2.4 | Habitat Classification
Habitat classifications were developed using Light Detection And Ranging (LiDAR) surveys and high-resolution aerial photo in 2009 by the Applied Geomatics Research Group (Nova Scotia Community College, Middleton, Nova Scotia; van Beest et al. 2014). Non-vegetated habitat classes included bare sand, ocean, human structures (buildings with fenced perimeters), and freshwater ponds. Vegetated habitat classes were characterized by their dominant plant species: grassland (marram grass), heath (mixed juniper, crowberry, and blueberry), sandwort, and beach pea. Vegetated classes subcategorized into ‘sparse’ or ‘dense’ (e.g. sparse grassland and dense grassland) in original classification efforts were combined in our analyses.
To quantify individual variation in local environment, we calculated the area of habitat classes overlapping a 150-m radius circular buffer centred on the location of sample collection in R (v3.5.1). A 150-m radius buffer corresponds approximately to the observed median daily movement of horses in 2014 (positive skewed distribution, median: 108 m/day; median 317 m/day), and so is expected to coarsely reflect the types of environment, and therefore forage, encountered during the 24 hours preceding defecation. Habitat class variables were calculated as the area of a given habitat class relative to the total occupiable terrestrial area included within an individual’s buffer (AreaBuffer – AreaBuilding– AreaOcean). Sandwort abundance was zero-inflated and non-normally distributed. Further, resource selection analysis of Sable Island horse foraging behaviour suggests horses actively select for sandwort when it is present, while other vegetated habitat classes are used in proportion to their abundance on the local landscape (personal communication K. Johnsen). For these statistical and biological reasons, sandwort was parameterized as ‘present’ or ‘absent’ in our analyses. Only vegetated habitat classes were parameterized in analyses to limit model inflation and avoid collinearity between terms.