Results

Microhabitat variability

Microhabitat variation within sites was higher than among site variation for half of the microhabitat parameters for which we could fit nested LMMs. Out of these 9 microhabitat parameters that we measured, only 3 varied predictably with elevation (canopy openness, winter minimum soil temperature, and spring days of snow). The parameters that followed an elevational pattern are both expected to be directly (winter minimum temperature, spring days of snow) as well as indirectly (canopy openness) affected by climate (Fig. 2).
Our principal component analysis plot shows that the microhabitat variables we measured cluster broadly by soil characteristics (carbon:nitrogen, fungus:bacteria, water holding capacity), soil moisture (summer minimum soil moisture, spring days of snow), light availability (canopy openness), soil temperature (summer maximum soil temperature), and above-ground temperature (summer minimum plant-height temperature) (Fig. S3). An additional parameter describing soil temperature (winter minimum soil temperature) clusters in the middle of the microhabitat space.

Species establishment

Out of the 25 species that we sowed, 84% (21/25) recruited in at least 1 plot and 56% (14/25) recruited in 8 plots or more (Table S2). Of the species that recruited, seedlings (i.e. at least one seedling) survived for one year for 57% (8/14) of species and survived for two years for 43% (6/14) of species. Almost all species that had sites beyond their range edge recruited and survived beyond their range edge, a pattern seen for species with either high- or low-elevation range edges (Table S2). Almost all of our models met the GLM assumption of independence of residuals. However, many of our models did not meet homogeneity of variance, variance did not equal the mean, or the link function was sometimes inappropriate (> 0.5 difference from 1 for slope of link function). Together with the low proportion of variance (Table S3) from each set of candidate models, we are therefore cautious in interpreting our results.
Overall, we found microhabitat parameters both related and not related to climate identified as the most important parameters in model selection of the effects of microhabitat on establishment, but models had low explanatory power (Tables 1, S3). At the community level, the main patterns we found were that certain microhabitat variables had largely negative or positive effects, but the direction of these effects changed with different life stages (Fig. 3). At the species level, we found that some species only had few microhabitat parameters chosen in model selection, whereas others had many (Fig. S4). Within the same species, parameters usually had the same directionality of effect on likelihood of recruitment, recruit counts, and seedling survival except for T. grandiflora. For two species, A. lasiocarpa andE. lanatum , the directionality of effects was consistent across life stages.

Microhabitat suitability

We generated predictions of microhabitat suitability to better understand if the distribution of microhabitats along climatic gradients suggests range shifts will likely be facilitated (i.e. where suitability increases with elevation), constrained (i.e. decreases), or unaffected (i.e. is constant). We note that due to poor model fit in the original models, our predictions cannot be used to make definitive forecasts of which species are likely to shift their ranges with climate change, but rather use these predictions to assess how microhabitat suitability changes across a large elevational gradient. Our predictions indicate that only range shifts for L. latifolius are likely to be facilitated with increasing microhabitat suitability with elevation (Fig. 4), although our experimental sites extended only just beyond the range limit for the species. For all other species, microhabitat suitability either declines or has no pattern with elevation and thus range shifts are likely to either be constrained or unaffected, respectively (Figs. S5, S6, S7). We also found that suitability patterns were modified by transect and life stage.
Together with these suitability predictions, the microhabitat parameters that significantly vary with elevation can give further insight on what aspects of microhabitat can facilitate or constrain range shifts. For example, the increasing microhabitat suitability for L. latifolius (Fig. 4a) together with the positive effects of canopy openness on the species (Table 1) and canopy openness increasing with elevation (Fig. 2a) point to a likely facilitated range expansion for the species. Spring snow cover shows the same pattern, and thus is a further microhabitat variable that will likely facilitate range expansion for this species. However, winter minimum soil temperature decreases with elevation in the Okanagan transect but positively affectsL. latifolius recruit counts and seedling survival, so this microhabitat variable could constrain range expansion.