Résumé
Among the wide variety of existing agricultural practices, some amendments can lead to changes in soil properties, including those perceived as permanent, such as clay content. These modifications pose new challenges for mapping and monitoring soil characteristics. In South India, one such practice is the application of sediment from reservoirs onto agricultural fields. Our study investigates the ability of Sentinel-2 imagery to detect changes in topsoil clay content induced by this practice. We used data from a cultivated watershed in South India, including 164 topsoil samples, and a set of control fields (19 without and 43 with sediment application). Two Sentinel-2 images acquired in February and April 2017 (before and after sediment spreading) were used to estimate soil clay content via Multiple Linear Regression (MLR) models. Comparing the clay content maps from both dates revealed significant increases in areas where sediment was applied. Validation with control fields demonstrated that Sentinel-2 images successfully captured clay content changes attributable to sediment application. When extended to the entire cultivated area, our approach identified 242 fields exhibiting a significant clay increase in 2017. These results highlight the potential of multi-temporal Sentinel-2 data to monitor short-term changes in soil properties driven by agricultural practices. Future work could leverage the full Sentinel-2 archive (from 2017 onwards) to quantify long-term dynamics of sediment application across broader regions. More generally, this approach contributes to the development of Earth observation-based frameworks for tracking soil property changes associated with evolving agricultural systems at multiple spatial and temporal scales.