Eight significant predictors were represented in the final biotic
response model (Table 1). These included leaf area index, softwood basal
area, normalized vegetation difference, slope curvature, canopy height,
hardwood basal area, terrain ruggedness, and geographic distance. Just
over half of the significant predictors were biotic, together explaining
41.3% of deviance. Of these, leaf area index was the strongest
predictor (22.02% of deviance explained), while the most important
abiotic predictor was slope curvature (5.32%). Although geographic
distance was a significant predictor, it contributed little to model fit
(3.02%).
In the ecosystem response model, most fitted I-spline curves increased
non-linearly and some were asymptotic (Figure 3). Higher change in model
transformed predictor values (y-axis, Figure 3) corresponded directly
with higher predictor values (x-axis, Figure 3) for leaf area index,
normalized difference vegetation index, and hardwood basal area. Spline
curves for softwood basal area, canopy height, and terrain ruggedness
plateaued at mid to high predictor values. Geography did not contribute
significantly to model fit, suggesting variations in dissimilarity were
driven by relationships with biotic and abiotic predictors that did not
vary markedly across the study area. In general, steeper changes in
I-spline slope (e.g., where canopy height varied between 1 and 15 meters
– Figure 3B) indicated greater dissimilarity between site-pairs,
while flatter curve segments indicated less change (e.g., Figure 3A –
where the terrain ruggedness index exceeds 2). Spline shapes for leaf
area and normalized difference vegetation indices were relatively
congruent. Here they showed prominent change in dissimilarity, between
site-pairs, throughout most of these predictor ranges; parallels in
I-splines shape and magnitude indicated their contributions to
ecosystem-level responses were quite comparable.
In our biotic response model, fitted I-spline curves for slope curvature
and hardwood basal area showed direct correspondence between higher
predictor values and higher ecological distance values throughout most
of their range (Figure 3). In contrast, spline curves for canopy height,
leaf area index, softwood basal area, and normalized difference
vegetation index plateaued at mid or higher predictor values. Terrain
ruggedness and geographic distance leveled off at lower values.