydrostatic Mesoscale Model (NMM) forecasts for Spain and the European radar mosaic and COSMO-DE forecasts for central Europe. The different configurations show different results depending on the meteorological situation, so what, the dynamics of the meteorological processes of each situation and the capacity of the model to reproduce it dearly affect the results. Additionally, the impact of the diurnal cycle of precipitation on the extrapolation results and the meteorological models has been analyzed and two new configurations have been implemented to incorporate, in some way, the impact of the diurnal cycle of precipitation in the warm season. The results of these two configurations depend to a great extent on the quality of the NWP forecasts and when analyzing a large number of cases, their results are similar to those of simpler configurations. Finally, it has been found that the combination of predictions of the radar extrapolation technique and the meteorological models provides better results than each of the two separately, especially between the third and the fifth hour of forecasting, allowing to have a single forecast that combines the available information, with a temporal coherence in the range of zero to six hours. In the operational mode, a configuration has been implemented that uses fixed weights due to its simplicity and because his forecast quality was slightly superior to the one of the linear configurations and equal to those of more complex configurations.