Three-way circulation–soil moisture–temperature feedbacks amplify the most extreme mid-latitudinal heat waves

Abstract The most extreme temperatures are increasing more rapidly than moderate ones, yet the physical mechanisms behind this acceleration remain unclear. Here, we show that nonlinear land–atmosphere interactions play a pivotal role in intensifying heatwave trends across midlatitude humid regions. Using an explainable machine learning framework, we quantify a nonlinear nature of these interactions, characterized by critical thresholds in temperature and soil moisture, beyond which heat extremes are strongly amplified. This behavior is evident across 30% of midlatitude humid areas, contributing to a 10% increase in individual heatwave magnitudes. In these regions, 15% of the accelerated rise in extreme temperatures is attributable to the increasing probability of crossing the critical thresholds—a value that rises to 53% in Central Europe, a prominent heatwave hotspot. We also find that biases in CMIP6 extreme heat projections stem from an underestimation of these nonlinear interaction processes, which is further led by a lagged soil moisture response during heatwave events. After statistically correcting these projections, we estimate that the difference between the most extreme and moderate temperature could be 4.2 ± 0.16 times greater than currently projected in midlatitude humid regions by the end of the 21st century.