Improving weather and climate predictions by Cross Pollination in Time

Abstract

Historically, weather and climate forecasting has always been important. But the changing climate with the increasing frequency of extreme weather events and larger societal impacts has made good forecasting skill of even more interest (IPCC 2013). Improving individual weather and climate models is a difficult task, but models have improved steadily over time as witnessed by objective skill scores. Recently it has been proposed to combine imperfect models dynamically in order to further improve predictions. In this thesis we explore a technique called Cross Pollination in Time (CPT, Smith 2001). In the CPT approach the models exchange states during the prediction. The number of possible predictions grows quickly with time and a strategy to retain only a small number of predictions, called pruning, needs to be developed. In the training phase a pruning strategy is proposed based on retaining those solutions that remain closest to the truth. From the training phase probabilities are derived that determine weights to be applied to the imperfect models in the forecast phase. The CPT technique is explored using low-order dynamical systems and applied to a global atmospheric model. The results indicate that the CPT approach improves the forecast quality over the individual models. The technique is suited for application to state-of-the art high-dimensional weather and climate models.