Scholarlink

<p>Conventional landslide susceptibility maps for China’s Three Gorges region primarily rely on static inventories and slope-gradient analysis, potentially overlooking other contributing factors. Using seven machine-learning models with eleven conditioning factors, we found that elevation and monsoon rainfall showed the strongest statistical association with historical landslide distribution, together accounting for approximately 90% of the aggregated feature importance, while slope gradient contributes only 11%. Five years of Sentinel-1A SBAS-InSAR data further revealed active deformation in areas not captured by the static inventory. Deformation fields reveal spatially variable subsidence patterns—river-facing toes subside at 12–40&#xa0;mm&#xa0;yr⁻<sup>1</sup>, with subsidence rates decreasing toward the crest. Incorporating these deformation data into the best-performing CNN model resulted in susceptibility upgrades for 43% of pixels and a threefold increase in the very-high susceptibility zone, reclassifying five planned villages as very-high-hazard areas. Moreover, deformation acceleration lags extreme rainfall by three to five days, suggesting time-lagged pore-pressure buildup rather than instantaneous surface runoff response. These findings demonstrate that hydro-gravitational loading is a primary triggering mechanism in high-relief monsoon settings and provide a transferable framework for landslide-risk forecasting in rapidly urbanizing reservoir regions.</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

Hydro-gravitational dominance and differential landslide deformation unveiled by SBAS-InSAR-enhanced susceptibility mapping in Wushan, Three Gorges

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Pinglang Kou,
  • Yuxiang Tao,
  • Jinhu Cui,
  • Zhao Jin,
  • Yijian Huang,
  • Wenli Liang,
  • Andrés Alonso-Rodriguez,
  • Qiang Xu

摘要

Conventional landslide susceptibility maps for China’s Three Gorges region primarily rely on static inventories and slope-gradient analysis, potentially overlooking other contributing factors. Using seven machine-learning models with eleven conditioning factors, we found that elevation and monsoon rainfall showed the strongest statistical association with historical landslide distribution, together accounting for approximately 90% of the aggregated feature importance, while slope gradient contributes only 11%. Five years of Sentinel-1A SBAS-InSAR data further revealed active deformation in areas not captured by the static inventory. Deformation fields reveal spatially variable subsidence patterns—river-facing toes subside at 12–40 mm yr⁻1, with subsidence rates decreasing toward the crest. Incorporating these deformation data into the best-performing CNN model resulted in susceptibility upgrades for 43% of pixels and a threefold increase in the very-high susceptibility zone, reclassifying five planned villages as very-high-hazard areas. Moreover, deformation acceleration lags extreme rainfall by three to five days, suggesting time-lagged pore-pressure buildup rather than instantaneous surface runoff response. These findings demonstrate that hydro-gravitational loading is a primary triggering mechanism in high-relief monsoon settings and provide a transferable framework for landslide-risk forecasting in rapidly urbanizing reservoir regions.