Distinct Energy and Hydro-Thermal Coupling Regimes at the Land-Atmosphere Interface Shape Global Compound Hot-Dry Extremes

Under global climate change, intensified land-atmosphere coupling has amplified the synergy between droughts and heatwaves, triggering a nonlinear escalation of compound hot-dry events (CHDEs) that threatens Earth systems. However, current research often lacks rigorous process-based classification at the atmosphere-soil interface, and understanding of energy partitioning and hydro-thermal feedback mechanisms remains limited, impeding systematic comprehension of these extremes. Using daily ERA5 and GLDAS data from 1965–2024, this study develops a mutually exclusive event identification framework based on four variables—vapor pressure deficit (VPD), air temperature (Tair), soil moisture (SM), and soil temperature (Tsoil)—classifying events into Single-Atmosphere (SA), Single-Soil (SS), and Compound Atmosphere-Soil (CAS) types. We systematically analyze event characteristics, identify high-risk regions, and conduct nonlinear trend analysis using Ensemble Empirical Mode Decomposition (EEMD). A progressive framework integrating event evolution analysis, Copula-based dependence modeling, and Structural Equation Modeling (SEM) is employed to elucidate the underlying physical mechanisms. Key findings are as follows: (1) Spatial patterns reveal mechanistic divergence. SA events display a "tropical zonal clustering" pattern with the highest frequency (8.12 events/decade) but shortest duration (5.09 days) and moderate intensity. SS events show a scattered distribution along land-sea margins with intermediate frequency (3.82 events/decade), longest duration (6.37 days), and lowest intensity. In contrast, CAS events expand extensively across mid-latitudes with the largest frequency increase (251%) and highest intensity (6.60 standardized units), marking a global shift from tropical single-process dominance toward mid-latitude land-atmosphere coupling dominance. (2) Evolution trends exhibit nonlinear acceleration. EEMD outperforms traditional linear regression in trend significance and fitting accuracy, demonstrating superior capability in capturing the nonlinear dynamics of extreme events. SA events intensify persistently in tropical regions but decelerate in later periods; SS events exhibit regional heterogeneity with abrupt shifts in arid zones; CAS events show synchronized global acceleration, with late-period growth rates exceeding early-period rates by 119%–232%. (3) Physical mechanisms differ fundamentally. Analysis reveals that SA events represent rapid boundary-layer responses to radiative forcing (energy-limited) with passive soil moisture depletion. SS events are driven by cumulative hydrological deficits (memory-dominated) with significant recovery lags. CAS events involve synergistic positive feedback: once SM drops below critical thresholds, a self-reinforcing loop (SM↓→LH↓/SH↑→Tair↑→VPD↑→SM↓↓) is triggered, fundamentally altering surface energy partitioning and hydro-thermal coupling regimes. Copula and SEM analyses confirm that SA events exhibit linear synchronous dependence under atmospheric forcing; SS events show lower-tail threshold effects dominated by soil memory; CAS events demonstrate significant cumulative atmospheric driving effects (with soil response lagged by 10–15 days) and enhanced tail dependence under extreme conditions, reflecting strong coupling between atmospheric triggers and soil feedbacks. Furthermore, large-scale climate modes such as ENSO modulate these processes by regulating regional background wet-dry states. This study establishes a comprehensive framework from event identification and characterization to mechanistic interpretation, elucidating the transformation of global hot-dry risks from "tropical single-process dominance" to "mid-latitude land-atmosphere coupling dominance," providing a robust scientific basis for monitoring, early warning, and risk management of compound extreme events.