Assessment of Vegetation Dynamics in Xinjiang Using NDVI Data and Machine Learning Models from 2000 to 2023

This study utilizes NASA's Normalized Difference Vegetation Index (NDVI) data from the Google Earth Engine (GEE) platform and employs methods such as mean analysis, trend analysis, and the Hurst index to assess NDVI dynamics in Xinjiang, with a particular focus on desert, meadow, and grassland vegetation. Furthermore, multiple linear regression, random forest, support vector machines, and XGBoost models are applied to construct and evaluate the NDVI prediction models. The key driving forces are identified and ranked based on the results of the optimal model. Changes in the vegetation cover in response to these driving forces are analyzed using the Mann-Kendall test and partial correlation analysis. The results indicate the following: (1) From 2000 to 2023, the annual variation in NDVI in Xinjian fluctuates at a rate of 0.0012 per year. The intra-annual trend follows an inverted U shape, with meadow vegetation exhibiting the highest monthly NDVI fluctuations. (2) During this period, the annual average NDVI in Xinjiang ranges from 0 to 0.3, covering 74.74% of the region. Spatially, higher NDVI values are observed in the north and northwest, while lower values are concentrated in the south and southeast. (3) The overall slope of the variation in NDVI in Xinjiang between 2000 and 2023 ranges between -0.034 and 0.047, indicating no significant upward trend. According to the Hurst index, future projections suggest a shift from vegetation improvement to potential degradation. (4) Machine learning models are developed to predict NDVI, with random forest and XGBoost showing the highest precision. Soil moisture, runoff, and potential evaporation are identified as key drivers. In the last 24 years, the temperatures in Xinjiang have generally increased, while precipitation, soil moisture, and runoff have declined. There is a significant negative correlation between NDVI and both temperature and potential evaporation, while the correlation between NDVI and precipitation, soil moisture, and runoff is positive and significant, with distinct spatial variations throughout the region. The overall trend of vegetation cover in Xinjiang has been increasing, but the future outlook is less promising. Enhanced environmental monitoring and protective measures are essential moving forward.