Gilad, D. (2019)

Remote sensing of rural landscape dynamics and its applications in movement ecology: The case of the Red Kite (Milvus milvus) in Switzerland.

Further information

Georg-August-Universität Göttingen





The expansion of agricultural areas, the intensification of land use and the increasing homogeneity of landscapes landscapes lead to biodiversity loss. One aspect of agricultural intensification is the timing and frequency of agricultural events, i.e. mowing, harvesting and ploughing. These events harm bird species by causing habitat destruction, increased mortality rates and decreased breeding success. But they also provide opportunities for other species, e.g. increasing the abundance and accessibility of prey. Quantifying rural landscape dynamics could offer new possibilities for large area analyses and provide insights on impacts and benefits that agricultural events create for species. We developed agricultural event models to identify and predict the occurrence of those events based on Sentinel-1 satellite data, which provide high spatial and temporal resolution interferometric images. Google Earth Engine, a cloud-based platform, was used to analyse the remotely sensed data. We trained our models with collected ground-truth data of agricultural events within 76 fields in western Switzerland and predicted the occurrence of mowing, harvesting and ploughing events within the entire study area. We were able to detect agricultural events, controlling for meteorological factors which influence Sentinel-1 data. The detection of mowing events had high accuracy during May (82%) and it decreased in June (71%). Mowing events in July had lower accuracy (64%) or were not detectable due to high heterogeneity in meadow structure. Harvesting and ploughing events were identified during May (86%), June (91%) and July (79%). Resulting agricultural management probability maps were used to assess habitat use based on movement trajectories of red kites (Milvus milvus), a diurnal European raptor species, which forages for small rodents in highly managed agricultural areas. The results of our habitat use model with low interval GPS data did not show that Red Kites select mowed, harvested or ploughed fields over fields without recent management events. Nevertheless, our models can be further developed to quantify land use intensity, detect landscape changes, monitor agricultural conservation practices and to investigate the impact of agricultural events on various species.