Abstract

Extreme weather events and water security are challenges for global society across multiple dimensions, from water-related disasters and complex variations in the global water cycle to the unsustainable water supply for growing populations and highly irrigated agricultural systems. The first step towards addressing these challenges is to accurately identify weather phenomena that often lead to extreme events. Being able to recognise these phenomena in space and time can facilitate understanding of their developing mechanisms and life cycles. Machine and deep learning, topological data analysis and causal time series analysis offer a wide range of automatic methods that can help recognise weather phenomena and their quantitative assessment in a rapidly increasing amount of observational and simulated climate data. In this talk, I will present recent applications of machine and deep learning, topological data analysis and causal time series analysis to study weather phenomena, such as atmospheric rivers, the Indian summer monsoon, and atmospheric blocks. I will also give perspectives on major challenges where machine learning methods and causal analysis have the potential to advance the state-of-the-art in extreme weather detection.