Verlag der Österreichischen Akademie der Wissenschaften

Abstract:
For the conservation of biodiversity in general and the monitoring of meadow birds in particular, actual grassland-use intensity maps are highly desirable. A method to map and assess grassland management intensity was developed using C5.0 decision tree classification on Sentinel-2 satellite data. Monoculture and extensively managed grasslands on both peat and clay soils could be accurately detected at parcel level in Friesland, the Netherlands. Field-survey-based validation returned an overall classification accuracy of 84.3% (KHAT 0.65). The Sentinel-2 Red-Edge Position vegetation index was found to be a good indicator of fertilization. Availability of springtime imagery, preferably acquired in April before the first mowing date, is essential for accurate classification. The spectral responses of grassland types on peat and clay soils differ significantly. Hence, successful classification requires training data for both soil types. The resulting grassland management map was used to assess the distribution of meadow bird nests. Redshank (79%) and godwit (77%) in particular choose to breed on extensive parcels. With the increasing availability of satellite imagery, remote sensing techniques can be used to monitor agri-environmental measures (at parcel and landscape scale) that impact the conservation of grassland biodiversity.