Detecting continuous lichen abundance for mapping winter caribou forage at landscape spatial scales

Spatial variation of available food resources can be difficult to accurately quantify for wide ranging organisms at landscape scales. Lichens with usnic acid, a yellowish pigment, constitute a large portion of caribou winter diet across much of their range. We take a new approach of modeling lichen abundances by capitalizing on unique spectral characteristics of usnic acid lichens. We utilize a recently completed ground reference vegetation data set extending over 12,000km2 in Denali National Park and Preserve, Alaska to model the abundance of usnic lichen and other forage vegetation groups. Spectral signatures were obtained for more than 700 vegetation monitoring plots in Denali from Landsat 7 ETM+ imagery. We fit models of the absolute percent cover of vegetation groups corresponding to caribou diet items, with a focus on lichens. We used non-parametric multiplicative regression to capture the non-linear relationships between vegetation cover and spectral and environmental data. Different groupings of lichen cover were tried as response variables in addition to usnic lichens to see if other lichen color groups were more detectable. The best fitting lichen model was for usnic acid lichens, which explained 37% of the variation using only three predictors (elevation, bands 1 and 7). Elevation had a non-linear, double-humped shaped relationship to usnic lichen abundance while bands 1 and 7 were positively correlated with usnic lichen cover. These results support previous spectroradiometric ground measurements that indicated usnic lichens were distinctive at those wavelengths. Other vegetation groups had models that explained between 31% and 51% of the variation in cover. Maps of estimated abundance of usnic lichens and other vegetation groups covering the northern half of Denali were generated using our models. These maps enable the study of the role of food resources as a continuous resource in winter habitat selection by caribou, rather than assuming food as a coarser, categorical or thematic variable assigned to discrete areas of the landscape as has been done in most previous studies.