Vulnerability classification.

To realize the sustainable development of agro-ecology in Northeast China and help our country achieve the goal of wide-area agro-ecological maintenance, this study takes “sensitivity⁃resilience⁃pressure” as the evaluation model, selects 13 evaluation indicators and adopts the principal component analysis method to calculate and grade the vulnerability of agroecosystems in the three northeastern provinces and 36 prefectural-level cities from 2004 to 2022 using principal component analysis, and the driving factors were evaluated using parameter-optimized geodetector methods aiming at exploring the changes in spatial and temporal patterns of agroecological vulnerability in the northeastern region and the influence of the driving factors of each of its indicators on its vulnerability index. The conclusions are as follows: (1) In terms of temporal evolution, the overall vulnerability of agricultural ecosystems in Northeast China has shown a fluctuating downward trend. The proportion of different vulnerability levels has fluctuated significantly, with moderately vulnerable areas continuously transitioning toward mild vulnerability, indicating a positive trend. (2) In terms of spatial evolution, the vulnerability of the agricultural ecosystem in the northeastern region shows a distribution pattern of “high in the southwest, low in the northeast,” gradually decreasing from south to north, with an overall trend of gradual improvement. (3) In terms of driving forces, the spatial pattern is mainly determined by the terrain, with surface vegetation playing a core regulatory role by buffering climate pressures and enhancing ecological functions. At the same time, agricultural production conditions serve as a key human intervention to enhance resilience. This relies on the deep interaction between surface vegetation and human activity factors, which, through positive feedback between ecological and economic systems, and synergy between vegetation and irrigation, jointly drive the system’s resilience pattern.