[Landscape Ecological Risk and Its Driving Factors in Jiangxi Province Based on Geodetector-GTWR]

The landscape pattern has been under stress by both human and natural factors in recent years， resulting in many ecological risks. In this study， the landscape pattern index method was used to assess the landscape ecological risk （LER） in Jiangxi Province from 2000 to 2020， and the spatial autocorrelation approach was applied to investigate the spatial clustering patterns of LER. Then， we used exploratory regression to select the driver combination with the strongest explanatory power and the smallest multicollinearity problem from nine candidate factors. The Geodetector and GTWR model were used to explore the relationship between landscape ecological risk and its driving factors. The results were as follows： ① During the study period， the mean value of landscape ecological risk in Jiangxi Province increased from 0.071 2 to 0.072 5， which is a generally low level. ② From 2000 to 2020， the low-risk area decreased in size by 6.14%， whereas the medium-high risk area and high-risk area expanded by 1.06% and 1.45%， respectively. ③ There was significant spatial autocorrelation of landscape ecological risk in Jiangxi Province. Although the area of spatial agglomeration decreased， the distribution pattern was relatively stable， with low-low agglomeration area in the west， central region， and south and high-high agglomeration area in the north. ④ The explanatory power of natural factors for landscape ecological risk in Jiangxi Province is high， and the explanatory power of the interaction among all factors is stronger than that of a single factor. ⑤ The driving factor combination of population， GDP， annual precipitation， DEM， and NDVI is the combination with the strongest explanatory power. The maximum variance inflation factor （VIF） value of this driving factor combination is 0.06. Population and GDP have positive driving effects on landscape ecological risk in Jiangxi Province， whereas NDVI， annual precipitation， and DEM have inhibitory effects. From the fitting effect， the AICc value and R_adj² value of the GTWR model are -66 139.80 and 0.55， respectively， which is better than the OLS and GWR model result. The results of this study provide a reference for ecological risk management in Jiangxi Province and other regions.