Scholarlink

<p>Understanding how land use/land cover (LULC) dynamics influence landslide hazards is crucial for mountainous urban regions experiencing rapid landscape disturbance. This study integrates multi-year LULC datasets (2017–2024), neighborhood-scale spatial analysis, landslide inventories, machine-learning–based susceptibility modeling, and LULC transition-path analysis to clarify their interactions in Lin’an District, China. Results show that landslide-prone locations consistently exhibit reduced tree cover and increased Built-up areas and Rangeland before failure. Incorporating LULC information (2018) significantly improved the 2019 landslide susceptibility model, with 10-fold cross-validation confirming higher and more stable AUC values compared with the baseline model. SHAP analysis identified Built-up areas, Farmland, and elevation as major contributors, while tree cover suppressed susceptibility; these interpretations were validated independently through field surveys. LULC transition matrices further revealed a characteristic evolutionary sequence—Disturbance → Degradation → Reconstruction—linking forest loss, slope damage, and post-event human modification. The findings highlight the dominant role of LULC transitions in landslide development and demonstrate the value of integrating land use/land cover change (LULCC) monitoring with susceptibility modelling for sustainable hazard management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Land use/land cover controls and neighborhood signatures of landslide hazards in a mountainous urban region of China

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • Jingyu Xia,
  • Lixia Chen,
  • Xingchen Zhang,
  • Shilin Zhu,
  • Jiang Wu,
  • Xuri Shi,
  • Yishun Zhang

摘要

Understanding how land use/land cover (LULC) dynamics influence landslide hazards is crucial for mountainous urban regions experiencing rapid landscape disturbance. This study integrates multi-year LULC datasets (2017–2024), neighborhood-scale spatial analysis, landslide inventories, machine-learning–based susceptibility modeling, and LULC transition-path analysis to clarify their interactions in Lin’an District, China. Results show that landslide-prone locations consistently exhibit reduced tree cover and increased Built-up areas and Rangeland before failure. Incorporating LULC information (2018) significantly improved the 2019 landslide susceptibility model, with 10-fold cross-validation confirming higher and more stable AUC values compared with the baseline model. SHAP analysis identified Built-up areas, Farmland, and elevation as major contributors, while tree cover suppressed susceptibility; these interpretations were validated independently through field surveys. LULC transition matrices further revealed a characteristic evolutionary sequence—Disturbance → Degradation → Reconstruction—linking forest loss, slope damage, and post-event human modification. The findings highlight the dominant role of LULC transitions in landslide development and demonstrate the value of integrating land use/land cover change (LULCC) monitoring with susceptibility modelling for sustainable hazard management.