Predicting future climate extremes in China using causality driven statistical models based on scenario model simulations

Abstract:

Accurate prediction of extreme weather and climate events in the current and future climate remains one of the major challenges in scientific research and any upcoming future climate service, e.g. for environmental disaster risk assessment. This poses a problem from the policy and stakeholder point of view (e.g., for disaster risk reduction and mitigation), where failure to take appropriate precautionary action before the occurrence of catastrophic events can have disastrous consequences. While the performance of climate models has improved substantially over the past few decades, predicting extreme events remains challenging. This is partially due to limited simulation periods, therefore the full intensity distribution is not sampled sufficiently in the high end tail. It is also due to their coarse spatial resolution (and therefore not being able to represent all characteristics of those events) as well as their inability to fully represent necessary processes on relevant scales for extreme event generation: from small to synoptic and hemispheric scales. The latter will provide important driving information via scale interactions and is generally seen as better captured by AOGCMs.

In this presentation, we provide an overview of a new project, predicting future climate extremes in China by blending predictive performance of generalized linear models with causal network discovery (PRE-CAX). We aim to develop a novel approach by combining knowledge, which has been developed in recent years (including causality network discovery technique), to produce a better prediction of extreme event occurrences – extreme windstorms and extreme precipitation associated with (a) Mei-Yu (or Meiyu-Baju) front (MBF) and (b) typhoons in the Western North Pacific (WNP), using climate model outputs in CMIP5/6. Consequently, a better risk assessment will be enabled by using these tools in any future model simulation setting. This information can essentially support the climate service component for improving the decision making chain e.g., for disaster risk reduction and mitigation strategies in China. The methodology and preliminary findings are also discussed.