Scholarlink

<p>Floods are a critical natural catastrophe that has a significant impact on the Teesta River Basin in India, resulting in substantial losses to agriculture, property, and life. This investigation introduces a sophisticated geospatial artificial intelligence framework for the evaluation of flood susceptibility, which employs spatial data augmentation for enhancing models’ reliability. The study uses six machine learning algorithms, involving K-Nearest Neighbours (KNN), Na¨ıve Bayes (NB), Decision Tree (DT), Conditional Inference Tree (CIT), BlackBoost (BBoost), and Model Averaged Neural Network (MANN), to integrate multisource geospatial data, including 30 × 30&#xa0;m Cartosat-1 satellite derived Digital Elevation Model (DEM). A thorough flood inventory found 317 inundated areas, with original data sources systematically enhanced at a 1:3 ratio to expand the dataset, and the Boruta feature selection technique identified 12 essential flood conditioning elements. The models were assessed using metrics such as accuracy, Kappa, precision, recall, F1 score, and Area Under the Curve (AUC). The KNN model has the maximum accuracy of 91.62% (Kappa = 0.8325), precision of 90.41%, recall of 93.12%, and AUC of 97.80%, classifying around 35% of the basin as high-susceptibility zones. This study underscores the efficacy of machine learning and geospatial analysis in enhancing flood risk management tactics, offering critical insights for disaster management policymakers. Future efforts will concentrate on improving model robustness via comprehensive evaluation across many geographical settings, incorporating real-time meteorological data for dynamic flood forecasting, and creating user-friendly decision support tools for stakehol-ders in disaster management.</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

Flood Susceptibility Assessment in the Teesta River Basin, India: An Advanced Geospatial Artificial Intelligence Approach Leveraging Spatial Data Augmentation

  • 期刊论文

  • 发表日期：

    2025年4月9日

查看全文
  • Deepanjan Sen,
  • Swarup Das

摘要

Floods are a critical natural catastrophe that has a significant impact on the Teesta River Basin in India, resulting in substantial losses to agriculture, property, and life. This investigation introduces a sophisticated geospatial artificial intelligence framework for the evaluation of flood susceptibility, which employs spatial data augmentation for enhancing models’ reliability. The study uses six machine learning algorithms, involving K-Nearest Neighbours (KNN), Na¨ıve Bayes (NB), Decision Tree (DT), Conditional Inference Tree (CIT), BlackBoost (BBoost), and Model Averaged Neural Network (MANN), to integrate multisource geospatial data, including 30 × 30 m Cartosat-1 satellite derived Digital Elevation Model (DEM). A thorough flood inventory found 317 inundated areas, with original data sources systematically enhanced at a 1:3 ratio to expand the dataset, and the Boruta feature selection technique identified 12 essential flood conditioning elements. The models were assessed using metrics such as accuracy, Kappa, precision, recall, F1 score, and Area Under the Curve (AUC). The KNN model has the maximum accuracy of 91.62% (Kappa = 0.8325), precision of 90.41%, recall of 93.12%, and AUC of 97.80%, classifying around 35% of the basin as high-susceptibility zones. This study underscores the efficacy of machine learning and geospatial analysis in enhancing flood risk management tactics, offering critical insights for disaster management policymakers. Future efforts will concentrate on improving model robustness via comprehensive evaluation across many geographical settings, incorporating real-time meteorological data for dynamic flood forecasting, and creating user-friendly decision support tools for stakehol-ders in disaster management.