Scholarlink

<p>Bangladesh lies downstream of one of the world’s highest rainfall zones, where monsoon precipitation drives strong seasonal and interannual variability of surface water in the northeastern Sylhet Division, particularly within the haor basins. Monitoring these dynamics is critical for water resource management, wetland conservation, flash-flood mitigation, and climate-related decision-making. This study develops a robust remote sensing framework to map surface water seasonality using Sentinel-1 Synthetic Aperture Radar (SAR) data and machine learning within the Google Earth Engine (GEE) platform. Four classification methods were evaluated: Otsu thresholding and three supervised algorithms—Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Tree (CART). The SVM classifier applied to the SAR Water Index (SWI) achieved the highest accuracy, showing strong agreement with field observations (R² = 0.9575). Monthly composites of Sentinel-1 imagery from 2014 to 2025 were used to generate detailed seasonality maps, categorizing water bodies as permanent, seasonal, or ephemeral. Results indicate that Sunamganj exhibits the most persistent water coverage due to its extensive haor wetlands, while Maulvibazar and Habiganj show more localized, short-duration inundation. Comparison with the Global Surface Water dataset of the Joint Research Centre highlights the advantages of SAR-based monitoring in cloud-dominated environments, effectively capturing transient and seasonally flooded areas often missed by optical sensors. Overall, the framework demonstrates the reliability of SAR-driven, cloud-independent water monitoring and provides valuable insights to support hydrological planning, wetland management, and climate-sensitive decision-making in monsoon-prone landscapes.</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

Assessing seasonal dynamics of surface water using multitemporal SAR data

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Mohammad Imrul Islam,
  • M. Mahmudur Rahman,
  • S. A. M. Arif-Ul-Haque,
  • Md. Abdur Rahman-Al-Mamun,
  • Md. Ashraful Haq Rahat,
  • Md. Naim Islam Talukder

摘要

Bangladesh lies downstream of one of the world’s highest rainfall zones, where monsoon precipitation drives strong seasonal and interannual variability of surface water in the northeastern Sylhet Division, particularly within the haor basins. Monitoring these dynamics is critical for water resource management, wetland conservation, flash-flood mitigation, and climate-related decision-making. This study develops a robust remote sensing framework to map surface water seasonality using Sentinel-1 Synthetic Aperture Radar (SAR) data and machine learning within the Google Earth Engine (GEE) platform. Four classification methods were evaluated: Otsu thresholding and three supervised algorithms—Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Tree (CART). The SVM classifier applied to the SAR Water Index (SWI) achieved the highest accuracy, showing strong agreement with field observations (R² = 0.9575). Monthly composites of Sentinel-1 imagery from 2014 to 2025 were used to generate detailed seasonality maps, categorizing water bodies as permanent, seasonal, or ephemeral. Results indicate that Sunamganj exhibits the most persistent water coverage due to its extensive haor wetlands, while Maulvibazar and Habiganj show more localized, short-duration inundation. Comparison with the Global Surface Water dataset of the Joint Research Centre highlights the advantages of SAR-based monitoring in cloud-dominated environments, effectively capturing transient and seasonally flooded areas often missed by optical sensors. Overall, the framework demonstrates the reliability of SAR-driven, cloud-independent water monitoring and provides valuable insights to support hydrological planning, wetland management, and climate-sensitive decision-making in monsoon-prone landscapes.