Abstract In June 2018, the European Commission approved a modification of the Common Agricultural Policy (CAP) that, among other measures, proposed the use of Copernicus data for the verification process of farmers’ declarations. In recent years, several research efforts have been conducted to develop operational tools to accomplish this aim, among this the Interreg-POCTEFA PyrenEOS project. This article describes the methodological strategy proposed in the PyrenEOS project, which is based on the identification of the most probable crop using the Random Forests algorithm. Originally, the strategy builds a training sample from the CAP declarations file based on their NDVI time series. In addition, a series of rules are proposed to establish the level of uncertainty in the classification, and the criteria used to represent each parcel in the verification map with a simple colour coding (traffic light), where green represents correctly declared parcels, red indicates that the declaration is dubious, and orange corresponds to parcels with a high classification uncertainty. This verification strategy has been applied to two Agricultural Regions of Navarre, during an agricultural campaign where valuable field inspections were available, with a sampling intensity of 7% of the declared parcels. The results obtained, report overall accuracies close to 80% when the most probable crop was considered, and 90% when the two most probable crops were considered. This proves it is possible to identify correctly declared parcels (green parcels) with an error below 1%. Orange and red parcels should be considered for further analysis and inspection by technicians from the paying agencies, though they represent a small percentage of declarations (~6% of parcels), and include most of the wrong declarations.
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|