Scholarlink

This study proposes a multi-model ensemble framework for landslide susceptibility assessment in Shenzhen, China, integrating static and dynamic spatiotemporal data. A Stacking ensemble model, combining CNN, MLP, GRU, and SVR, incorporates twelve static environmental factors with time-series rainfall and NDVI data, achieving high predictive accuracy (AUC 0.9893, F1 0.9705) and reduced uncertainty compared to individual models. Additionally, a Temporal-Spatial Deep Neural Network (TSDNN) framework, leveraging MSCNN for spatial feature extraction and Bi-LSTM for temporal dynamics, enhances susceptibility mapping by effectively fusing multi-source data. Validation using InSAR deformation data confirms that high-susceptibility zones align with historical landslide occurrences, ensuring reliable and precise risk assessment for disaster mitigation.

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

Multi-model Integrated Landslide Susceptibility Mapping Method Based on Static-Dynamic Spatiotemporal Data Fusion

  • 章节

  • 发表日期：

    2025年

查看全文
  • Yi He

摘要

This study proposes a multi-model ensemble framework for landslide susceptibility assessment in Shenzhen, China, integrating static and dynamic spatiotemporal data. A Stacking ensemble model, combining CNN, MLP, GRU, and SVR, incorporates twelve static environmental factors with time-series rainfall and NDVI data, achieving high predictive accuracy (AUC 0.9893, F1 0.9705) and reduced uncertainty compared to individual models. Additionally, a Temporal-Spatial Deep Neural Network (TSDNN) framework, leveraging MSCNN for spatial feature extraction and Bi-LSTM for temporal dynamics, enhances susceptibility mapping by effectively fusing multi-source data. Validation using InSAR deformation data confirms that high-susceptibility zones align with historical landslide occurrences, ensuring reliable and precise risk assessment for disaster mitigation.