Nowcasting probabilísitico basado en observaciones de lluvia con radar meteorológico
- Autores: Alejandro Buil Martinez
- Directores de la Tesis: Marc Berenguer Ferrer (dir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2017
- Idioma: español
- Tribunal Calificador de la Tesis: Nicola Rebora (presid.), Daniel Sempere Torres (secret.), Eduardo F. Cassiraga (voc.)
- Materias:
- Enlaces
- Tesis en acceso abierto en: TDX
- Resumen
The nowcasting of rainfall based on the extrapolation of the radar precipitation field is a common technique used in different operational and research centers. However, this deterministic forecasting technique is subject to different sources of uncertainty that should be taken into account. It is known with quite certainty that a storm will occur somewhere, but its exact devolepment in time and space is not known, therfore including a probabilistic approach in nowcasting allows us to characterize the different sources of uncertainty. In this case, instead of a single amount of rainfall, a probability value is predicted for each point in the domain to find a certain rainfall intensity.
In this sense, the main objectives of this thesis are: ¿ to compare and evaluate a set of nowcasting techniques based on radar precipitation observations, ¿ to improve SBMcast probabilistic nowcasting technique, ¿ to develop a new nowcasting algotithm that allows the incorporation of precipitation information of the NWP model to those obtained from the observations of radar precipitation.
The set of probabilistic forecasting techniques that are evaluated can be separated into two blocks: those based only on forecasting the distribution function of the precipitation field at each point of the domain and those that calculate the predicted distribution function at each point of the domain from a set of precipitation fields compatible with the observations.
To assess their skill in different meteorological situations, an evaluation and verification system is established that allows quantifying the degree of accuracy for different precipitation thresholds. Within this context, a new version of SBMcast is proposed that allows to use a space-time model for each of the spatial scales that form the precipitation field.
Another aspect that is studied in this thesis is the impact of estimates of the global mean (IMF) and coverage (WAR) of the precipitation field prediction ability in the two nowcasting techniques, Lagrangian persistence and new version of SBMcast. Also included is a new approach to improve forecasting of the IMF-WAR using NWP model.
Finally, a new probabilistic rainfall forecasting technique is proposed based on ensembles that allows to combine the information of each point of the domain provided by the NWP model with the probabilistic forecasts of the new version of SBMcast. The objective is to identify the regions where rainfall growth and decay is most likely to occur through NWP model information and to determine future forecasts to these locations of the observation domain.
