Scholarlink

<p>The synclinal valleys in the outer wedge of the Indo-Burmese Ranges (IBR), part of zipper tectonics, are more prone to earthquake-induced liquefaction due to thick unconsolidated, immature sediments with shallow groundwater conditions. The present study, a data-driven ensemble learning approach—Random Forest (RF)—was employed to assess liquefaction susceptibility in the Barak Valley by considering nine contributing factors such as geology, land use/land cover, lineament density, slope, peak ground acceleration, predominant frequency, peak amplification, shear wave velocity, and distance from water bodies. The RF model implicitly integrates the historical data and the conditional influence of these features to predict the susceptibility zones. Based on the model output, the study area was classified into four susceptibility zones: Very High (39 km<sup>2</sup>), High (45 km<sup>2</sup>), Moderate (36 km<sup>2</sup>), and Low (89 km<sup>2</sup>). The model was tested (30%) and trained (70%) using liquefaction evidence from multiple historical earthquakes. The area under the receiver operating characteristic curve (ROC-AUC) shows both training (0.929) and testing (0.871) values above 0.8, indicating the accuracy and robustness of the model. The liquefaction susceptibility map based on RF model can be integrated with the earthquake risk mitigation, and urban planning in tectonically active areas for achieving sustainable development.</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

Assessment of liquefaction susceptibility using machine learning technique in the inner wedge of Indo-Burmese range

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • Drishya Girishbai,
  • Arun Bhadran,
  • A. L. Achu,
  • Girish Gopinath,
  • B. P. Duarah

摘要

The synclinal valleys in the outer wedge of the Indo-Burmese Ranges (IBR), part of zipper tectonics, are more prone to earthquake-induced liquefaction due to thick unconsolidated, immature sediments with shallow groundwater conditions. The present study, a data-driven ensemble learning approach—Random Forest (RF)—was employed to assess liquefaction susceptibility in the Barak Valley by considering nine contributing factors such as geology, land use/land cover, lineament density, slope, peak ground acceleration, predominant frequency, peak amplification, shear wave velocity, and distance from water bodies. The RF model implicitly integrates the historical data and the conditional influence of these features to predict the susceptibility zones. Based on the model output, the study area was classified into four susceptibility zones: Very High (39 km2), High (45 km2), Moderate (36 km2), and Low (89 km2). The model was tested (30%) and trained (70%) using liquefaction evidence from multiple historical earthquakes. The area under the receiver operating characteristic curve (ROC-AUC) shows both training (0.929) and testing (0.871) values above 0.8, indicating the accuracy and robustness of the model. The liquefaction susceptibility map based on RF model can be integrated with the earthquake risk mitigation, and urban planning in tectonically active areas for achieving sustainable development.