2023年3月15日 (水) 16:00-18:00
東大本郷キャンパス + zoom
Noel Sebastian Keenlyside (ベルゲン大学, Bjerknes Centre for Climate Research)
Supermodelling as means to improve climate models and predictions
Climate models suffer from long-standing biases that introduce uncertainties in their prediction and projection of climate. The standard approach to address this issue is to perform simulation with different models, run independently. Combining outputs of such simulations leads to cancellation of errors but cannot mitigate common biases. In this respect, storm-resolving models offer great promise, but technical challenges exist in performing extensive simulations with such models. Supermodelling is an alternate approach to address these issues. In a supermodel, multiple models are combined interactively, so that errors can be compensated as the models runs. Large-scale model errors can be mitigated, through the compensation of model errors at an early stage before they are amplified by non-linear interactions. The model connections are trained using observations. The supermodelling approach has been tested in a hierarchy of models, from conceptual, to intermediate complexity climate models, and CMIP class models. In this presentation I will show how this approach can reduce errors common errors, such as the double ITCZ, and reduce uncertainties in climate change projections.