Scholarlink

<p>Mangrove forests are among the most efficient coastal carbon sinks, playing a vital role in climate change mitigation and blue carbon accounting. Accurate estimation of aboveground biomass (AGB) is therefore essential, particularly in complex and inaccessible ecosystems such as the Indian Sundarbans. However, comprehensive large-scale AGB mapping using advanced deep learning techniques remains limited for this region. This study evaluates the effectiveness of artificial neural networks (ANN) and convolutional neural networks (CNN) for estimating mangrove AGB using Landsat-9 derived vegetation indices transformed through principal component analysis. Field-measured biomass data from representative plots were used for model training and validation. Model performance was assessed using the coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), mean absolute deviation (MAD), and mean bias error (MBE). Results show that the CNN model outperformed the ANN model, achieving higher predictive accuracy (R<sup>2</sup> = 0.88; RMSE = 46.89 t/ha; MAD = 38.36 t/ha; MBE = 4.02 t/ha), while ANN yielded lower performance (R<sup>2</sup> = 0.76; RMSE = 51.10 t/ha; MAD = 41.17 t/ha; MBE = 7.23 t/ha). Overall, this study demonstrates that integrating deep learning with freely available Landsat imagery provides a robust and cost-effective approach for large-scale mangrove biomass estimation, supporting improved spatial carbon stock assessment and mangrove conservation planning.</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

Deep neural network-based approaches for aboveground biomass mapping in Indian Sundarbans

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Sinchan Banerjee,
  • Abhisek Santra,
  • Tuhin Bhadra,
  • Kasturi Mukherjee,
  • Abinit Saha,
  • Roshan Das,
  • Rupayan Sardar,
  • Soumak Roy,
  • Shreyashi S. Mitra

摘要

Mangrove forests are among the most efficient coastal carbon sinks, playing a vital role in climate change mitigation and blue carbon accounting. Accurate estimation of aboveground biomass (AGB) is therefore essential, particularly in complex and inaccessible ecosystems such as the Indian Sundarbans. However, comprehensive large-scale AGB mapping using advanced deep learning techniques remains limited for this region. This study evaluates the effectiveness of artificial neural networks (ANN) and convolutional neural networks (CNN) for estimating mangrove AGB using Landsat-9 derived vegetation indices transformed through principal component analysis. Field-measured biomass data from representative plots were used for model training and validation. Model performance was assessed using the coefficient of determination (R2), root mean square error (RMSE), mean absolute deviation (MAD), and mean bias error (MBE). Results show that the CNN model outperformed the ANN model, achieving higher predictive accuracy (R2 = 0.88; RMSE = 46.89 t/ha; MAD = 38.36 t/ha; MBE = 4.02 t/ha), while ANN yielded lower performance (R2 = 0.76; RMSE = 51.10 t/ha; MAD = 41.17 t/ha; MBE = 7.23 t/ha). Overall, this study demonstrates that integrating deep learning with freely available Landsat imagery provides a robust and cost-effective approach for large-scale mangrove biomass estimation, supporting improved spatial carbon stock assessment and mangrove conservation planning.