Abstract: Understanding the responses of vegetation to climate change is crucial for predicting changes in ecosystem processes and developing adaptive strategy. In this study, we analyzed the spatial variations of normalized difference vegetation index (NDVI) of different vegetation types in the Loess Plateau in the main growing season from April to October of 2002 to 2019 and its relationship with meteorological factors using geographical detector model and geographically weighted regression. Affected by the fulfillment of “Grain for Green Program”, vegetation coverage on the Loess Plateau had been significantly improved, but the NDVI of steppe and desert were unstable due to their high sensitivity to environmental changes, especially in the summer. Increased sunshine duration indirectly led to the decrease of precipitation, which had a strong photoinhibition effect. The coefficient of variation of sunshine duration had the greatest explanatory power (0.124) for the spatial variation of vegetation NDVI in the Loess Plateau. Deserts in the northwest arid area of the Loess Plateau were affected by altitude, with low air pressure and large water vapor evaporation, which inhibited plant photosynthesis and was positively correlated with the variation of vegetation. In summer, precipitation increased sharply in desert, and the response of vegetation to temperature change slowed down, showing the hysteresis of NDVI change in response to the changes of temperature. From August to October, the sensitivity of high vegetation coverage of broadleaved forests to water stress decreased, resulting in a time lag in the responses to precipitation.

Key words: Loess Plateau, NDVI, geographically weighted regression, geographical detector, climate change.