Results
With the assumptions of normal distributions and homogeneity of variance not being met, it became apparent to use non-parametric statistics to analyse the response of grizzly bear foods to prescribed burning. A univariate, pairwise Wilcoxen-rank sum test, was first used to analyse each response variable independently. This illustrated any significant differences in the abundance of each grizzly bear food between prescribed fire sites and forested controls. A multivariate analysis, permutational ANOVA, was used to determine which fire, terrain or forest predictor variables have a significant influence on the out come of grizzly bear food items. This multivariate approach also demonstrated an "overall" significance between the two treatments on the abundance of grizzly bear food items.
The 21 response variable were measured in three different ways: 1) Item or shrub density, 2) Percent cover or 3) Fruit density. During the statistical analyses, these response variables were maintained in these categories for simplicity and better graphical representation.
Pairwise Wilcoxon-Rank Sum Test
The results from the pairwise Wilcoxon rank sum indicated that there was a significantly greater density (W=717, P <0.05) of Amalancher anifolia shrubs in prescribed fire sites than in controls (Fig. 7, Table. 3). Ant mounds and logs were also significantly greater in prescribed fire sites than controls (W=738, P<0.05, table. 3). When analyzing the percent cover of herbaceous grizzly bear foods (Fig. 8; Table 4), Rubus ideas was found to cover significantly greater area in prescribed fire sites than controls (W=752, P<0.01), while Vaccinium scoparium was found at greater percent coverage in forested controls (W=455, P<0.01). Fruit density for Ribes oxycanthoides was significantly greater (W=717, P<0.05) in prescribed fire sites than in forested controls (Table 5).
The results from the pairwise Wilcoxon rank sum indicated that there was a significantly greater density (W=717, P <0.05) of Amalancher anifolia shrubs in prescribed fire sites than in controls (Fig. 7, Table. 3). Ant mounds and logs were also significantly greater in prescribed fire sites than controls (W=738, P<0.05, table. 3). When analyzing the percent cover of herbaceous grizzly bear foods (Fig. 8; Table 4), Rubus ideas was found to cover significantly greater area in prescribed fire sites than controls (W=752, P<0.01), while Vaccinium scoparium was found at greater percent coverage in forested controls (W=455, P<0.01). Fruit density for Ribes oxycanthoides was significantly greater (W=717, P<0.05) in prescribed fire sites than in forested controls (Table 5).
Multivariate PermANOVA
To determine the influence that fire, terrain and forest predictor variables have on grizzly bear foods at prescribed fire and control sites, permutational ANOVA was modelled in four ways. First, all response variables were incorporated into a single model, followed by three additional models where the response variables were seperated based on phenotypic characteristics of the bear foods and the resulting method they were sampled. When modelling all response variables together there was no significant difference between the two treatment levels, however the resulting P-value was very close to the alpha level of 0.05 (Table 6). Elevation (R^2=0.0365; P< 0.05) and burn severity (R^2=0.0793; P<0.01; Fig.10) were found to have significant influences on the productivity of grizzly bear foods (Table 6). |
When modelling the shrub and item density predictor variables (Table 7), elevation (R^2=0.0520; P< 0.01) and burn severity (R^2=0.0805; P< 0.05; Fig. 10) were again significant factors. As well, the timing in which fires occurred was also a significant factor (R^2=0.0870; P< 0.05) that influenced the density of shrubs and other bear food items.
There was a significant difference in herbaceous coverage between prescribed fire and control sites (R^2=0.0870; P< 0.05); (Table 8). A.uva-ursi, R.ideas, Triffolium spp, and R. parviflourus occurred at greater coverage in prescribed fire sites while Equisetum spp, T.offininale and V.scoparium were found in greater coverage in forested controls (Fig. 9). Compound topographic index (R^2=0.0805; P< 0.05), elevation (R^2=0.0805; P< 0.05) , and aspect (R^2=0.0805; P< 0.05) were significant factors influencing the cover of herbaceous grizzly bear foods. |