Agricultural landscapes support multiple functions and
are of great importance for biodiversity. Heterogeneous agricultural mosaics
of cropland and grassland commonly result from variable land use practices
and ecosystem service demands. Switzerland's agricultural land use is
considerably spatially heterogeneous due to strong variability in
conditions, especially topography and climate, thus presenting challenges to
automated agricultural mapping. Nationwide knowledge of the location of
cropland and grassland is necessary for effective conservation and land use
planning. We mapped the distribution of cropland and permanent grassland
across Switzerland. We used several indices largely derived from Sentinel-2
satellite imagery captured over multiple growing seasons and parcel-based
training data derived from landholder reporting. The mapping was conducted
within Google Earth Engine using a random forest classifier. The
resulting map has high accuracy in lowlands as well as in mountainous areas.
The map will act as a base agricultural land cover dataset for researchers
and practitioners working in agricultural areas of Switzerland and
interested in land cover and landscape structure. The map as well as the
training data and calculation algorithms (using Google Earth
Engine) are freely available for download on the EnviDat platform https://doi.org/10.16904/envidat.205 (Pazúr et al., 2021)....

A national extent map of cropland and grassland for Switzerland based on Sentinel-2 data

10.5194/essd-14-295-2022