Scholarlink

<p>Accurate measurement and mapping of above-ground biomass (AGB) is critical for SDG 15, facilitating comprehensive ecosystem assessments and carbon sequestration evaluations within biodiversity hotspots like the Indo-Myanmar region. This study employs Random Forest (RF), Classification and Regression Trees (CART), and Gradient Tree Boosting (GTB) models in Google Earth Engine (GEE) to predict AGB in Manipur, India. Using remote sensing data (NDVI, EVI, LAI, FPAR, land cover, and DEM), models were trained on 2010 reference data for 2024 predictions. The Random Forest model achieved an R² of 0.797, outperforming CART and GTB in predictive accuracy (RMSE: 13.678 Mg/ha). NDVI emerged as the key predictor. Prediction maps reveal 24–75% biomass decline, highlighting degradation in this biodiversity hotspot. This approach supports SDG 15 and REDD + by enabling scalable forest monitoring. The findings also underscore the urgent need for improved forest protection, restoration strategies, and long-term monitoring frameworks. This integrated modelling approach offers a scalable tool for supporting evidence-based policies related to forest conservation, carbon accounting, and sustainable landscape management.</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

Machine Learning-Based Above-Ground Biomass Prediction in Indo-Myanmar Biodiversity Hotspot of India, Using Google Earth Engine

  • 期刊论文

  • 发表日期：

    2026年5月19日

查看全文
  • Kangujam Premkumar,
  • Vivek Vaishnav,
  • Sanjay Singh,
  • Gopal Shukla

摘要

Accurate measurement and mapping of above-ground biomass (AGB) is critical for SDG 15, facilitating comprehensive ecosystem assessments and carbon sequestration evaluations within biodiversity hotspots like the Indo-Myanmar region. This study employs Random Forest (RF), Classification and Regression Trees (CART), and Gradient Tree Boosting (GTB) models in Google Earth Engine (GEE) to predict AGB in Manipur, India. Using remote sensing data (NDVI, EVI, LAI, FPAR, land cover, and DEM), models were trained on 2010 reference data for 2024 predictions. The Random Forest model achieved an R² of 0.797, outperforming CART and GTB in predictive accuracy (RMSE: 13.678 Mg/ha). NDVI emerged as the key predictor. Prediction maps reveal 24–75% biomass decline, highlighting degradation in this biodiversity hotspot. This approach supports SDG 15 and REDD + by enabling scalable forest monitoring. The findings also underscore the urgent need for improved forest protection, restoration strategies, and long-term monitoring frameworks. This integrated modelling approach offers a scalable tool for supporting evidence-based policies related to forest conservation, carbon accounting, and sustainable landscape management.