Mapping winter soil cover in Belgian agricultural fields using Sentinel-2 imagery for carbon farming monitoring.

Carbon farming practices, such as cover crops, are crucial for increasing soil organic carbon (SOC) content and promoting sustainable agriculture. Cover crops, including yellow mustard and Japanese oats, play a key role in improving SOC by increasing biomass inputs during the non-growing season. Accurate identification and monitoring of winter soil cover in agricultural fields is essential to assess the effectiveness of these practices. 
This study uses Sentinel-2 satellite imagery to map winter soil cover in Belgian agricultural fields. Target cover types include bare soil, grassland, winter cereals, maize residues and various cover crops. Sentinel-2’s high temporal and spatial resolution, combined with its multispectral bands, enables the differentiation of soil cover types based on their spectral signatures. Feature engineering allows to derive covariates for supervised classification models from interpolated time-series of Sentinel-2 spectral bands and indices. Ground truth data collected during four consecutive winter seasons (2021-2024) were used to train and validate the random forest model.
This research demonstrates the potential of Sentinel-2 imagery for monitoring winter soil cover in agricultural fields, allowing to assess the implementation and effectiveness of carbon farming practices. Future work will focus on refining classification methods and on integrating winter cover management practices in SOC modelling to quantify their long-term benefits, contributing to more effective carbon sequestration strategies in croplands.