High spatial resolution vegetation mapping for assessment of wildlife habitat

Abstract

Management of large herbivore populations requires assessment of both the quality and quantity of habitats available to the animals. Classifying and quantifying their habitats, particularly the availability of forage, is often challenging because of the time and expense associated with collection of the appropriate data over large landscapes. Land-cover maps generated from remote-sensing imagery generally provide only qualitative information regarding species composition of the vegetation, and do not adequately quantify forage species availability. Our objective was to determine the feasibility of mapping important forage species for moose (Alces americanus) in South-central Alaska, USA, using fine-resolution aerial photography. We used infrared and color digital photography acquired at an altitude of 150 m across 2 study areas with distinctly different vegetation to provide spectral data for supervised classifications of plant species using the maximum likelihood algorithm. We analyzed spring, autumn, and composite sets of the imagery to determine which imagery provides the most accurate classifications. Willow (Salix spp.) was separable from non-forage species in autumn images with an average overall accuracy of 76% in the Nelchina study area and 78% in Placer Valley. The classifications of composite images separated important moose forages, but with lower average overall accuracies (68% for Nelchina study area and 75% for Placer Valley) compared with the autumn classifications. Spring classifications were least accurate (57% for Nelchina study area and 63% for Placer Valley), likely because the green-up of plants in the spring occurred rapidly for all species in both study sites, and this resulted in relatively small differences in spectral response patterns between species. © 2013 The Wildlife Society. © The Wildlife Society, 2013