Distinct intraseasonal pathways driving rapidly escalating heat extremes in North China and their subseasonal predictability

Rapidly escalating heat extremes (REHEs), characterized by abrupt onset and limited preparation time, pose disproportionate risks to public health, agriculture, and regional economies, yet their governing mechanisms and subseasonal predictability remain poorly understood. This study identifies 107 boreal-summer REHEs over North China during 1940–2024 and classifies them into four spatially distinct types using an unsupervised machine-learning approach. A thermodynamic diagnosis identifies horizontal temperature advection and adiabatic heating as the two primary warming pathways, whose relative contributions differ systematically among the four types. The four types are further distinguished by their intraseasonal forcing configurations, spanning dual high-latitude–tropical interference (Type I), predominantly tropical forcing (Type IV), primarily high-latitude wave-train forcing (Type III), and a superposition of mid–high-latitude wave trains (Type II). Using the subseasonal-to-seasonal dynamical prediction models, we demonstrate pronounced type dependence in forecast skill: tropical-driven events (Type IV) are the most predictable, whereas Type II events show little skill because their evolution is dominated by the least predictable extratropical intraseasonal mode. The predictability hierarchy mirrors that of the underlying intraseasonal modes, with the Boreal Summer Intraseasonal Oscillation exhibiting the longest prediction horizon. These results establish a process-based framework linking intraseasonal dynamics to rapid heat escalation and provide a physical basis for improving subseasonal prediction of high-impact heat extremes.