Scholarlink

<p>Soil Organic Carbon (SOC) is critical for soil functions and ecosystem services. This study evaluates machine learning (ML) models and environmental predictors to map SOC stocks in hilly regions, highlighting their spatial variability. Random Forest (RF) models outperformed Cubist and Artificial Neural Networks, with R² values of 0.41 (RMSE is 8.14) for surface soils and 0.36 (RMSE is 8.56) for subsurface soils. Surface SOC stocks ranged from 14.83 to 47.27 Mg C ha⁻¹, while subsurface stocks ranged from 16.91 to 45.86 Mg C ha⁻¹, reflecting higher organic inputs and root biomass in surface layers. Key predictors included geology and terrain features for surface soils, and rainfall and ruggedness for subsurface layers. These insights aid sustainable land 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

Spatial Prediction of Soil Organic Carbon Stock Using Machine Learning in a Himalayan Watershed

  • 期刊论文

  • 发表日期：

    2026年6月18日

查看全文
  • Nilkamal Konra,
  • Justin George Kalambukattu,
  • Suresh Kumar

摘要

Soil Organic Carbon (SOC) is critical for soil functions and ecosystem services. This study evaluates machine learning (ML) models and environmental predictors to map SOC stocks in hilly regions, highlighting their spatial variability. Random Forest (RF) models outperformed Cubist and Artificial Neural Networks, with R² values of 0.41 (RMSE is 8.14) for surface soils and 0.36 (RMSE is 8.56) for subsurface soils. Surface SOC stocks ranged from 14.83 to 47.27 Mg C ha⁻¹, while subsurface stocks ranged from 16.91 to 45.86 Mg C ha⁻¹, reflecting higher organic inputs and root biomass in surface layers. Key predictors included geology and terrain features for surface soils, and rainfall and ruggedness for subsurface layers. These insights aid sustainable land management.