Advances in Agriculture and Agricultural Sciences

ISSN 2756-326X

Advances in Agriculture and Agricultural Sciences | Vol. 12, No. 11, November 2026 | pp. 1061–1069

DOI: 10.46882/2026.AAAS.121131

Research Article

Title: Hyperspectral Soil Organic Carbon Mapping Using Sentinel-2 Time-Series Composites and Random Forest Kriging Across Semi-Arid Croplands

Names of Authors: Ben-Salem, H.¹, MacDonald, G. W.², and Rojas-Paredes, L.³

Authors’ Affiliations: ¹Department of Soil and Environmental Sciences, National Agronomic Institute of Tunisia, Tunis, Tunisia. ²Department of Soil Science, University of Saskatchewan, Saskatoon, Canada. ³Department of Earth and Environmental Sciences, National Agrarian University La Molina, Lima, Peru.

Abstract: Accurate regional mapping of topsoil organic carbon (SOC) is essential for verifying carbon-farming incentives and guiding site-specific soil restoration in semi-arid croplands, yet dense ground sampling is prohibitively expensive. Bare-soil median reflectance composites derived from five-year Sentinel-2 multispectral time series were combined with terrain wetness indices and 680 georeferenced soil cores (0–30 cm) across 180,000 hectares of rainfed cereal farmland. Regression kriging paired with random forest machine learning was benchmarked against ordinary kriging and support vector regression. Integrating short-wave infrared (SWIR1/SWIR2) bare-soil indices with topographic wetness covariates explained 81.4% of spatial SOC variance (RMSE = 1.42 g kg⁻¹) during independent spatial cross-validation. The resulting 10-meter resolution SOC baseline map identified degraded erosion-prone hillslopes requiring targeted organic matter inputs. Multi-temporal satellite composites enable scalable, cost-effective digital soil carbon monitoring.

Keywords: Digital soil mapping; Soil organic carbon; Sentinel-2 composites; Random forest kriging; Semi-arid croplands; Carbon monitoring.

Manuscript Timeline: Received October 20, 2026; Revised November 19, 2026; Accepted November 27, 2026; Published November 30, 2026.

Citation: Ben-Salem, H., MacDonald, G. W., and Rojas-Paredes, L. (2026). Hyperspectral Soil Organic Carbon Mapping Using Sentinel-2 Time-Series Composites and Random Forest Kriging Across Semi-Arid Croplands. Advances in Agriculture and Agricultural Sciences, 12(11), pp. 1061–1069.