Soil tillage detection performances using Sentinel time series and Copernicus product in Wallonia

Details

Supervisors

Defourny, Pierre ; Kenda, Tom

Faculty

Faculté des bioingénieurs

Degree label

Master [120] en sciences agronomiques et industries du vivant, à finalité spécialisée: Geo-Information Science and Earth Observation for Environmental Modelling and Management

Abstract

enMonitoring soil disturbance activities such as tillage is essential for promoting sustainable agricultural practices and supporting soil health. However, detecting these short-lived events through remote sensing presents significant challenges due to their subtle surface signatures and the influence of external factors like precipitation and vegetation cover. This study investigates the potential of Sentinel-1 C-band synthetic aperture radar (SAR) data—specifically backscatter and interferometric coherence metrics—for tillage detection in Wallonia, Belgium. Two rule-based algorithms are developed: a baseline version that filters out soil moisture using a landscape-based approach, and a modified version that replaces this with precipitation data. Both algorithms use VV and VH polarized backscatter and coherence signals and apply a two-level detection framework to distinguish high- and low-confidence tillage events, based on the joint behaviour of radar metrics. The modified version achieves higher overall accuracy, suggesting that integrating precipitation data can enhance detection performance. The results demonstrate that the combined use of backscatter and coherence improves the sensitivity of detection. The analysis further indicates that satellite pass direction (ascending vs. descending) significantly influences detection performance, with ascending passes yielding more convincing results—likely due to the reduced impact of morning dew. Importantly, the absence of precise tillage dates and pre-tillage field conditions in the ground truth dataset limit both detection accuracy and the ability to validate results with high temporal confidence. Despite these limitations, the study confirms the viability of SAR time series for tillage monitoring and contribute to ongoing efforts in agricultural monitoring under the Common Agricultural Policy and support the development of more sustainable, accountable land management practices.