Scholarlink

<p>Landslide susceptibility assessment in basaltic terrain of the Deccan Volcanic Province, India, remains challenging due to complex geomorphic controls and limited evaluation of integrated modelling approaches. This study aims to compare commonly used statistical and heuristic models and to examine whether their integration improves susceptibility prediction. Bivariate statistical models of Modified Frequency Ratio (MFR), Statistical Index, Weight of Evidence and the heuristic Analytic Hierarchy Process (AHP) models were applied using six conditioning factors: slope inclination, slope aspect, slope curvature, geomorphology, land use and land cover and slope forming material. A landslide inventory comprising 99 documented landslides was used for model development and validation. Landslide Susceptibility Index and Landslide Occurrence Favourability Scores were calculated to generate four susceptibility maps, which were evaluated using Receiver Operating Characteristic analysis. Binary classification using mean, median and Natural Breaks (Jenks) thresholds enabled accuracy assessment through contingency tables, including Type I and Type II errors. Among individual models, MFR showed the best performance and was therefore integrated with AHP. The integrated model achieved an Area Under the Curve of 0.827, representing improvements of 4.9% and 13.5% over the MFR and AHP models, respectively, with reduced misclassification errors. The results demonstrate the effectiveness of hybrid modelling for landslide susceptibility mapping and provide a robust tool for regional-scale hazard assessment and land-use planning in volcanic terrains.</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

Improving Landslide Susceptibility Prediction Through Hybrid Framework: A Case from the Deccan Province

  • 期刊论文

  • 发表日期：

    2026年2月23日

查看全文
  • Parmita Dasarwar,
  • Jagadish Nandan Hindayar,
  • Prakash Khushaldas Gajbhiye

摘要

Landslide susceptibility assessment in basaltic terrain of the Deccan Volcanic Province, India, remains challenging due to complex geomorphic controls and limited evaluation of integrated modelling approaches. This study aims to compare commonly used statistical and heuristic models and to examine whether their integration improves susceptibility prediction. Bivariate statistical models of Modified Frequency Ratio (MFR), Statistical Index, Weight of Evidence and the heuristic Analytic Hierarchy Process (AHP) models were applied using six conditioning factors: slope inclination, slope aspect, slope curvature, geomorphology, land use and land cover and slope forming material. A landslide inventory comprising 99 documented landslides was used for model development and validation. Landslide Susceptibility Index and Landslide Occurrence Favourability Scores were calculated to generate four susceptibility maps, which were evaluated using Receiver Operating Characteristic analysis. Binary classification using mean, median and Natural Breaks (Jenks) thresholds enabled accuracy assessment through contingency tables, including Type I and Type II errors. Among individual models, MFR showed the best performance and was therefore integrated with AHP. The integrated model achieved an Area Under the Curve of 0.827, representing improvements of 4.9% and 13.5% over the MFR and AHP models, respectively, with reduced misclassification errors. The results demonstrate the effectiveness of hybrid modelling for landslide susceptibility mapping and provide a robust tool for regional-scale hazard assessment and land-use planning in volcanic terrains.