聚类分析
可解释性
生态学
加权
人口
城市化
计算机科学
环境资源管理
生态系统理论
大都市区
地理
数据挖掘
数据库扫描
水准点(测量)
离群值
环境科学
成对比较
突出
人工智能
鉴定(生物学)
作者
Sicheng Yan,Zhenqin Shi,Wenbo Zhu,Jingeng Huo,Zechen Wang,Jichao Ma,Yanhui Yan
摘要
ABSTRACT Worsening ecological security crises, driven by human activities and rapid urbanization, require more effective ecological security patterns (ESP). Traditional ESP construction relies on multi‐factor overlay analysis and subjective weighting for source identification and resistance surface construction, introducing homogenization bias and failing to capture nonlinear ecological relationships. Moreover, previous studies typically improved individual construction stages in isolation rather than within a unified framework. This study developed a data‐driven framework coupling K‐means clustering algorithm, LightGBM‐SHAP interpretable machine learning, and circuit theory to construct and optimize the ESP for the Zhengzhou Metropolitan Area (ZMA). K‐means clustering identifies ecological sources from intrinsic data distributional patterns without subjective weighting, while LightGBM‐SHAP constructs the ecological resistance surface and reveals how natural and anthropogenic factors influence ecological security. The results showed the following: (1) 63 ecological sources concentrated in western mountainous areas and 131 ecological corridors along with ecological pinch points, barrier points, and breakpoints were identified. (2) Ecological security levels decreased from the western mountainous areas to the eastern plains, with over half the study area classified as medium‐low or low security. The interpretability analysis revealed nonlinear relationships and threshold effects in key factors including elevation, slope, population density, and nighttime light. (3) A spatial optimization framework of “One Core, Two Barriers, Three Corridors, Four Zones” was proposed with differentiated restoration strategies for each zone. This study presents a replicable integrated framework that reduces subjectivity across multiple ESP construction stages and provides a reference for ecologically fragile regions facing rapid urbanization and conservation pressures.
科研通智能强力驱动
Strongly Powered by AbleSci AI