Introduction

Whitebark pine (Pinus albicaulis), a keystone (Ellison et al., 2005), subalpine tree species in western North America is declining throughout its range (Tomback and Achuff, 2010). White pine blister rust (Smith et al., 2008; McKinney et al., 2009; Geils et al., 2010) and mountain pine beetle (Dendroctonus ponderosae; Logan et al., 2010; Raffa et al., 2013) have drastically increased mortality and lowered reproduction of this species. In addition, increased fire return intervals have resulted in a reduction in the open sites thought to be necessary for regeneration and a concomitant increase in competition from later-successional conifers (Murray et al., 1998, 2000; Arno, 2001). Indeed, whitebark pine may have already entered into the spiral of extirpation (Tomback and Achuff, 2010).

Restoration of whitebark pine is ongoing throughout its range (Keane et al., 2012). Actions include thinning competing tree species, developing and planting blister rust-resistant seedlings, and emulating natural disturbance regimes such as fire that are thought to help perpetuate whitebark pine in the lower subalpine (Schwandt et al., 2010). The success of restoration efforts designed to increase the species' recruitment depends upon a good understanding of the regeneration niche (Grubb, 1977). Due to whitebark pine's large geographic range (Weaver, 2001) and wide potential niche (Arno and Hoff, 1989; Larson and Kipfmueller, 2012), research on the regeneration niche would benefit from targeted sampling in a variety of regions. Indeed, most recovery plans for whitebark pine suggest research be conducted on its regeneration niche (Aubry et al., 2008; Keane et al., 2012; Alberta Whitebark and Limber Pine Recovery Team, 2014).

For successful natural whitebark pine regeneration to occur, seed must be produced, removed from the cone and cached by a Clark's nutcracker, germinate, survive, and grow. Seedling abundance is greatly impacted by proximity to seed source and seed source size (Moody, 2006; McKinney et al., 2009; Larson and Kipfmueller, 2010; Barringer et al., 2012). The impact of slope and aspect varies with latitude, with whitebark pine most prevalent on north-facing slopes in the southern part of its range and on south-facing slopes in the north (Arno and Hoff, 1989). This may be due to the balance between moisture and temperature required for germination and early growth. Germination rates appear to be highest on warm sites (McCaughey and Weaver, 1990; Moody, 2006; Larson and Kipfmueller, 2010) as long as adequate moisture is available (McCaughey and Weaver, 1990; Tomback et al., 1993, 2001; Mellman-Brown, 2005; Moody, 2006).

Recruitment of whitebark pine seedlings occurs in a variety of forest and treeline habitats. Whitebark pine is moderately shade tolerant (Arno and Hoff, 1989) and may persist and grow slowly in the understory of late successional stands (Campbell and Antos, 2003), but it is typically outcompeted by spruce and fir (Campbell and Antos, 2003; Moody, 2006; McCaughey et al., 2009; Larson and Kipfmueller, 2010). Whitebark pine seedlings can escape from competition by occupying more harsh, open environments where competing conifers struggle to survive (Callaway et al., 2002; Maher and Germino, 2006). However, seedlings in open environments still benefit from the presence of vegetation or objects that may provide shelter, helping to alleviate water stress induced by high winds or solar insolation (Tomback et al., 1993; Mellman-Brown, 2005; Resler et al., 2005; Izlar, 2007; McCaughey et al., 2009).

There is a dearth of research on whitebark pine at the northern edge of its distribution where environmentally limiting factors might differ. Research in this region could help to inform larger geographic trends in the distribution of whitebark pine and provide insight into the species' adaptation and survival under a changing climate (McLane and Aitken, 2012). Filling this void is important for defining critical habitat under the Canadian Species at Risk Act (SARA) and informing provincial (Alberta and British Columbia) and federal (Canada) recovery plans for whitebark pine by identifying (1) specific microsites where seedlings are favored, (2) forest habitats where disturbance would benefit whitebark pine, and (3) important habitat where development should be avoided.

In this study, we characterized the biophysical associates of whitebark pine regeneration at the northern edge of its distribution in Alberta, Canada, examining environmental factors affecting presence/absence and abundance of whitebark in a variety of mesohabitat types at different elevations. We created statistical candidate models to test competing hypotheses about factors that had the strongest effect. For occurrence we compared between: soil microhabitat, light, and plant community. For density we compared between: soil microhabitat, plant community, and climate and seed source. We predicted that coarser scale variables, such as climate and seed source proximity, would better explain variation in seedling abundance/density, while finer scale factors, such as light, competition, and microsite type, would be best related to occurrence. In addition, we hypothesized that the main drivers of whitebark occupancy would vary among mesohabitats. For example, in forested environments we expected that light availability would drive whitebark pine occurrence, whereas in open and treeline environments we expected occurrence to be impacted mainly by microhabitat factors such as soil and the plant community (which could either have negative effects due to competition or positive effects due to protection from the harsh microclimate).

Methods

Mesohabitat and Microsite Characterization

At each study site we sampled whitebark pine established in different mesohabitats: (1) mixed conifer or pure whitebark pine forest (hereafter referred to as “forest”); (2) open canopied areas below treeline (hereafter referred to as “open”); and (3) within the transition zone from forest to alpine tundra (hereafter referred to as “alpine-treeline” [AT]). In each mesohabitat at each study site, transects—typically 100 m long—were established within a homogenous area representative of the mesohabitat. The forest transect was selected using the same starting point as the formerly established health transects. Open and alpine-treeline transects were selected based on a brief exploration of the site. At several sites, two transects were established in a given mesohabitat type in order to increase replication and capture additional variation (e.g., we might establish transects in two open mesohabitats that exhibited different slopes or aspects at the same site).

At each transect we recorded slope, aspect, pH of the surface mineral soil at the transect start point, and a qualitative description of soil moisture regime and of nutrient regime. Aspect was changed to heat load index (HLI) by converting degrees to a linear value from 0–1 and transforming the data so the warmest aspect (SW or 225°) equals 1 and coldest aspect (NE or 45°) equals 0 (Beers et al., 1996; as cited in Moody, 2006). We retrieved several climate variables using a climate model based on elevation, longitude, and latitude developed by Hamann et al. (2013). These variables are accurate to 1 km2 and can be used with confidence to predict climatic variation as related to elevation. Further details on each of these climate variables can be found in Table 1. For each transect, we calculated canopy cover, total stand density, whitebark pine stand density (as a proxy for seed production), understory vegetation cover, species richness, and availability of different regeneration substrates by taking the average of data collected in the microsite plots (procedure below). Seedling density was quantified by recording the seedlings within 2 m of either side of each transect. All individual whitebark stems shorter than 1.3 m were included in the count. Transects that had no seedlings within 2 m were eliminated from the analysis of density (one forest, three open, and two alpine-treeline transects).

To examine the associates of whitebark pine seedling occurrence, we compared the biophysical characteristics of microsite plots (1 × 1 m) containing whitebark pine seedlings (hereafter referred to as “occupied plots”) to plots without whitebark pine seedlings (hereafter referred to as “unoccupied plots”). We divided each transect into five 20-m segments and selected an occupied and unoccupied plot in each segment for further characterization. The occupied plot was established at the first whitebark pine seedling found in a transect segment (within 2 m either side). If no seedling was found within this area, we extended our search to within 4 m of the transect line, ensuring that this area still fell within the target mesohabitat. The unoccupied plot was randomly located with the constraint that there should be no whitebark pine seedling within 2 m. In each plot we quantified: organic layer depth, litter depth, canopy cover, and tree basal area for each tree species. We visually estimated percent cover for ground cover types (moss, lichen, rock, mineral soil, downed woody material, and cryptogamic crust) and percent cover to species for vascular plants, separately recording seedlings and trees of the same species (Table 1). Canopy cover was determined using a convex spherical densiometer. Basal area was determined by using a basal area prism (BAF 4). From the plant community data, we derived several variables: total vegetation cover (sum of covers for all species); species richness (number of vascular species per plot to use as a proxy for soil productivity; Kirkman et al., 2001); cover by functional group—trees (cover below 1.4 m in microsite plot of branches and stems of saplings and mature trees), seedlings (all trees <1.4 m in height, other than whitebark pine), graminoids (grass, sedge, and rush species), forbs (herbaceous broadleaf vascular species), prostrate shrubs (woody vascular species <5 cm in height), and upright shrubs (woody vascular species >5 cm height).

TABLE 1

Predictor variables used in linear models, the level (mesohabitat or microsite) at which they were sampled, and the candidate model of which each variable was considered a part.

Results

Comparisons of the candidate models for seedling occurrence illustrated that the most important factors differed among the forest, open, and AT mesohabitats, although some factors were important in all mesohabitats (Table 2). Notably, tree cover in the microsite plots was negatively related to occupancy in all three mesohabitats. For occupancy in the forest mesohabitat, the soil microhabitat model was the best, whereas the plant community model was the best in the open mesohabitat; in the AT mesohabitat, the microhabitat and plant community models had very similar AIC values. In all cases, the combined model incorporating variables from the three candidate models had the lowest AIC. The validation showed reasonably good performance of these models with correlation coefficients between observed and predicted values of 0.60, 0.56, and 0.56 for the Forest, Open, Alpine-Treeline mesohabitat models, respectively.

In the forest mesohabitat (Table 2, part A), the soil microhabitat model was the best; it included cover of bare mineral soil and species richness, which were both positively related to whitebark pine seedling occupancy. The combined model included these two variables plus tree cover in the microsite plot, which had a negative association with whitebark pine occupancy.

In the open mesohabitat (Table 2, part B), the plant community model was the best; it included cover of trees and seedlings as having a negative impact, and cover of prostrate shrubs as having a positive impact on the probability of whitebark pine seedling occupancy. The combined model was the best overall and it included these three variables along with a negative impact of the cover of rock.

In the alpine-treeline (AT) mesohabitat (Table 2, part C), the plant community, light, and soil microhabitat models had similar AIC values of 386.2, 387.0, and 387.5, respectively. In the plant community model, occupancy was negatively associated with tree cover and positively associated with cover of prostrate shrubs in the microsite plots, while the light availability model included only the negative association with tree cover. Similar to open mesohabitats, in the soil microhabitat model, cover of rock was negatively associated with whitebark pine occurrence and this model also included a positive effect of litter depth. The combined model incorporated the negative impact of tree and rock cover along with the positive association with litter depth.

TABLE 2

Results for the optimum candidate and combined models for whitebark pine seedling occupancy in: (A) forest; (B) open; and (C) alpine-treeline mesohabitats. For each model the estimate for the intercept, the terms and their coefficients are presented. Akaike's Information Criterion (AIC) for each model is also given and was used to choose among competing models. Coefficients represent the effect (negative or positive) on the probability (log odds) of a plot being occupied by a whitebark pine seedling. Also presented is the correlation between predicted and observed values for the model with the lowest AIC value in each mesohabitat. Explanation of abbreviations for the terms in the models can be found in Table 1.

For seedling density, the climate model consistently outperformed the seed source, soil microhabitat, or plant community models for all three mesohabitats (Table 3). Indeed, in the forest and open mesohabitats, no additional terms entered into the seed source, microhabitat, or plant community models. Further, for the forest and open mesohabitats, only climate variables entered into the combined model; for the AT mesohabitat the combined model included the climate variables along with pH and had a slightly lower AIC than the climate model. The validation showed good performance of the optimal models with correlation coefficients between observed and predicted values of 0.83, 0.92, and 0.72 for the best forest, open, and alpine-treeline mesohabitat models, respectively.

Heat load index had a strong positive effect in all three mesohabitats, indicating that seedling density increased as the aspect approached the southwest. Mean warmest month temperature (MWMT) was negatively associated with seedling density in the open mesohabitat but positively associated with density in the AT mesohabitat; further, the models for both these mesohabitats included an interaction between MWMT and heat load index (HLI) (Table 3) (Fig. 1). In open mesohabitats, increasing MWMT was associated with decreased seedling density. This contrasted the effect of MWMT in AT mesohabitats where increased MWMT was associated with increased seedling densities. The interaction between HLI and MWMT was similar in the optimal models for both open and AT mesohabitats (although stronger in the latter). The interaction demonstrated increased MWMT resulted in a decrease in the positive effect of HLI. Only in the AT mesohabitat did any terms other than climate variables enter any models. In the seed source model, there was a positive association of the basal area of mature whitebark pine with seedling density, and a small negative effect of pH was observed in the soil microhabitat model.

TABLE 3

Results for the optimum candidate and combined models for whitebark pine seedling density in: (A) forest; (B) open; and (C) alpine-treeline mesohabitats. For each model the estimate for the intercept, the terms, and their coefficients are presented. Akaike's Information Criterion (AIC) is also given and was used to choose among the competing models. Coefficients represent the linear effect (negative or positive) on the density (base-two log) of whitebark pine seedlings (seedlings ha-1). For models that include only the intercept, no variables were found to improve the model (based on AIC). Also presented is the correlation between predicted and observed values for the model with the lowest AIC value in each mesohabitat. Explanation of abbreviations for the terms in the models can be found in Table 1.

Discussion

The comparison of candidate models for occupancy of microsite plots by whitebark pine seedlings showed that occupancy is driven by soil microhabitat in forest mesohabitats and by plant community in open mesohabitats below treeline and alpine-treeline mesohabitats. Comparison of candidate models for seedling density, on the other hand, supported the hypothesis that climate is the driver of density. For seedling occupancy, the combined model that incorporated components of both the plant community and soil microhabitat models was superior to any of the simple candidate models in all mesohabitats. Evaluation of the simple candidate models provided information on how different ecological factors influence occurrence; this will be helpful for determining where to focus resources on conservation efforts. However, ecological systems are complex and the combined model better reflected the complexity of factors influencing whitebark pine occurrence.

For occupancy in the forest mesohabitat, the soil microhabitat was the best model, and this contradicted our hypothesis. As whitebark pine is considered moderately shade intolerant and isn't known to compete well in understory environments (Arno and Hoff, 1989), we had expected light availability to account for the most variation. It is important to note, however, that we modeled occupancy and not growth. Research has shown that whitebark pine seedlings have the ability to survive suppressed for many years in the understory (Campbell and Antos, 2003). Our results suggest that in these environments availability of appropriate soil microhabitats that facilitate germination and initial growth of seedlings is a key limiting factor, specifically bare mineral soil and areas of high plant species richness. Bare mineral soil is an important substrate for whitebark pine establishment (McCaughey and Schmidt, 1990), while species richness may indicate sufficient moisture availability (Kirkman et al., 2001) or shared mycorrhizae, both of which are vital in early seedling survival and growth (McCaughey and Weaver, 1990; Tomback et al., 1993, 2001; Mellman-Brown, 2005; Moody, 2006). Our results suggest that in areas where appropriate soil microsite conditions exist for whitebark pine germination and early survival, conservation efforts might be better focused on selective removal of competing tree cover than to employ prescribed fire, because while the latter would destroy existing seedlings the former could allow for their release (Gelderman, 2015).

FIGURE 1.

Graphical representation of the optimal model predicting seedling density as a function of the interaction between mean warmest month temperature (MWMT) and heat load index (HLI) (see Table 3, parts B and C) for: (A) Open and (B) Alpine-Treeline (AT) mesohabitats. Each line was constructed using a range of values for HLI while holding MWMT at the value shown. The range of MWMT at our sites was 9.1–11.8 °C for open and 8.3–9.9 °C for AT mesohabitats, and the range of HLI at our sites was 0–1 for both mesohabitats. Note that the optimal model for seedling density in alpine-treeline mesohabitats also includes a negative effect of pH, which is not shown here.

As hypothesized, occupancy for both the open and AT mesohabitats was best accounted for by the surrounding plant community. This result suggests availability of suitable soil microhabitats for establishment is not a limiting factor in these mesohabitats, while the adverse effects of nearby competition and harsh environmental conditions become more important. Negative impacts included cover of mature trees and seedlings of other species, while cover of prostrate shrubs was positively related to seedling occupancy. While mature trees can play a facilitative role in AT environments (Resler et al., 2014), many studies have demonstrated a negative effect of trees on whitebark pine seedling occurrence (Arno and Hoff, 1989; Campbell and Antos, 2003; Moody, 2006; Larson and Kipfmueller, 2010). The negative impact of nearby mature trees is clear in our study, as occupancy in all three mesohabitats had a significant negative association with tree cover in microsite plots.

Whitebark pine seedlings are noted as having the greatest ability of all subalpine conifer species to survive in open exposed habitats (Callaway et al., 2002; Maher and Germino, 2006), but they still benefit from some protection from wind and solar desiccation (Izlar, 2007; McCaughey et al., 2009). Our results add to the growing literature regarding the benefit of prostrate shrubs in such habitats. Prostrate shrubs do not limit light availability and provide several facilitative effects, such as reducing wind speeds, dampening variation in ground temperature, holding moisture (Körner, 2003), and potentially sharing mycorrhizal associations (Nara and Hogetsu, 2004). Given the potential benefits, further research should investigate specific effects of prostrate shrubs on whitebark pine seedling survival and growth and determine whether a specific mycorrhizal association exists between these shrubs and whitebark pine.

Contrary to our predictions, climate rather than seed source predominantly drove whitebark pine seedling density. Heat load index (HLI) had a strong positive association with seedling density across all three mesohabitats, which emphasizes the importance of aspect in whitebark pine regeneration near the northern limits of its distribution. Seedlings on southwest aspects are likely exposed to longer growing seasons and additional heat units in comparison to seedlings on more northeast aspects, both of which could explain increased seedling densities. Another factor causing increased whitebark pine density on southwest slopes is perhaps an increase of Clark's nutcracker seed caching on southwest-facing slopes, where late snow arrival and early snow melt provide access to caching sites; this has been observed in studies in other areas of whitebark pine's range (Tomback, 1978; Lanner, 1982; Lorenz et al., 2011).

We observed a contrasting effect of mean warmest month temperature (MWMT) in open and AT mesohabitats. Increasing MWMT was associated with a decrease in seedling density in open mesohabitats but an increase in density in AT mesohabitats. This contrasting effect may be explained simply by the location of open and AT mesohabitats in relation to each other. Recall that MWMT was largely based off of elevation, and increases or decreases in MWMT likely reflect downward or upward changes in slope position, respectively. Increasing MWMT in open mesohabitats reflects a movement downslope into areas where the climate is too warm to support seedlings. Likewise, decreasing MWMT in AT transects reflects a movement upslope where growth of trees becomes more and more difficult. The interaction of MWMT with HLI in these two mesohabitats also makes intuitive sense. In open mesohabitats, the positive effect of HLI is reduced as you move downslope (where MWMT is higher) because the southwest exposure exacerbates the effect of increased temperature. Further, in AT mesohabitats the positive effect of HLI is increased as you move upslope (to lower MWMT values) where additional heat units are even more important for the survival of whitebark pine seedlings.

Prior to this study, seed availability had been identified as one of the most significant predictors of seedling abundance (Moody, 2006; McKinney et al., 2009; Larson and Kipfmueller, 2010; Barringer et al., 2012). Our seed source candidate model included only percent canopy cover and mature whitebark pine basal area, which likely did a poor job of representing the seed availability in this region. While whitebark pine basal area did have a small positive effect on seedling density in AT mesohabitats, the lack of seed source impact on forest and open mesohabitats likely reflects our poor characterization of seed source and should not be used as strong evidence that seed source is not important.