Detection and counting of meadow cuts by copernicus sentinel-2 imagery in the framework of the common agricultural policy (CAP)

Abstract

ABSTRACT Greening is a subsidy provided by the Common Agricultural Policy (CAP), related to mowing and designed to protect environment. National or regional paying agencies (PP) monitor and verify compliance of farmers’ declarations with CAP rules. In this work, an operational procedure is proposed aimed at supporting PPs in detecting, mapping and quantifying the number of times mowing occurred in a meadow field. In particular, 72,539 meadows fields within the Piemonte region (NW – Italy) were analysed with a time series of Sentinel-2 (S2) data. The procedure is based on the processing of filtered and regularized time series of NDVI maps. The Fast Fourier Transform (FFT) was applied at field level to decompose the local NDVI temporal profile. The frequency ( ω p e a k ) corresponding to the maximum amplitude ( A p e a k ) was therefore considered. A p e a k value was used to detect not-mowed meadows by thresholding based on the application of the non-parametric Kolmogorov-Smirnov test. Mowing counting were achieved with reference to ω p e a k and the correspondent map (called MCM) generated for the study area. MCM was, finally, tested against the validation set (285 fields). Results showed an Overall Accuracy (OA) > 87%, confirming the effectiveness of the proposed procedure in detecting, mapping and quantifying the number of times mowing occurred.