Abstract
Understanding how the changes of vegetation sensitivity to climate and human activities under drought is essential for evaluating ecosystem resistance in subtropical humid regions. This paper focused on Guangdong Province, China, and used SPEI-3 (Standardized Precipitation Evapotranspiration Index) to identify short-term drought events (< 6 months). LMG (Lindeman-Merenda-Gold method) and XGBoost (Extreme Gradient Boosting) were used to quantitatively analyze the contributions of temperature, precipitation and nighttime light (NTL) to the NDVI of evergreen forest, grassland and urban region. LMM (Linear Mixed Model) with month and year as random effects were applied. The conclusions revealed that: (1) Short-term drought significantly reduced the sensitivity of vegetation to temperature in the south of the Tropic of Cancer, while vegetation in the north of the Tropic of Cancer maintained a stable relationship with temperature. (2) Under drought, the temperature contribution to evergreen forest decreased the most (10-13%), followed by urban region (6-15%) and grassland (7-12%). The precipitation contribution to evergreen forest increased by 7-10% and 2-7% for grassland. (3) Drought weakened the positive temperature-NDVI correlation in the west located in the south of the Tropic of Cancer, while the positive temperature-NDVI correlation persisted in the north located in the north of the Tropic of Cancer. (4) Under drought, vegetation maintained a stable positive response to climate. The temperature-NDVI relationship in more than half of the cities shifted from positive to negative.