Grassland vegetation decline is exacerbated by drought and can be mitigated by soil improvement in Inner Mongolia, China

Grassland activity is highly susceptible to drought while drivers from climate and soil attributes can largely affect drought propagation. However, understanding how these drives regulate the risk of vegetation decline under drought conditions remains limited, potentially impeding the adoption of appropriate adaptation strategies. To address this knowledge gap, we conducted a case study focusing on grassland activity in Inner Mongolia, China. In this study, we applied copula theorem to estimate the conditional probabilities of vegetation decline under drought conditions. Additionally, we utilized a structural equation model and a machine learning approach to identify the relative contributions of external drivers to the risk of vegetation decline. Our findings demonstrated a positive correlation between anomalies in vegetation activity and the status of water balance, and grassland vegetation in drier regions exhibited a more rapid response to water deficit. Increasing water deficit continuously reduced vegetation activity with risks of 77.27 %, 83.83 %, and 88.35 % under moderate, severe, and extreme drought conditions, respectively. Furthermore, the risks of vegetation decline under drought conditions were primarily governed by climate attributes, followed by soil properties and topography. Soil with high soil organic carbon stock content contributed significantly to mitigating the adverse effects of drought on grassland vegetation. In addition, we detected nonlinear patterns among environmental drivers and vegetation decline risks caused by drought. These findings highlight the importance of climate, soil properties, topography, and their intricate interconnections in regulating vegetation decline. This knowledge provides valuable insights into drought risk management for vegetation in advance and offers potential solutions to enhance vegetation resistance in the face of extreme drought events.