= 461.8, SD = 73.1 g/m²) for each meadow (n = 9,015) within the study area.Ecological Archives E096-161-A2
Jerod A. Merkle, Seth G. Cherry, and Daniel Fortin. 2015. Bison distribution under conflicting foraging strategies: site fidelity vs. energy maximization. Ecology 96:1503–1511. http://dx.doi.org/10.1890/14-0805.1
Appendix B. Details of methods for estimating range size and quality.
B.1 Range size
We calculated summer and winter range size using GPS collar data and winter survey observations using kernel density methods (Worton 1989). We subsampled GPS collar location data because fix success rate was lower with GPS collars deployed in the 1990s. In maximizing the number of GPS locations that could be used, we delineated 2 seasons: summer (May to August), and winter (November to April). We removed September and October from this analysis because ≤ 3 individuals were collared during these months in the 1990s. To keep comparisons consistent throughout the entire analyses, we identified the start of a given year as 1 May. We chose a minimum number of locations per month that would allow > 95% of the individuals in each season and year in the 1990s to be kept in the analysis. We then randomly selected this minimum number per month for each individual and year, providing every individual within each year the same number of GPS locations per month. To avoid including individuals collared in the middle of a season, we removed data for an individual during a season if > 1 month of data were missing.
We calculated seasonal population range size by pooling individual GPS relocation data for each year. Because collars were only deployed on females and observations of bison seen during aerial surveys provided a consistent data set over the study period, we also calculated winter range size using survey observations. Using a 100 m resolution grid, we calculated annual bivariate normal kernels (Worton 1989) for each year for each of the three data sets (GPS collar data in summer and winter, and aerial survey data for winter). Using the reference technique, we used fixed smoothing factors of 1,207 and 1,669 m for summer and winter GPS points, respectively, and 2,757 m for the aerial survey locations. We calculated range size as the 95% contour of each annual kernel. Because a low number of animals were collared during the 1990s and between 2008 and 2009 (i.e., < 4 individuals per year), we pooled location data for these two periods, and estimated one seasonal range size for each.
B.2 Range quality
We calculated mean expected profitability (kJ of digestible energy per min) of the winter population range within the park each year. Using a high resolution land cover classification derived from a SPOT5 multispectral image (taken in August 2008; 10 m resolution; classification accuracy 89%; Dancose et al. 2011), we first parsed the landscape into four land cover types based on forage availability: meadows (open areas > 0.4 ha in size), deciduous forest (including riparian and forest gaps), coniferous forest, and non-vegetated areas (including lakes, rivers, bare ground and roads). Then, following the methods of Dancose et al. (2011), we used a relationship between total aboveground dry biomass (based on field efforts) and the normalized difference vegetation index (NDVI) calculated from the SPOT5 image to estimate mean biomass (= 461.8, SD = 73.1 g/m²) for each meadow (n = 9,015) within the study area.
For pixels classified as deciduous and coniferous forest, we assigned values of 103.2 and 47.2 grams of grasses and forbs per square meter, respectively, as reported by Fortin (2007). For non-vegetated areas, we assigned a biomass value of 0. We removed pixels from the population range that fell outside of the park boundaries (= 24.9% SD = 5.4%), because habitat quality is not well documented outside the park and our focus was on range quality and size within the park. We detected no relationship between year and proportion of pixels falling outside the park, both linearly (F1,13 = 2.07, P = 0.174) and using piecewise regression with a break point in 2006 (F2,12 = 1.56, P = 0.251). Therefore the removal of pixels outside of the park likely had little effect on our inference on range quality within the park. Finally, we calculated expected profitability of each pixel based on its mean biomass estimate following the methods of Merkle et al. (2014), where profitability is derived from a quadratic relationship between mean biomass and mean expected profitability within meadows (R² = 0.31, F1, 21 = 9.55, P = 0.005). Range quality was calculated as a mean of the pixel values within the winter population range each year.
Literature Cited
Dancose, K., D. Fortin, and X. Guo. 2011. Mechanisms of functional connectivity: the case of free-ranging bison in a forest landscape. Ecological Applications 21:1871–1885.
Fortin, M.-È. 2007. Effets de la taille du groupe sur la sélection de l'habitat à plusieurs échelles spatio-temporelles par le bison des plaines (Bison bison bison). Thesis. Université Laval, Québec, Canada.
Merkle, J. A., D. Fortin, and J. M. Morales. 2014. A memory-based foraging tactic reveals an adaptive mechanism for restricted space use. Ecology Letters 17:924–931.
Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in home-range studies. Ecology 70:164–168.