Tillage refers to the agricultural practice involving the preparation of soil through mechanical operations such as digging, stirring and overturning. In the context of the Common Agricultural Policy (CAP), tillage practices have been gaining significant interest for their impact on environment and climate. Reduced or no tillage, which imply a low level of soil disturbance, may have significant positive effects in terms of soil health, carbon sequestration, provision of ecosystem services and soil erosion. Given these positive effects, Member States (MSs) have included in their CAP Strategic Plans (CSPs), interventions involving reduced or no tillage. With the 2023-2027 CAP, MSs are responsible for the development of effective monitoring approaches able to detect agricultural practices and verify the eligibility conditions (requirements) established in the MS CSPs. The ensemble of such monitoring approaches forms the so-called Area Monitoring System (AMS), which essentially relies on different technologies and products including those from the freely available Sentinel-1 and Sentinel-2 missions and in-situ data such as geo-tagged pictures. In this context, also tillage detection may be subject to monitoring and the discrimination between tillage and no-tillage using satellite products, such as those from Sentinel-1 and Sentinel-2, is thus of significant importance for MS administrations. This report summarizes the work done by the Joint Research Centre (JRC) toward the detection of tillage conditions using Sentinel-1 and Sentinel-2 products. The analysis is based on reference data received from AGREENA (https://agreena.com/), a company devoted to the development of Monitoring Reporting and Verification (MRV) technologies for regenerative agriculture. The analysis involved the dataset preparation including the retrieval of the parcel geometries from open Geo-Spatial Application (GSA) databases, the extraction of Sentinel-1 and Sentinel-2 products, the selection of appropriate features for classification and the performance analysis of different classification approaches. The analysis shows that Sentinel-2 derived products play a major role for the discrimination of tillage/no-tillage. More specifically, the Normalized Difference Vegetation Index (NDVI) provides the most informative features for the classification process. The major difficulty encountered in this study was represented by the fact that, in the reference datasets, the tillage and no-tillage classes were significantly unbalanced. Most of the parcels were interested by conventional tillage with a significant level of soil disturbance. Approaches for rebalancing the two classes, including the generation of synthetic features, were considered. The importance of Sentinel-1 products, in the absence of optical information, for instance due to clouds, also emerged. The analysis provides useful insights for the design of tillage/no-tillage detection approaches, identifying the most relevant features and data sources.
BORIO Daniele;
PIGNOCCHINO Gianmarco;
ERDOGAN Hakki;
LOUDJANI Philippe;
2025-12-05
Publications Office of the European Union
JRC143611
978-92-68-33757-8 (online),
1831-9424 (online),
EUR 40532,
OP KJ-01-25-571-EN-N (online),
https://publications.jrc.ec.europa.eu/repository/handle/JRC143611,
10.2760/7914269 (online),
