A0682
Title: Dynamic combination and calibration for climate predictions
Authors: Roberto Casarin - University Ca' Foscari of Venice (Italy) [presenting]
Dario Palumbo - University Ca Foscari of Venice (Italy)
Francesco Ravazzolo - BI Norwegian Business School (Norway)
Abstract: When multiple forecasts are available from different models or sources, it is possible to combine these to make use of all relevant information on the variable to be predicted and, as a consequence, to produce better forecasts. This is particularly important when working with uncertain environments, and the selection of relevant information a priori is not an easy task. Climate change and challenges related to global warming are essential issues for science; therefore, modeling their uncertainty and producing reliable forecasts to deal with them are crucial tasks for econometricians. Dynamic combination and calibration are introduced to produce accurate climate predictions. The density calibration literature is extended, and the application of dynamic combinations when calibrating models is proposed. The static model of the prior study is extended to a score-driven dynamic model for calibration and combination of predictive distributions. The time-varying weights are fitted by an observation-driven model with dynamics driven by the score of the assumed conditional likelihood of the data-generating process. The model is very flexible and can handle different shapes, instability, and model uncertainty. It is illustrated analytically and in simulation exercises. Then, the methodology is applied to climate predictions.
