GPS, LiDAR and VNIR data to monitor the spatial behavior of grazing sheep

ABSTRACT

Traditional knowledge about the behavior of grazing livestock is about to disappear. Shepherds well know that sheep behavior follows non-random patterns. As a novel alternative to seeking behavioral patterns, this study quantified the grazing activities of two sheep flocks of Churra breed (both in the same area but separated by 10 years) based on Global Position System (GPS) monitoring and remote monitoring sensing techniques. In the first monitoring period (2009-10), geolocations were recorded every 5 min (4,240 records), while in the second one (2018-20), records were taken every 30 min (7,636 records). The data were clustered based on the day/night and the activity (resting, moving, or grazing). An airborne LiDAR dataset was used to study the slope, aspect, and vegetation height. Four visible-infrared orthophotographs were mosaicked and classified to obtain the land use/land cover (LU/LC) map. Then, GPS locations were overlain on the terrain features, and a Chi-square test evaluated the relationships between locations and terrain features. Three spatial statistics (directional distribution, Kernel density, and Hot Spot analysis) were also calculated. Results in both monitoring periods suggested that the spatial distribution of free-grazing ewes was non-random. The flocks showed strong preferences for grazing areas with gentle north-facing slopes, where the herbaceous layer formed by pasture predominates. The geostatistical analyses of the sheep locations corroborated those preferences. Geotechnologies have emerged as a potent tool to demonstrate the influence of environmental and terrain attributes on the non-random spatial behavior of grazing sheep.