The Role of Field Survey in the Identification of Farmland Abandonment in Slovakia Using Sentinel-2 Data

摘要

Agricultural land abandonment is a dynamic process characterized by both significant spectralvariability and spectral similarity to areas of agricultural land. The identification of abandonedagricultural land (AAL) based on remote sensing data must be preceded by fieldsurveys that are focused on the acquisition of the physiognomic characteristics (the compositionof species, height of vegetation, textures, and clustering into patterns) of theseareas. This study aims to document the physiognomic and spectral differences betweenAAL and other land cover/land use classes in Slovakia. The Normalized DifferenceVegetation Index (NDVI), derived from Sentinel-2 time series for vegetation from April toSeptember 2018, was applied. NDVI values were calculated for each Sentinel-2 scene, andNDVI profiles for selected samples were used to create phenological profiles for each AALand land cover/land use class. The dispersion of the NDVI values for these classes, theirmedian, and the root mean square error between NDVI data show that overgrowth by herbaceousplants is characterized by more significant dynamics (0.40–0.75), resulting in betterspectral discriminability than classes overgrown by shrubs and trees (0.70–0.80). Field surveydata are a fundamental prerequisite for the correct explanation of the discriminability ofthese AAL classes.