Scholarlink

<p>Landslides is one of the most destructive geological hazards worldwide, where susceptibility assessment serves as a critical component in regional landslide risk management. Based on the geological disaster investigation report and media reports, this study systematically collected and compiled a landslide list and constructed a multi-dimensional landslide influencing system. To address the limitations of conventional methods in spatial representation, class imbalance handling and temporal feature extraction, this study proposes a Buffer-SMOTE-Transformer comprehensive optimization framework (BST). The framework integrates geospatial buffer sampling techniques to refine negative sample selection, employs SMOTE algorithm to effectively resolve class imbalance issues, and incorporates a weighted hybrid Transformer network to enhance modeling capability for complex geographical features. An empirical analysis conducted in China’s Guangdong Province demonstrates that the BST model reveals the varying impacts of sample selection, dataset construction, and model performance on assessment results. The framework achieves significant superiority over conventional machine learning methods (i.e., Random Forest and LGB) in key metrics, with AUC reaching 0.964 and Recall reaching 0.953. These findings not only elucidate the cascading amplification effects of comprehensive optimization in susceptibility modeling but also establish a novel technical method for large-regional-scale geological hazard risk assessment.</p> Graphical Abstract <p>This work represents the effort to address future geological extremes (i.e., landslides) in Guangdong Province, China, significantly supporting decision-makers and authorities in tackling the region’s environmental challenges related to natural disasters in the future. Based on the geological disaster investigation reports and media reports, we systematically collected 4140 landslides and compiled them into a landslide list, and constructed a multi-dimensional landslide influencing system. In this study, to address the limitations of conventional methods in spatial representation, class imbalance handling and temporal feature extraction, we propose a Buffer-SMOTE-Transformer comprehensive optimization framework (BST) to evaluate the landslide susceptibility in this region. The framework integrates geospatial buffer sampling techniques to refine negative sample selection, employs SMOTE algorithm to effectively resolve class imbalance issues, and incorporates a weighted hybrid Transformer network to enhance modeling capability for complex geographical features. Ultimately, this study identifies high-risk and highly sensitive areas for landslides, providing a new technical approach for large-regional-scale geological hazard risk assessment.</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

A Hybrid Neural Network-Based Model for Landslide Susceptibility Mapping

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Long Chen,
  • Yuchen Jia,
  • Sen He,
  • Shanjunxia Wu,
  • Fei Wang

摘要

Landslides is one of the most destructive geological hazards worldwide, where susceptibility assessment serves as a critical component in regional landslide risk management. Based on the geological disaster investigation report and media reports, this study systematically collected and compiled a landslide list and constructed a multi-dimensional landslide influencing system. To address the limitations of conventional methods in spatial representation, class imbalance handling and temporal feature extraction, this study proposes a Buffer-SMOTE-Transformer comprehensive optimization framework (BST). The framework integrates geospatial buffer sampling techniques to refine negative sample selection, employs SMOTE algorithm to effectively resolve class imbalance issues, and incorporates a weighted hybrid Transformer network to enhance modeling capability for complex geographical features. An empirical analysis conducted in China’s Guangdong Province demonstrates that the BST model reveals the varying impacts of sample selection, dataset construction, and model performance on assessment results. The framework achieves significant superiority over conventional machine learning methods (i.e., Random Forest and LGB) in key metrics, with AUC reaching 0.964 and Recall reaching 0.953. These findings not only elucidate the cascading amplification effects of comprehensive optimization in susceptibility modeling but also establish a novel technical method for large-regional-scale geological hazard risk assessment.

Graphical Abstract

This work represents the effort to address future geological extremes (i.e., landslides) in Guangdong Province, China, significantly supporting decision-makers and authorities in tackling the region’s environmental challenges related to natural disasters in the future. Based on the geological disaster investigation reports and media reports, we systematically collected 4140 landslides and compiled them into a landslide list, and constructed a multi-dimensional landslide influencing system. In this study, to address the limitations of conventional methods in spatial representation, class imbalance handling and temporal feature extraction, we propose a Buffer-SMOTE-Transformer comprehensive optimization framework (BST) to evaluate the landslide susceptibility in this region. The framework integrates geospatial buffer sampling techniques to refine negative sample selection, employs SMOTE algorithm to effectively resolve class imbalance issues, and incorporates a weighted hybrid Transformer network to enhance modeling capability for complex geographical features. Ultimately, this study identifies high-risk and highly sensitive areas for landslides, providing a new technical approach for large-regional-scale geological hazard risk assessment.