Scholarlink

<p>The increasing coastal population with dynamic coastal environmental conditions in India necessitates robust land flood risk assessment frameworks that promote sustainable development. The precise mapping and evaluation of land use hazards in susceptible coastal environments, where traditional approaches often fail to account for complex flood risk interactions. This study presents an integrated flood risk-based land use mapping approach that combines land use classification with multi-criteria flood risk assessment, utilizing machine learning (ML) algorithms to improve accuracy and efficiency. Focusing on the Baleswar CD block in India, comparing the performance of four machine learning algorithms—Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Decision Tree (DT) with traditional Weighted Overlay Analysis (WOA) for flood risk zonation. The study employs ASTER DEM and Sentinel-2 satellite data, alongside OpenStreetMap, to extract flood risk parameters. A sequentially analytical approach is utilized, combining flood intensity score, area of impact score, and flood hazard score to generate a comprehensive flood risk assessment. In the setting of vulnerability zoning, the calculation of factor weights is conducted using the Analytical Hierarchy Process (AHP). The ML-based flood risk maps categorize land cover into four categories-Agriculture, Vegetation, Barren Land, and Settlement, each assessed under high, medium, and low-flood risk levels. The Results highlight that barren land demonstrates a medium-flood risk coverage of 19.64% while high-flood risk exposure is minimal at 0.25%. Accuracy assessment via the Kappa index confirms outperformance of machine learning (ML) algorithms (SVM: 98.54%, KNN: 97.96%, RF: 98.45%, DT: 98.45%) over Weighted Overlay Analysis (91.74%). Among ML models, SVM demonstrates the highest precision, positioning them as the most effective method for integrated flood risk-based land use mapping. This finding offers essential insights for coastal land-use management and disaster resilience planning for vulnerable regions, as it illustrates the spatial distribution and proportional representation of a variety of land-use categories across flood risk zones.</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

Integrated flood risk-based land use mapping using machine learning and Geospatial technologies: a case study of Baleswar CD block, India

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月20日

查看全文
  • Auroshreeta Arpita,
  • Sk. Mustak,
  • Pritam Mitra,
  • Prosenjit Barman,
  • Pulak Barman,
  • Amol Kumar Garg,
  • Dharmaveer Singh

摘要

The increasing coastal population with dynamic coastal environmental conditions in India necessitates robust land flood risk assessment frameworks that promote sustainable development. The precise mapping and evaluation of land use hazards in susceptible coastal environments, where traditional approaches often fail to account for complex flood risk interactions. This study presents an integrated flood risk-based land use mapping approach that combines land use classification with multi-criteria flood risk assessment, utilizing machine learning (ML) algorithms to improve accuracy and efficiency. Focusing on the Baleswar CD block in India, comparing the performance of four machine learning algorithms—Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Decision Tree (DT) with traditional Weighted Overlay Analysis (WOA) for flood risk zonation. The study employs ASTER DEM and Sentinel-2 satellite data, alongside OpenStreetMap, to extract flood risk parameters. A sequentially analytical approach is utilized, combining flood intensity score, area of impact score, and flood hazard score to generate a comprehensive flood risk assessment. In the setting of vulnerability zoning, the calculation of factor weights is conducted using the Analytical Hierarchy Process (AHP). The ML-based flood risk maps categorize land cover into four categories-Agriculture, Vegetation, Barren Land, and Settlement, each assessed under high, medium, and low-flood risk levels. The Results highlight that barren land demonstrates a medium-flood risk coverage of 19.64% while high-flood risk exposure is minimal at 0.25%. Accuracy assessment via the Kappa index confirms outperformance of machine learning (ML) algorithms (SVM: 98.54%, KNN: 97.96%, RF: 98.45%, DT: 98.45%) over Weighted Overlay Analysis (91.74%). Among ML models, SVM demonstrates the highest precision, positioning them as the most effective method for integrated flood risk-based land use mapping. This finding offers essential insights for coastal land-use management and disaster resilience planning for vulnerable regions, as it illustrates the spatial distribution and proportional representation of a variety of land-use categories across flood risk zones.