Scholarlink

Monitoring water bodies is important for sustainable water resource management, particularly in regions like India, where rapid urbanization and climate change have led to significant environmental challenges. This paper presents a novel approach for analyzing and predicting temporal changes in water bodies across India using multi-year Land Use and Land Cover (LULC) data from the Sentinel-2 Image Collection obtained from Google Earth Engine (GEE), which provides multispectral imagery at high spatial resolution (10, 20, and 60 m). We leverage the Normalized Difference Water Index (NDWI) to identify water bodies from satellite imagery collected between 2018 and 2023. To analyze trends and predict future changes, we propose a hybrid deep learning model combining Convolutional Neural Networks (CNN) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal trend analysis. The proposed method identifies water bodies, evaluates year-over-year changes, and predicts potential future shifts. Additionally, the system generates alerts for local authorities when significant changes are detected, enabling timely intervention. Comprehensive preprocessing techniques, including cloud masking and NDWI calculation, are applied to ensure data accuracy. The dataset used for this research was obtained from the Sentinel-2 Image Collection, which covers the entire territory of India. Experimental results demonstrate a significant improvement in accuracy, with the proposed model achieving 92.5% precision and 90.3% recall, outperforming traditional methods such as threshold-based approaches, Random Forest, and SVM. This research offers a new, data-driven approach to monitoring water resources, contributing to better environmental management and decision-making.

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

Hybrid CNN-LSTM Model for Water Body Monitoring and Prediction Using Sentinel-2 Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ramamoorthy Hariharan,
  • M. Dhilsath Fathima,
  • S. Kavitha,
  • N. Vinitha,
  • C. Madan Kumar

摘要

Monitoring water bodies is important for sustainable water resource management, particularly in regions like India, where rapid urbanization and climate change have led to significant environmental challenges. This paper presents a novel approach for analyzing and predicting temporal changes in water bodies across India using multi-year Land Use and Land Cover (LULC) data from the Sentinel-2 Image Collection obtained from Google Earth Engine (GEE), which provides multispectral imagery at high spatial resolution (10, 20, and 60 m). We leverage the Normalized Difference Water Index (NDWI) to identify water bodies from satellite imagery collected between 2018 and 2023. To analyze trends and predict future changes, we propose a hybrid deep learning model combining Convolutional Neural Networks (CNN) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal trend analysis. The proposed method identifies water bodies, evaluates year-over-year changes, and predicts potential future shifts. Additionally, the system generates alerts for local authorities when significant changes are detected, enabling timely intervention. Comprehensive preprocessing techniques, including cloud masking and NDWI calculation, are applied to ensure data accuracy. The dataset used for this research was obtained from the Sentinel-2 Image Collection, which covers the entire territory of India. Experimental results demonstrate a significant improvement in accuracy, with the proposed model achieving 92.5% precision and 90.3% recall, outperforming traditional methods such as threshold-based approaches, Random Forest, and SVM. This research offers a new, data-driven approach to monitoring water resources, contributing to better environmental management and decision-making.