Table 2. Model comparisons for the different a priori hypotheses tested in this study.

ID model	Hypothesis	Variables	pD	REP	DIC
4	distance to road (= distroad)	land cover + distroad	20	85	1766
7	distroad x covar	land cover + distroad + distroad:logging + distroad: woodlichen + distroad:water	23	74	1766
6	geog x covar	land cover + X + Y + Y:logging + Y:lichen woodland + Y:water ¶	24	71	1769
9	distroad x covar + climate x covar	land cover + distroad + meanprec + meantemp + meantemp:logging + meantemp: lichen woodland + meantemp:water + distroad: lichen woodland + distroad:logging + distroad:water	28	61	1769
8	climate x covar	land cover + meanprec + meantemp + meantemp: logging + meantemp: lichen woodland + meantemp:water	24	71	1795
3	Geographic (= geog)	X + Y + land cover ¶	21	80	1800
5	climate	land cover + meanprec + meantemp	21	81	1801
2	land cover only	land cover §	19	89	1803
1	intercept only	intercept	9	201	2094

§ land cover = intercept + wildfire + logging + lichen woodland + bare land dominated by lichen +

water + wetlands + coniferous stands PC1 + coniferous stands PC2 + deciduous +

mixed deciduous (see Table 1 for a definition of variables)

^¶X and Y represent the geographic coordinates of centroids for each 100 km² cell.

Geographic-driven selection is represented by model IDs 3 and 6, climate-driven selection by model IDs 5 and 8, and selection that is driven by distance to roads is represented by model IDs 4 and 7. Interaction effects between two variables are indicated by a colon (:). The hypothesis of an additive effect of climate and distance to roads is represented by model ID 9. Variable names: pD = the number of effective parameters; REP = the mean number of data points for each effective parameter; and DIC = Deviance Information Criteria. See Table 1 for a description of variables.