Remote sensing-based soil organic carbon monitoring using advanced machine learning techniques under conservation agriculture systems

Abstract

Accurate soil organic carbon (SOC) monitoring is essential for sustainable agriculture and climate change mitigation. This study integrates remote sensing and machine learning to improve SOC estimation in agricultural soils across two contrasting sites: Niigata, Japan (temperate, sandy soils) and Agbelouve, Togo (tropical, clayey soils). Conservation agriculture (CA) practices, including no-tillage and mulching, were assessed for their role in carbon sequestration. Using freely available satellite data from Sentinel-1 synthetic aperture radar (SAR) and Sentinel-2 multispectral imagery (MSI), machine learning models were trained and validated. The eXtreme Gradient Boosting (XGBoost) model achieved the highest accuracy, with a cross-validation coefficient of determination (R²) of 0.88, a test R² of 0.91, and a root mean square error (RMSE) of 0.17 metric tons of carbon per hectare (t C ha⁻¹). Other models, including Random Forest (RF) and Support Vector Machine (SVM), showed competitive but slightly lower performance. Vegetation indices such as the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Soil-Adjusted Vegetation Index (SAVI) were identified as key predictors of SOC variation. SOC maps revealed significant spatial variability, ranging from 1.2 to 3.8 t C ha⁻¹ in Niigata and 0.9 to 3.2 t C ha⁻¹ in Togo, reflecting land use and climate differences. The results demonstrate the potential of integrating satellite-based observations with machine learning for cost-effective, high-resolution SOC assessment. This approach provides a scalable solution for site-specific land management, carbon market verification, and sustainable farming practices.

More by Nail Beisekenov

• Sustainable lake-resort tourism under water-quality constraints: an integrated monitoring-carrying-capacity framework for destination managementTourism Recreation Research · 2026
• Delft3D Model of Nature-Based Solutions: Comparing a Sand Engine, Living Shoreline, and Reef at Sand Point and Keller Bay, TexasJournal of Waterway Port Coastal and Ocean Engineering · 2026
• Data for "Detecting tipping thresholds in U.S. tidal wetlands using machine learning"Zenodo · 2026
• Recovery of Secondary Metals and Concrete Modification from Recycled PC Electronic WasteRecycling · 2026 · Citations: 1