Identifying Probabilistic Weather Regimes Targeted to a Local-Scale Impact Variable

Abstract: Identifying large-scale atmospheric patterns that modulate extremes in local-scale variables such as precipitation has the potential to improve long-term climate projections as well as extended-range forecasting skill. We propose a novel machine learning method, RMM-VAE, for identifying probabilistic weather regimes targeted to a local-scale scalar impact variable. Based on a variational autoencoder architecture, this method combines targeted and non-linear dimensionality reduction with probabilistic clustering in a coherent architecture. We apply the new method to identify robust circulation patterns that are predictive of precipitation over Morocco while still capturing the complete phase space of atmospheric dynamics over the Mediterranean. The results are compared to three existing approaches - two established linear methods and another machine learning method. The RMM-VAE method performs well across all different objectives, outperforming linear methods in terms of reconstructing the input space and predicting the target variable, and the other machine learning method in terms of identifying robust and persistent clusters. The results reveal a trade-off between the different objectives of targeted clustering and highlight the benefits of the novel RMM-VAE method in terms of balancing these different objectives for various climate applications.

Biography: Fiona is a PhD student at the University of Reading, working with Prof Marlene Kretschmer and Prof Ted Shepherd on improving seasonal forecasts using causal models of atmospheric teleconnections. Prior to starting her PhD, Fiona worked for three years at a not-for-profit organisation on the alignment of the European financial sectors with climate goals. Fiona co-developed an open-source python package for the comparison and evaluation of statistical bias adjustment methods of climate models and holds a degree in Physics (MSc, University of Edinburgh) as well as Environmental Change and Management (MSc, University of Oxford).