Abstract
In order to determine the evolution characteristics of net primary productivity of vegetation in Nanchang City and the main driving factors influencing its spatiotemporal evolution, based on the ArcGIS and Matlab platforms, ReliefF, Random Forest (RF),BP neural network, GRNN machine learning algorithm and geographic detector were used to quantitatively evaluate the evolution characteristics and spatiotemporal driving factors of Nanchang City from 1998 to 2015.The results show: 1) From a temporal perspective, NPP overall shows a fluctuating upward trend with distinct seasonal variations; spatially, it follows a distribution pattern of higher values in the middle and lower values around the edges; 2) The ReliefF algorithm has the highest fitting accuracy and is more suitable for regression analysis of NPP, with both algorithms indicating that air temperature and precipitation have the most significant impact on NPP evolution; 3) According to the results of the geographic detector, the NPP in Nanchang City is most significantly influenced by precipitation factors spatially, while the temporal dimension is dominated by human factors. In-depth study of the evolution characteristics of NPP can provide a scientific basis for quantifying the health of regional ecosystems and the balance of the ecological environment under the background of climate change.