A multivariate analysis of atmospheric drivers for Western European heatwaves

Kurzfassung

Understanding the dynamics of heatwaves is critical for accurate climate risk assessment. Traditional definitions, based solely on surface temperature thresholds, can detect heatwaves but cannot be used to analyze the complex, multivariate nature that causes them. We apply a spatiotemporal Variational Autoencoder (VAE) to ERA5 reanalysis data to identify compact representations of multivariate, year-round heatwave patterns over Western Europe. Focusing on key atmospheric variables (e.g. circulation, humidity, temperature, geopotential height, cloud cover, stream function, and radiation), we extract 11 d heatwave samples from ERA5 reanalysis data over the North Atlantic, centered on near-surface temperature extremes in Western Europe. The VAE was trained on heatwave samples from 1941 to 1990 and evaluated using samples from 2001 to 2022, effectively clustering heatwave events by season and identifying known dynamical regimes, such as summer blocking highs and winter omega blocks. The VAE model captures the interplay and temporal evolution between different atmospheric variables in their contributions to heatwaves over Western Europe. Notably, the VAE identifies distinct atmospheric circulation patterns that align with seasonal dynamics even when applied to detrended and deseasonalized data. When analyzing non-detrended data, recent summer heatwaves occupy a distinct region in the latent space, consistent with the strong warming-driven intensification of these events. However, analyses of detrended data reveal significant structural shifts in atmospheric configurations across all seasons, indicating that the VAE captures changes in the multivariate structure of heatwave-related variables beyond simple mean warming. Composite anomaly maps further show coherent pre-onset patterns across variables. These results demonstrate the utility of VAEs for extracting physically meaningful, multivariate representations of heatwave dynamics from reanalysis data, highlighting changes in their atmospheric drivers over recent decades.