Nonlinear interactions amplify the most extreme midlatitude heatwaves

Recent occurrences of record-breaking heat extremes and their profound societal impacts on health, infrastructure, food systems, and the energy sector underscore the urgent need to improve our physical understanding and modeling capacities for future projections. In mid-latitude regions, persistent high-pressure systems and dry soils have been identified as key contributors to heatwave severity. Moreover, non-linear interactions between these two drivers and temperature have been suggested to play a critical role in some of the most extreme recent heat events, such as the 2021 Pacific-North America heatwave (Bartusek et al., Nat. Clim., 2022). However, the universality and regional significance of such non-linear interactions remain largely unquantified.

Using an explainable machine learning approach, we quantitatively decompose surface air temperature anomalies during heat extremes into three components: direct contributions from (i) geopotential height anomalies, (ii) soil moisture deficits, and (iii) the interaction between the two. Our analysis reveals that non-linear interactions make statistically significant contributions across 19% of the land area in the northern hemisphere mid-latitudes (40°N–60°N). In these regions, the interactive contribution increases with temperature at a rate of 0.1 K/K when temperatures exceed a critical threshold of 4.0 K above the local summer mean. Hotspots of such behavior are especially pronounced in Central Europe, where 40% of the land area exhibits significant non-linear interactions, amplifying the most extreme heatwave events by up to 13%.

Furthermore, we identify a 2.4-fold increase in the regional mean non-linearity of interactions in Central Europe over the past 45 years, accompanied by a 25% expansion in the affected area. This accounts for 18% of the observed widening in the temperature distribution’s upper tail reported in other studies (Kornhuber et al., PNAS, 2024). Additionally, our findings show that CMIP6 climate models underestimate the non-linearity of extratropical interactions by 80%, contributing to biases in projections of extreme heat changes. Our findings underscore the critical role of these non-linear physical processes in amplifying extreme heatwave events, emphasizing the need to account for these processes in climate models to better anticipate and mitigate the impacts of climate extremes in current and future climates.