Influence of cumulative extreme climate risk on agricultural ecosystem resilience in the karst mountainous region of southwestern China

Agricultural ecosystem resilience (AES) underpins food security and sustainable agricultural development. Using panel data for 60 prefecture-level administrative units in the karst mountainous region of southwestern China from 2005 to 2023, this study adopts a cumulative-disturbance perspective to construct a Cumulative Extreme Climate Risk (CECR) index and an AR(1)-based recovery-persistence indicator derived from first-order autoregressive coefficients of gross primary productivity (GPP) time series. The latter captures the short-term recovery dimension of AES. Within an analytical framework that integrates cumulative disturbance, spatial diffusion, and socioeconomic modulation, we combine the entropy weighting method, a dynamic spatial Durbin model, mediation analysis, and panel threshold models to examine the multiple pathways linking CECR to short-term recovery dynamics. The results reveal pronounced spatial heterogeneity and temporal variability in AR(1), with a degree of spatial mismatch between its evolution and that of CECR. CECR is positively associated with AR(1), indicating that accumulated climatic disturbances are linked to greater recovery persistence and slower recovery. The indirect spatial effect is substantially larger than the direct effect, suggesting that the association operates not only locally but also through marked cross-regional spillovers. Rural residents' per capita disposable income and the urban-rural income ratio provide evidence of socioeconomic transmission, highlighting resource availability and allocation as potential links between climatic disturbance and system recovery. The CECR-AR(1) association is also strongly condition-dependent: it becomes more pronounced under higher poverty incidence, warmer temperature regimes, and tighter household resource constraints, consistent with condition-dependent amplification under greater vulnerability. By placing agricultural resilience within a coupled socio-ecological and cumulative-disturbance framework, the study extends existing resilience research and provides an empirical basis for differentiated policies to support sustainable agricultural development in ecologically fragile regions.