干旱区地理2026，Vol.49Issue(9)：1826-1839,14.DOI:10.12118/j.issn.1000-6060.2025.703
基于XGBoost-SHAP方法的天山中段景观生态风险时空演变及驱动因素
Spatiotemporal evolution and driving factors of landscape ecological risk in the middle section of the Tianshan Mountains based on the XGBoost-SHAP method
摘要
Abstract
Landscape ecological risk refers to the potential adverse effects on ecological processes due to changes in regional landscape structure.Understanding the spatiotemporal dynamics and driving factors of landscape eco-logical risk is crucial for optimizing territorial spatial planning.Focusing on the middle section of the Tianshan Mountains,this study applied a landscape ecological risk assessment model to analyze the spatiotemporal evolu-tion of risk from 2005 to 2023.The extreme gradient boosting-Shapley additive explanations method was em-ployed to identify the driving factors and their interactions.The results revealed the following:(1)Temporally,landscape ecological risk in the study area remained generally stable between 2005 and 2023,with a slight de-crease in the ecological risk index.Although 66.12%of the area maintained consistent risk levels,33.88%experi-enced transitions between risk levels,indicating overall stability with localized sensitivity.(2)Spatially,land-scape ecological risk exhibited a polarized"low in the north,high in the south"pattern,roughly delineated by the main ridge of the Tianshan Mountains in a"人"-shaped configuration.High-risk areas were concentrated in the southern Taklimakan Desert,whereas low-risk areas were predominantly in the northern Ili River Valley.Land-scape composition significantly influenced this spatial differentiation.(3)In terms of driving factors,single-factor analysis indicated that natural geographical factors,particularly the normalized difference vegetation index(ND-VI)and temperature,primarily determined the spatial pattern of landscape ecological risk in the middle Tianshan Mountains.Interaction analysis showed that high-risk areas are collectively driven by the coupled effects of"low NDVI and relatively high temperature"and"low GDP and high population density".These multi-threshold non-linear feedback mechanisms warrant particular attention for ecological risk management in arid regions.关键词
景观生态风险/XGBoost-SHAP/时空演变/驱动因素/天山中段Key words
landscape ecological risk/XGBoost-SHAP/spatiotemporal evolution/driving factors/middle sec-tion of the Tianshan Mountains引用本文复制引用
毕文焌,闫庆武,赵甫心,移明昊,吴子豪..基于XGBoost-SHAP方法的天山中段景观生态风险时空演变及驱动因素[J].干旱区地理,2026,49(9):1826-1839,14.基金项目
第三次新疆综合科学考察项目(2022xjkk1004) （2022xjkk1004）
国家自然科学基金(42101459,42201447)资助 （42101459,42201447）
