Human activities accelerate recovery of gross ecosystem product following vegetation disturbances

Post-disturbance vegetation recovery is critical for sustaining ecosystem services, yet the mechanisms underlying this process, particularly the role of human activities, remain insufficiently understood. This study focuses on Zhejiang Province, China, over the period 2000–2020, employing Gross Ecosystem Product (GEP) as a comprehensive indicator to assess vegetation recovery, which integrates key ecosystem services such as water conservation, soil conservation, air purification, and carbon storage. Disturbance dynamics were identified using the LandTrendr algorithm, and high-resolution GEP was estimated from multi-source remote sensing datasets. We examined recovery drivers through a Random Forest model and the SHAP (SHapley Additive exPlanations) analysis. The results show that most disturbed vegetation exhibited strong resilience, with 78.5% recovering within one to two years, while 13.1% of the disturbed area, primarily located in urbanized regions, failed to recover fully. The Random Forest model achieved robust explanatory performance (R 2 = 0.69, RMSE = 3.01), with anthropogenic factors, particularly nighttime light intensity and road proximity, emerging as dominant drivers surpassing climatic and topographic variables. SHAP partial dependence plots revealed that human activities generally accelerated recovery by shortening recovery intervals. Furthermore, a counterfactual simulation excluding anthropogenic factors demonstrated that the average recovery interval would increase substantially from 1.47 to 3.96 years in the absence of human activities. Collectively, these findings underscored the critical role of human activities in facilitating ecosystem recovery following disturbance, providing valuable implications for ecological restoration and sustainable ecosystem management. • Gross ecosystem productivity shows strong negative correlation with disturbance area. • Over 78.5% of disturbed vegetation recovered within one to two years, indicating high resilience. • SHAP analysis identifies human activity factors as dominant drivers of GEP recovery. • Human activities shortened recovery interval by ∼2.5 years compared with natural scenario.