- Autor: Luis Ricardo Lage Rodrigues
- Lloc: Aula 507 "Pere Pasqual", Facultat de Física.
- Data/Hora: 29 gener 2016/11:30
- President: Dr. Jon Sáenz Aguirre
- Secretari: Dra. María Gonçalves Ageitos
- Vocal: Dra. Eisa Mohino Harris
- Substituts: Dr. Antonio Santiago Cofiño González, Dr. Oriol Jorba Casellas
Dr. Francisco Javier Doblas Reyes
Dr. Caio Augusto dos Santos Coelho
Dr. Ileana Blade Mendoza (Tutora)
Ctment technology allows the proliferation of llmltiple forecast systems developed by different research institutions from all over the world. However, most decision makers need a reliable probabilistic prediction instead of a set of predictions to take an action given a probability of an event. Severa! studies have shown that the combination of predictions derived from severa! forecast systems yields on average to better redictions than the best single forecast system. Nevertheless, none of these studies has shown the existence of a combinatíon method that produces the best predictions. Therefore, this thesis aims at applying different statistical techniques to cómbine predictions derived from different statistical and dynamical forecast systems. Some of these techniques attempt to combine the predictions assigning tmequal weights to the different forecast systems based on their past performance and one of them combines ali forecast Systems without assigning weights, considering ali of them having the same leve! of forecast skill. This Iater are referred to as simple multimodel (SMM). A unique feature of this study is !lle broad nature of the forecast quality assessment, performed using múltiple deterministic and. probabilistic verification measures and the same verifying observations. Thus, allowing to compare the predictions produced by the diferent combination methods and forecast systems in a uníform fushion. Besides, most of the forecast systems used in this study are publicly available on the intemet or could be easily implemented by the user. This thesis focuses on seasonal prediction of sea surface temperature (SST), near-surface temperature and precipitation in tropical and extratropical regions. lt is shown that the predictions of the SMM are often better than those combination methods that assign unequal weights. The difficulty in the robust estimation of the weights dueto the small samples available is one of the reasons that Iimit the potential benefit of the combination methodq that assign unequal weights. However, some of the results shown here give light to further research 011 how to improve the SMM predictions using combination methods that assign unequal weights. For in.qtance, the combination methods that assígn unequal weights improve the SMM predictions when only a fraction of ali single forecast systems have skill as shown for !he predictions of SST. On the other hand, it is shown that there are cases when combining many forecast. systems <loes not lead to improved forecasts when compared to the best single forecast system. This suggests that a multimodel approach fa not necessarily better than an especially skillful forecast system, which highlights the importance of continuously assessing the forecast quality for the specific applicatio11 of the users.
Key words: Climate prediction, forecast verification, uncertainty quantification