Scholarlink

<p>Flood events cause significant loss of life and property, thus making early detection and reliable flood monitoring critical for mitigating their impacts. However, studies assessing floods in high mountains are limited due to the complex terrain, steep slopes and persistent cloud cover. In this study, we compared the usability of Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 multispectral imagery (MSI) and their fusion (SAR + MSI) for flood detection in the Melamchi River of Central Nepal employing Random Forest (RF), a supervised machine learning algorithm. Freely accessible Sentinel-1 and Sentinel-2 data were processed and analyzed within the Google Earth Engine platform to ensure methodological reproducibility. Two complementary validations were performed, addressing the lack of ground truth data that often lack during disastrous situations like flood: (a) a quantitative validation using a high-resolution PlanetScope (PS) imagery, and (b) qualitative assessment using Unmanned Aerial Vehicle (UAV) imageries. The RF classifier detected floods with an overall accuracy of 77% for Sentinel-1 SAR, 84% for Sentinel-2 MSI, and 88% for fused (SAR + MSI) dataset. Additionally, the fusion method achieved the highest F1 score of 0.77 and detected 329.73&#xa0;ha of flooding, outperforming SAR and MSI individually. These findings highlight the usefulness of integrating optical and SAR imagery to overcome weather-related and complex terrain-associated challenges, especially in regions characterized by dynamic weather conditions and frequent cloud cover, such as the Himalayas. Utilizing freely accessible data from remote sensing technology, we can enhance disaster management and response efforts, ultimately reducing the impact of floods on disaster-affected vulnerable communities.</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 Flood Detection in the Himalayas: Evaluating SAR-Optical Data Fusion with Machine Learning in a Mountainous River Basin of Nepal

  • 期刊论文

  • 发表日期：

    2026年7月17日

查看全文
  • Sandip Rijal,
  • Sonia Sharma Banjade,
  • Nitant Rai,
  • Anil K. Mandal,
  • Rabindra Parajuli

摘要

Flood events cause significant loss of life and property, thus making early detection and reliable flood monitoring critical for mitigating their impacts. However, studies assessing floods in high mountains are limited due to the complex terrain, steep slopes and persistent cloud cover. In this study, we compared the usability of Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 multispectral imagery (MSI) and their fusion (SAR + MSI) for flood detection in the Melamchi River of Central Nepal employing Random Forest (RF), a supervised machine learning algorithm. Freely accessible Sentinel-1 and Sentinel-2 data were processed and analyzed within the Google Earth Engine platform to ensure methodological reproducibility. Two complementary validations were performed, addressing the lack of ground truth data that often lack during disastrous situations like flood: (a) a quantitative validation using a high-resolution PlanetScope (PS) imagery, and (b) qualitative assessment using Unmanned Aerial Vehicle (UAV) imageries. The RF classifier detected floods with an overall accuracy of 77% for Sentinel-1 SAR, 84% for Sentinel-2 MSI, and 88% for fused (SAR + MSI) dataset. Additionally, the fusion method achieved the highest F1 score of 0.77 and detected 329.73 ha of flooding, outperforming SAR and MSI individually. These findings highlight the usefulness of integrating optical and SAR imagery to overcome weather-related and complex terrain-associated challenges, especially in regions characterized by dynamic weather conditions and frequent cloud cover, such as the Himalayas. Utilizing freely accessible data from remote sensing technology, we can enhance disaster management and response efforts, ultimately reducing the impact of floods on disaster-affected vulnerable communities.