Drivers of agricultural ecosystem resilience under compound atmospheric-soil drought in Northeast China

Agricultural ecosystem resilience is important to secure food supply and support sustainable progress during climate change. This study examines agricultural ecosystems in Northeast China and quantitatively evaluates their resilience using Gross Primary Productivity (GPP) data. Methodologically, this study integrates compound drought indices with resilience indicators through machine learning to identify nonlinear threshold effects in agricultural systems. Resilience was quantified using the first-order autoregressive coefficient derived from the GPP time series, which reflects the ecosystem's recovery rate after disturbances. Results show significant spatial heterogeneity in resilience across the region. Areas with no significant change in resilience cover 41.82 %, while areas with declining and increasing resilience account for 33.31 % and 24.87 %, respectively. Compared with non-agricultural regions, agricultural ecosystems in Northeast China display a trend of resilience improvement. Analysis of driving factors identifies precipitation and the proportion of large-scale cropland as the most influential. Although compound drought severity has a relatively minor impact in static resilience assessments, it plays a significant role in resilience evolution trends. In areas where resilience increases, a clear negative-to-positive switching effect appears, indicating that mild drought accelerates resilience enhancement, while extreme drought slows it down. This study identifies the complex regulatory mechanisms and nonlinear responses shaping agricultural ecosystem resilience in Northeast China, providing scientific support for resilience improvement and sustainable development. Targeted management measures, such as optimizing crop configuration and enhancing drought adaptation, can help strengthen agricultural resilience in the region. The findings offer valuable guidance for developing differentiated agricultural management strategies in the region. • NE China farm resilience varies, with 41.8 % stable, 33.3 % falling, 24.9 % rising. • Identified higher resilience in farmland than non-farmland. • Revealed nonlinear threshold in large-scale farmland • Drought severity impact shifts from positive to negative, shows long-term effects.