Separating the Contributions of Human Activities and Climate Change to Vegetation Dynamics on the Loess Plateau Using Machine Learning

Abstract

Against the dual backdrop of climate change and ecological restoration projects, vegetation on the Loess Plateau has experienced rapid recovery. However, the relative contributions of climate change and human activities to these vegetation changes remain controversial. This study employs the Leaf Area Index (LAI) as an indicator of vegetation greenness, identifies the onset of significant ecological restoration effects in different regions using change-point detection, and separates the contributions of human activities and various climate factors through machine learning and segmented residual analysis. The results indicate that the LAI of the Loess Plateau exhibits a fluctuating upward trend. The timing of significant ecological restoration effects shows pronounced spatial heterogeneity, with an average onset year of 2013. The machine learning model, which considers the optimal 3-month cumulative rainfall scale, achieves an average R² of 0.87. Based on this model and residual trend analysis, human activities are found to dominate 69.6% of the vegetation changes, while climate change contributes 30.4%. After 2010, the promoting effect of human activities became more pronounced, whereas climate change shifted from a promoting role before the change-point to an inhibiting role afterwards. This study systematically analyzes the mechanisms of vegetation change on the Loess Plateau, providing a scientific basis for formulating regional ecological management policies.