Community Structure and Functional Role of Limber Pine (Pinus flexilis) in Treeline Communities in Rocky Mountain National Park

Abstract

Research Highlights: Limber pine (Pinus flexilis) is abundant in some alpine treeline ecotone (ATE) communities east of the Continental Divide in Rocky Mountain National Park (RMNP) and the Colorado Front Range. Limber pine may be able to colonize the ATE under changing climate aided by directed seed dispersal by Clark’s nutcrackers (Nucifraga columbiana). Cronartium ribicola, white pine blister rust, is a growing threat to limber pine and may affect its functional role within the ATE. Background and Objectives: The ATE is sensitive, worldwide, to increasing temperature. However, the predicted advance of treeline under a changing climate may be modified by tree species composition and interactions. We aimed to (1) examine the conifer species composition and relative abundances in treeline communities with limber pine; (2) assess which functional roles limber pine assumes in these communities—tree island initiator, tree island component, and/or solitary tree; and (3) determine whether limber pine’s occurrence as a tree island initiator can be predicted by its relative abundance as a solitary tree. Materials and Methods: We selected four study sites in RMNP above subalpine forest limber pine stands. We sampled the nearest tree island to each of forty random points in each study site as well as solitary tree plots. Results: Across study sites, limber pine comprised, on average, 76% of solitary trees and was significantly more abundant as a solitary tree than Engelmann spruce (Picea engelmannii) or subalpine fir (Abies lasiocarpa). Limber pine was a frequent component of multi-tree islands in three study sites, the major component in one study site, and dominated single-tree islands at two study sites. At three of four study sites, no species had significantly greater odds of being a tree island initiator. Limber pine was found less often as a tree island initiator than predicted from its relative abundance as a solitary tree, given the likely role of solitary trees in tree island formation.

1. Introduction

The alpine treeline ecotone (ATE), the transitional zone between subalpine forest and alpine tundra, is a mosaic of alpine tundra vegetation such as cushion plants and graminoids, bare substrate, rocks, and semi-upright or krummholz trees growing individually or together in tree islands on the landscape [1,2,3,4,5,6]. The term “krummholz” is German for “crooked wood” and refers to trees that have a stunted, twisted, or mat-like growth form due to harsh conditions at high elevations [7]. The ATE, extending from the timberline (the upper limit of subalpine forest) to treeline (the upper limit of tree growth) [8], is considered highly sensitive to climate change given its close correlation with isotherms, especially mean air and root zone temperatures [4,9,10]. Fluctuations in treeline position also correlate with past climate records, reflecting temperature increases and decreases over time [2,11,12,13,14]. Treelines worldwide have been predicted to advance upward as global average temperatures increase [14]. However, only 52% of treelines—examined between 1900 and 2006—had advanced in response to increases in average temperature [15], indicating that the temperature–treeline relationship is complex and impacted by other variables.

The limits and patterns of tree occurrence in the ATE are controlled by processes at the local as well as regional scales [6,12,16]. At the landscape scale, slope aspect influences solar radiation. For example, in the Northern Hemisphere, southerly aspects receive more solar radiation than northerly aspects and are therefore associated with higher temperatures and lower soil moisture [12,16,17]. The ATE is also a highly heterogeneous system at the local scale, with variations in topography, soil, moisture, and temperature [12]. Nurse objects at treeline, such as plants, rocks, deadwood, and topographic depressions, provide protective microsites that enable seed germination and establishment of new trees [5,13,18,19,20,21], particularly by providing wind shelter and shade [13], but also by reducing convective heat loss to the sky [22,23] and by re-radiating solar energy via the black-body effect [24]. Trees that are well-established in the ATE also modify microclimatic conditions downwind and build soil [25], facilitating the survival of trees and ultimately forming tree islands—that is, patches of trees on the landscape [5,21,23]. In fact, facilitation interactions between plant species or with nurse objects are now considered an important mechanism for shaping communities in harsh environments [26,27,28]. Climatically controlled treelines may be limited by the availability of favorable microsites for seedlings [13,21,23,29,30]. The taxonomic (family-level) composition of ATE communities is also known to limit treeline position, in some cases causing the treeline to be 200–300 m lower than if a different tree taxon were dominant [9], and subtle differences in physiology among tree species may influence the sensitivity of the treeline to changes in climate. Community-level studies of species composition, seedling niche requirements, site conditions, and interactions between species are thus necessary to understand or predict idiosyncratic treeline responses to changing climate.

Given that many mountain ranges of the Rocky Mountains in western Colorado, USA, have peaks above 3200–4000 m elevation, ATE communities are widely represented in the state. Treeline communities in very few of these ranges have been studied; research on the ATE in Colorado has been primarily limited to the Front Range [1,3,7,8,31,32,33,34,35,36,37,38]. The history of research at high elevations in the Front Range provides a valuable context for ongoing work, but studies of ATE communities in the Front Range and elsewhere in Colorado have so far been limited to those dominated by Engelmann spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa). ATE communities including limber pine (Pinus flexilis James) as a community component, while noted to exist [1,32,38], have not been studied.

Limber pine is a five-needle white pine in the subgenus Strobus, section Quinquefoliae, subsection Strobus [39]. It is found from northern New Mexico at the southern limit of its range to Alberta and southeastern British Columbia, and from North Dakota at its eastern extent through eastern California [40]. It has a notably broad elevational distribution, with the lowest recorded tree at around 870 m in North Dakota and the highest at 3810 m in Colorado [40]. Limber pine opportunistically colonizes areas that have been burned or otherwise disturbed [41,42]. However, it is slow-growing and moderately shade intolerant, and thus a poor competitor. Limber pine is seral to faster-growing, shade tolerant conifers [40,43] except on xeric, often steep sites where its tolerance of arid conditions and poorly developed soils provides a survival advantage [43,44,45,46]. Limber pine can establish under arid, windy conditions above timberline where less hardy conifers may not survive without protection [41,45,46]. Its colonization of post-burn areas and of the ATE is due in large part to the seed-caching behavior of Clark’s nutcrackers (Nucifraga columbiana), which is limber pine’s primary mode of seed dispersal [47,48,49,50].

Clark’s nutcrackers harvest ripe limber pine seeds beginning in late August and bury caches of one to fifteen seeds under a few centimeters of soil in both forested and open communities across the landscape [51,52,53]. The seeds are retrieved from late winter through the following summer, and unretrieved seeds may germinate and produce seedlings [51,54]. Nutcrackers provide directed seed dispersal, selecting cache sites near objects that tend to facilitate seed germination and seedling survival [54,55,56]. Nutcrackers often choose cache sites in areas where snow is likely to blow free or melt early (e.g., open areas or near tree boles), allowing for seed retrieval in the winter and early spring. These same sites may be ecologically suitable for seed germination, with extended growing seasons due to earlier snowmelt, gaps in canopy where seedlings are released from light competition, or near objects that may provide shelter from high winds [56,57]. Clark’s nutcrackers will often cache seeds in recently disturbed areas, especially after fire, leading to rapid regeneration of these sites [48,58]. Nutcrackers have also been observed to cache in the ATE [51,52,57,59].

Directed dispersal by Clark’s nutcrackers may explain the close association of limber pine regeneration with a substrate of larger particle sizes and nurse rocks in the ATE, likely providing limber pine with an advantage over wind-dispersed species such as Great Basin bristlecone pine (Pinus longaeva) [60]. In the Great Basin region, limber pine is advancing upslope, with greater densities of limber pine regeneration than Great Basin bristlecone pine regeneration above the timberline, particularly on dolomite soils, which retain water better than other soil types [60,61,62]. Although limber pine seedlings are drought-tolerant [44,60], limber pine seedlings survive better with higher moisture, as demonstrated by simulated climate warming experiments [63,64,65]. Empirical evidence suggests overall that limber pine will be able to advance into the ATE with climate warming, especially if precipitation also increases, which is in agreement with bioclimatic envelope model predictions for the species [66].

Due to its hardiness, limber pine may serve as a nurse plant for other conifers and shrubs, and through this interaction, initiate tree islands in the ATE [45,67]. The pine is drought- and wind-tolerant [40,44]. A closely related species, whitebark pine (Pinus albicaulis), also dependent on nutcrackers for seed dispersal, acts as a tree island initiator in the ATE [68,69,70], with other conifers establishing in more favorable leeward microsites. The likelihood of whitebark pine serving as a tree island initiator can be predicted by its relative abundance as a solitary tree, growing isolated from other trees on the landscape [70,71]. Directed dispersal of whitebark and limber pine seeds by Clark’s nutcrackers to nurse objects and other favorable microsites likely results in facilitative benefits for seedlings in the harsh conditions at the treeline [55,59] and thus potentially helps these species to colonize the ATE.

Limber pine is the only five-needle white pine in Rocky Mountain National Park (RMNP) and is considered a keystone species for its production of large, nutrient-rich seeds that are an important food source for wildlife [58,72,73,74]. In RMNP at treeline, limber pine may contribute to snowpack redistribution and retention, both where it grows in isolation and in tree island associations with krummholz Engelmann spruce and subalpine fir [70]. Limber pine is a species of management concern in RMNP and elsewhere in the Colorado Front Range, where it has recently experienced mortality from mountain pine beetle (Dendroctonus ponderosae) outbreaks beginning in 2003 [75,76] and now faces white pine blister rust, a disease caused by the invasive fungal pathogen Cronartium ribicola [74,77,78]. Limber pine is listed as endangered in Alberta and is being evaluated for listing more widely in Canada due to blister-rust related decline [79,80]. Proactive management and restoration of this species must consider projections of upslope shifts in limber pine’s distribution under a changing climate. The importance and functional role of limber pine in the ATE of the Colorado Front Range, however, has not yet been studied. A better understanding of limber pine’s representation and role in the ATE may inform conservation and management of this species.

Trees in the ATE are often upright, resembling smaller versions of subalpine forest growth forms but with wind-flagged branches, or they are heavily wind-sculpted in shrub-like krummholz mats growing close to the ground. The prevailing wind direction can be determined from tree morphology, or a combination of these forms within a tree island [19]. Previous researchers have assigned functional roles to trees in the ATE based on their position outside of or within tree islands: single-tree island, multi-tree island initiator, multi-tree island component, solitary tree, and satellite tree [68,70,71]. In this study, we sought to (1) characterize the species composition (and relative abundance) and structure of treeline communities with limber pine in RMNP; (2) assess the functional role of limber pine in these communities (i.e., multi-tree island initiator, multi-tree island component, single-tree island, or solitary tree); and (3) determine whether a species’ role as an initiator can be predicted by its abundance as a solitary tree. We also discuss the management implications of our results for limber pine conservation and restoration.

2. Materials and Methods

The Front Range, a segment of the U.S. Rocky Mountains that extends from Casper, WY to Pueblo, CO includes RMNP. RMNP encompasses about a 60 km length of the Continental Divide (and a total park area of approximately 1074 km²) including extensive ATE communities between ~3300 and 3500 m in elevation. The treeline communities in RMNP and along the Front Range east of the Continental Divide are often climatically limited, growing on relatively gentle, convex slopes [6]. The substrate is most commonly a combination of Precambrian granite, schist, and gneiss with eolian deposits of finer sediment [1,6]. The terrain includes glacial cirques, flat-topped ridges, sharp peaks, and moraines sloping away from the continental divide [1]. Vegetation communities observed at the treeline in the park are likely post-Pleistocene glaciation, and are classified as dry and wet meadows above the timberline, and subalpine forest [1,3].

In RMNP, the dominant conifer species at the treeline are Engelmann spruce and subalpine fir [34], with limber pine scattered or in larger numbers in some areas and occasional lodgepole pine (Pinus contorta) [3]. At some sites in the southern parts of the Front Range, Rocky Mountain bristlecone pine (Pinus aristata) also occurs at the treeline, but not in RMNP. Subalpine fir seedlings are typically found in cirques and areas of greater snow accumulation, and Engelmann spruce seedlings are abundant in snowbeds [1,34]. Limber pine, in contrast, is found on dry, exposed sites [1,34].

2.1. Study Site Selection

We used shapefiles of the 2005 RMNP vegetation survey data to find large stands of subalpine limber pine, trails from RMNP park website [81], and a 30 m × 30 m digital elevation model (DEM) raster file from park staff (Scott Esser, personal communication) to select four study sites within RMNP where limber pine was present at the treeline and accessible by trail or road: Rainbow Curve, Ute Trail, Battle Mountain, and Longs Peak. We scouted study sites before including them to confirm limber pine presence and safe access (Figure 1).

We used a Trimble GeoXT geolocator (GeoExplorer 2008 Series) to record the vertices of each study site polygon. We designated timberline as the lower elevational boundary for each study site, and the treeline as the upper boundary of the ATE [8]. In the case of the Rainbow Curve study site, we placed the lower boundary at Trail Ridge Road, below which was subalpine forest. The boundaries for the Rainbow Curve and Ute Trail study sites were horizontally constrained by ridges, but the boundaries for the Longs Peak and Battle Mountain study sites were based on aspect; we determined these limits using a compass in the field. The study site area ranged from 2.74 ha at Rainbow Curve to 8.54 ha at Longs Peak based on the rules we used to delineate each polygon (

Table 1. Summary information for the four study sites in Rocky Mountain National Park.

Study Site	Area (ha)	Mode Elevation (m)	Aspect (s)	Prevailing Wind Direction
Rainbow Curve	2.74	3376	NW	283 (WNW)
Ute Trail	6.62	3472	SE, E	282 (WNW)
Longs Peak	8.54	3410	N, NE, E	268 (W)
Battle Mountain	3.35	3456	N, NW, W	227 (SW)

Aspect, area, and elevation were determined in ArcMap using a 30 × 30 m digital elevation model (DEM). Wind direction was determined by the direction of wind sculpting of tree islands and the table provides the average across islands for each site.

). Forty random points were generated in ArcMap within the boundaries of each polygon.

2.2. Tree Island Sampling

We located the nearest tree island to each random point. We defined a tree island either as a single tree ≥1.0 m in its longest dimension (single-tree islands, Figure 2A) or as a patch of trees with overlapping canopy (multi-tree islands, Figure 2D). The longest dimension was length in the case of krummholz trees and height in the case of upright trees. The definition of single-tree islands distinguished trees large enough to act as a wind break or snow fence from smaller solitary trees on the landscape. Individual trees were assigned one of four functional roles: satellite tree, tree island initiator, tree island component, or solitary tree. Satellite trees were found close to the leeward side of tree islands, and while separate from the tree islands (with no overlapping canopy), they likely received some facilitative benefits (Figure 2B). Tree island initiators occupy the most windward position of a tree island (rightmost tree in Figure 2D) and all leeward trees in the tree island are tree island components. Solitary trees are isolated on the landscape, receiving no facilitative aid from other trees, and are defined here as <1.0 m in their longest dimension (Figure 2C). Both single-tree islands and solitary trees may be considered to be colonizing the ATE and could become pioneers if they facilitate the establishment of other conifers.

Where two random points indicated the same tree island, one point was eliminated by random draw. Where a tree island had an elongate form and overlapped at the leeward end with a neighboring tree island, we differentiated one tree island from the other based on partial separation by sparse vegetation and exposed soil surface. In these cases of merging tree islands (found only at the Rainbow Curve study site), the tree islands clearly were the result of separate initiators and had simply grown together over time. All distinctions were evaluated separately by at least two or three observers.

We took measurements of each tree island to the nearest 5 cm, using a retractable measuring tape or transect tape if the islands were longer than 600 cm. We measured the height of each tree island from the ground to the top of the tallest woody stem. We measured the length as the longest dimension of the tree island, which was often oriented with the prevailing wind direction, and the width perpendicular to length. We recorded each tree island as being krummholz (entirely mat-like and wind-sculpted with no upright stems), krummholz-upright (mostly krummholz, but with some upright stems), or upright (an upright tree that may exhibit flagging but is not krummholz). Trees in the ATE may have an upright growth form, though this form is often flagged and stunted compared to subalpine trees. Engelmann spruce and subalpine fir may be classed as “upright” in the ATE, but often exhibit vegetative layering and multiple clonal stems.

We recorded the number of discrete patches of each species within multi-tree islands. Limber pine will often grow in a clustered or clumped form, due either to multiple seeds from a nutcracker cache germinating and growing in close association or a multi-trunk phenotype of a single individual [82]. The two are not visually distinguishable, and where limber pine trees were identified to have multiple stems, they were counted as one individual. Subalpine fir and Engelmann spruce produce clonal upright stems through underground vegetative growth, and individual trees (genets) are impossible to differentiate from ramets in dense, krummholz tree islands without genetic analysis or destructive sampling [14]. Estimations of proportion based on patch size or counts of discrete stems are likely to over-represent Engelmann spruce and subalpine fir. However, counts of patches would likely under-represent these species. Due to the difficulties discriminating individual trees in multi-tree islands, we chose to report the proportion of multi-tree islands with at least one individual of a species rather than count or area estimates.

For multi-tree islands, we recorded the species of the tree island initiator. We determined the initiator to be the most windward individual, with prevailing wind direction estimated based on wind sculpting and flagging of trees. In a few instances, two trees were equally windward–side-by-side with respect to prevailing wind. When these were of the same species, we recorded that species as most windward. When they were of two different species, they were excluded from the analysis.

2.3. Solitary Tree Sampling

At each random point, we created a circular, 5-m-radius plot (78.54 m²) delineated with surveyor flags to count and describe solitary trees, defined as trees ≤1.0 m in their longest dimension. Where a random point fell within a tree island, the plot center was located 5.0 m downslope in an open area. If two plots overlapped in area due to the proximity of the random points, one of the two was eliminated by random draw. The center of each solitary tree plot was marked with a geolocation using the Trimble. We counted the number of trees of each species within the solitary tree plot. We excluded satellite trees, which we defined as being within a distance from the tree island equal to the height of the leeward side of the tree island. Satellite trees were too few to allow for a separate analysis of this functional role. We measured the height of each solitary tree to the nearest 1.0 cm.

2.4. Statistical Analyses

We used the R package DescTools, version 0.99.36 (Andri Signorell, Zurich, Switzerland), to calculate the median length, width, and height of single- and multi-tree islands at each study site as well as 95% confidence intervals [83]. We used base R (version 3.6.3) (R Core Team, Vienna, Austria) to calculate the number and percent of tree islands classed as each growth form (krummholz, krummholz-upright, or upright) for each study site. We also used base R to calculate the median tree density of solitary plots for each study site and to calculate the mean heights of solitary trees in each site. We used the R package Rmisc, version 1.5 (Ryan M. Hope), to calculate the 95% confidence intervals for mean densities and heights [84].

For each study site, we used Equation (1) to calculate the proportion of single-tree islands for each species,

$$\widehat{p}=\frac{x}{n}$$

(1)

where x is the number of single-tree islands of a given species and n is the total number of single-tree islands. We also used Equation (1) for multi-tree islands, with x being the number of multi-tree islands containing at least one individual of a given species and n being the total number of multi-tree islands. Proportions for multi-tree islands therefore did not add to one because multi-tree islands often contained more than one species. We used Equation (2) to calculate the 95% confidence intervals (CI) for proportion estimates [85],

$$CI=\widehat{p} \pm 1.96\sqrt{\frac{p\left(1-p\right)}{n}}$$

(2)

where 1.96 is the critical value from the standard normal distribution; p is the proportion of single-tree islands for each species; and n is again the total number of single-tree islands. For each species, we used Equation (3) to calculate the mean species proportions ($\stackrel{-}{p}$) across solitary plots in each site,

$$\stackrel{-}{p}=\frac{{{\displaystyle \sum }}_i^y{\left(\frac{x}{n}\right)}_i}{y}$$

(3)

where x is the number of individuals of a given species in the plot; n is the total number of all individuals of all species in that plot; and y is the total number of plots in the study site. We used Equation (2) to calculate the 95% confidence intervals for mean proportions of each species in solitary plots.

To determine whether any one species had significantly greater odds of being the tree island initiator at a site, we did an odds ratio analysis [86]. We used Equation (4) to calculate the odds of a species being the tree island initiator at a site,

$$odds=\frac{p}{1-p}$$

(4)

where p is the proportion of multi-tree islands with a given species as the initiator. We used Equation (5) to calculate the odds ratios (OR), comparing the odds of one species (x) being the initiator to the odds of another species (y) being the initiator,

$$OR=\frac{odd{s}_x}{odd{s}_y}$$

(5)

where odds ratios are equal to one, the odds of one species being the initiator is the same as the odds of another species being the initiator. To estimate the uncertainty around these comparisons and to determine whether any significant differences existed, we used Equation (6) to calculate 95% confidence intervals for the odds ratios [85],

$$CI=e^{\ln \left(OR\right)\pm 1.96\ast \sqrt{\frac{1}{x}+\frac{1}{\left(n-x\right)}+\frac{1}{y}+\frac{1}{\left(n-y\right)}}}$$

(6)

where 1.96 is the critical value from a standard normal distribution; x is the count of the first species; y is the count of the second species; and n is a total count of individuals in the island. The 95% confidence interval was calculated for the log odds ratio, and then back transformed.

To determine whether the relative abundance of a species as a solitary tree could predict its occurrence as a tree island initiator, we used Equation (7) to calculate the 95% confidence intervals for differences in proportions [71,85],

$$\left({\widehat{p}}_1-{\widehat{p}}_2\right)\pm 1.96\ast \sqrt{\frac{{\widehat{p}}_1\left(1-{\widehat{p}}_1\right)}{n_1}+\frac{{\widehat{p}}_2\left(1-{\widehat{p}}_2\right)}{n_2}}$$

(7)

where $\widehat{p}$₁ is the proportion of multi-tree islands for which a species is the initiator; $\widehat{p}$₂ is the mean proportion for that species across solitary plots; $n_1$ is the number of multi-tree islands; and $n_2$ is the number of solitary plots. Where the difference between $\widehat{p}$₁ and $\widehat{p}$₂ is greater than 0, the species is the initiator more often than would be expected from its abundance as a solitary tree, and where the difference is less than 0, it is the initiator less often than expected from its abundance as a solitary tree. If the confidence interval includes 0, then the species’ representation as an island initiator is consistent with (could be predicted from) its representation in solitary plots.

3. Results

3.1. Community Structure

Study sites differed in community structure in terms of the proportion of tree islands that were multi-tree vs. single-tree, the dimensions of tree islands, the growth form of islands (krummholz, upright, or a mix of krummholz with upright stems), and the density of solitary trees. The Rainbow Curve study site had the greatest proportion of multi-tree islands (61.5%), which also tended to be long and narrow, earning the designation “stringers” (

Table 2. Descriptive statistics for single- and multi-tree islands at each study site.

Single-Tree Island Metrics
Study Site (No. Islands, %)	Median Island Length (cm) (95% CI)	Median Island Width (cm) (95% CI)	Median Island Height (cm) (95% CI)
Rainbow Curve (15, 38.5%)	300	260	200
	(245, 400)	(180, 340)	(140, 245)
Ute Trail (21, 53.8%)	180	205	120
	(115, 340)	(130, 285)	(65, 145)
Longs Peak (23, 54.8%)	235	200	115
	(170, 305)	(150, 240)	(95, 155)
Battle Mountain (27, 81.8%)	170	160	110
	(130, 240)	(115, 255)	(60, 210)
Multi-Tree Island Metrics
Rainbow Curve (24, 61.5%)	1018	378	265
	(430, 1550)	(255, 510)	(180, 380)
Ute Trail (18, 46.2%)	1128	715	250
	(790, 2235)	(560, 1100)	(190, 300)
Longs Peak (19, 45.2%)	560	385	130
	(305, 745)	(200, 645)	(75, 160)
Battle Mountain (6, 18.2%)	275	248	135
	(95, 475)	(90, 425)	(45, 305)

Metrics were calculated separately for single-tree and multi-tree islands. The number of single- or multi-tree islands is given in parentheses between the study as well as the percent of the total number of islands that were single- or multi-tree.

).

The stringer tree islands tended to increase in height from the windward to the leeward end, producing a wedge shape. The trees at the windward end of these tree islands were krummholz in form, and those at the leeward end were of mixed krummholz-upright growth form; at the Rainbow Curve study site, a large percentage of the growth forms in its stringer-shaped tree islands were classified as krummholz-upright (

Table 3. Growth form classifications for tree islands at each study site.

Study Site	No. Krummholz (%)	No. Krummholz-Upright (%)	No. Upright (%)
Rainbow Curve	6 (15.4)	31 (79.5)	2 (5.1)
Ute Trail	18 (46.2)	15 (38.5)	6 (15.4)
Longs Peak	23 (54.8)	19 (45.2)	0
Battle Mountain	15 (45.5)	13 (39.4)	5 (15.2)

Table 3 shows the number and percent of tree islands classified as krummholz, krummholz with some upright shoots, or upright at each study site.

).

The Ute Trail study site also had long multi-tree islands, but these were wider and so did not have a stringer form (Table 2). The Battle Mountain study site was comprised mostly of limber pine in single-tree islands (Table 2 and Figure 3), and so the tree islands tended to be smaller overall (Table 2).

Approximately half of the tree islands at the Longs Peak and Ute Trail study sites were multi-tree (Table 2), although the dimensions of the Longs Peak study site tree islands were generally smaller.

Tree islands across all study sites were stunted to some degree because of wind-stress and winter desiccation, with very few upright trees. Tree island heights were comparable among the study sites. The Longs Peak study site had the greatest percentage of krummholz tree islands and zero upright tree islands (Table 3) as well as lower median heights (Table 2), suggesting stronger or more frequent winds at this study site than the other study sites. The Ute Trail and Battle Mountain study sites indicated a greater number of sheltered microsites with 15.4 and 15.2% of tree islands classified upright, respectively. The Rainbow Curve study site had the greatest percentage of krummholz-upright tree islands due to the upright stems of trees at the leeward end of stringer tree islands, which benefitted from facilitation by windward trees (Table 3).

Solitary trees generally occurred at low densities across all study sites, except where moister microsites or shelter occurred such as in an area of high snowmelt runoff near the road at the Rainbow Curve study site (Figure 4). Within the 5-m-radius plots (78.54 m²), counts were almost always <10 trees (Figure 4). Median density ranged between 0.000 trees/m² at the Ute Trail study site and 0.038 trees/m² at the Battle Mountain study site. The Ute Trail study site had significantly lower solitary tree density than the Battle Mountain, Longs Peak, or Rainbow Curve study sites. The Battle Mountain study site had a significantly greater density of solitary trees than the Longs Peak study site (Figure 4).

No significant differences in solitary tree height were observed among the study sites (Figure 3).

3.2. Species Composition of Tree Islands

Limber pine was significantly more abundant as a single-tree island species than either Engelmann spruce or subalpine fir at both the Battle Mountain and Longs Peak study sites (Figure 5). At the Rainbow Curve study site, limber pine was significantly more abundant as a single-tree island than subalpine fir, and at the Ute Trail study site, Engelmann spruce was significantly more abundant as a single-tree island than limber pine.

We represented multi-tree island species composition as a proportion of tree islands containing each conifer species, rather than attempting to discriminate individuals from clonal stems in krummholz mats. At the Longs Peak and Rainbow Curve study sites, no significant differences were observed in species representation in multi-tree islands (Figure 6). At the Ute Trail study site, Engelmann spruce was found in a significantly greater proportion of multi-tree islands than limber pine. At the Battle Mountain study site, limber pine was found in all six multi-tree islands while Engelmann spruce was not found in any, and subalpine fir was found in fewer than 25% of multi-tree islands.

In sum, limber pine was a frequent component of multi-tree islands in three study sites, the major component in one study site, and dominated single-tree islands at two study sites.

3.3. Species Composition of Solitary Tree Plots

Limber pine was significantly more abundant in solitary tree plots than either Engelmann spruce or subalpine fir across all study sites (

Table 4. Estimated differences in solitary tree proportions between tree species.

Study Site (Number of Solitary Tree Plots)	Limber Pine Solitary $\widehat{\mathit{p}}$—Engelmann Spruce Solitary $\widehat{\mathit{p}}$ (95% CI)	Limber Pine Solitary $\widehat{\mathit{p}}$—Subalpine Fir Solitary $\widehat{\mathit{p}}$ (95% CI)
Longs Peak (38)	0.607 *	0.683 *
	(0.430, 0.784)	(0.523, 0.842)
Rainbow Curve (39)	0.588 *	0.776 *
	(0.409, 0.768)	(0.643, 0.909)
Ute Trail (39)	0.574 *	0.537 *
	(0.396, 0.752)	(0.352, 0.722)
Battle Mountain (33)	0.633 *	0.749 *
	(0.446, 0.819)	(0.593, 0.904)

Differences in proportions and 95% confidence intervals were calculated for each study site. Where confidence intervals do not include 0, differences are significant at the α = 0.05 threshold (as indicated by an *).

and Figure 7).

The differences were calculated to confirm the confidence interval assessment (Figure 7) that limber pine was significantly more abundant as a solitary tree than Engelmann spruce or subalpine fir across all study sites. We used Equation (7) to calculate 95% confidence intervals for these differences in proportions (Table 4). None of the confidence intervals included 0; the results indicate that limber pine was significantly proportionately more abundant than subalpine fir or Engelmann spruce.

3.4. Windward Species Analyses

Odds ratios were calculated to compare the probability of finding different species in the most windward position of multi-tree islands (Figure 8).

The Battle Mountain study site was excluded from this analysis because limber pine was the tree island initiator for all sampled multi-tree islands. The 95% confidence intervals of all odds ratios spanned 1.0 across all study sites apart from the Ute Trail study site, indicating that no species was significantly more likely to be the initiator than any other. At the Ute Trail study site, Engelmann spruce was the tree island initiator significantly more often than limber pine.

To determine whether the probability of finding a given species in the most windward position can be predicted from (is consistent with) its proportional abundance in solitary tree plots, we calculated differences (and 95% confidence intervals) between the windward tree proportions and solitary tree proportions for each species at each study site (

Table 5. Estimated differences between windward proportions and solitary proportions.

Study Site	Limber Pine	Engelmann Spruce	Subalpine Fir
Longs Peak	−0.278 *	0.066	0.247 *
	(−0.482, −0.074)	(−0.101, 0.232)	(0.085, 0.409)
Rainbow Curve	−0.440 *	−0.397 *	0.249 *
	(−0.637, −0.243)	(−0.595, −0.199)	(0.107, 0.391)
Ute Trail	−0.579 *	0.496 *	0.083
	(−0.756, −0.402)	(0.310, 0.680)	(−0.096, 0.263)
Battle Mountain	0.206 *	−0.161 *	−0.045
	(0.068, 0.344)	(−0.287, −0.036)	(−0.116, 0.026)

Difference between windward proportions and solitary proportions and 95% confidence intervals, were calculated for each species and study site. Where the entirety of the confidence interval was below zero, windward proportions were significantly lower than solitary plot proportions. Where the confidence interval was above zero, windward proportions were significantly greater solitary proportions. Significant differences are noted with an asterisk (*).

). We found that limber pine was found in the windward position significantly less often than would be expected from its proportional abundance as a solitary tree at three of four sites, and significantly more often than expected at the Battle Mountain study site. Engelmann spruce was found significantly more often in the windward position than expected at the Ute Trail study site, but less than expected at the Rainbow Curve and Battle Mountain study sites. Subalpine fir was found in the windward position significantly more often than would be expected from its representation in solitary plots at both the Longs Peak and Rainbow Curve study sites.

In sum, limber pine in general did not differ overall from other species in its proportional occurrence as a tree island initiator of multi-tree islands, and its proportional occurrence as a tree island initiator was less than its proportional abundance as a solitary tree.

4. Discussion

4.1. Limber Pine as a Colonizer in the ATE

Across sites, limber pine consistently comprised the greatest proportion of trees in solitary plots. Limber pine’s abundance in solitary plots suggests that limber pine is well-suited to colonize treeline environments without facilitation from other conifers. Its prevalence as a solitary tree may be due to directed dispersal by Clark’s nutcrackers, which bury limber pine seeds near objects (mnemonic aids for cache retrieval) and so likely provide seeds with protected microsites [59,67]; solitary trees were often noted to be located in the lee of rocks. In contrast, Engelmann spruce and subalpine fir are wind-dispersed species, and so their ability to colonize the ATE would depend on the co-occurrence of nearby seed sources, upslope wind patterns, and the prevalence of favorable microsites [6]. Seeds of wind-dispersed species may not land in suitable microsites or may be more vulnerable to predation.

Clark’s nutcrackers also tend to cache in sites that allow for easier seed retrieval in the winter and spring, and so limber pine’s distribution as a solitary tree may be a result of caching behavior in areas of low snowpack [40,56] in combination with the pine’s tolerance of moisture stress [44]. Limber pine may therefore be dominant at sites that are less suitable for Engelmann spruce or subalpine fir establishment due to wind exposure, moisture limitations, or edaphic conditions. Limber pine seedlings are generally drought tolerant [44] and are more tolerant of moisture stress than Engelmann spruce [65] (and likely also subalpine fir). As climate warming causes earlier snowmelt in the ATE [87,88,89,90], thus increasing moisture stress during the growing season, limber pine’s relative abundance in the ATE may increase, with beneficial influences on local hydrology through snow retention and for slope stabilization.

Bioclimatic envelope models predict that limber pine will move upslope in RMNP under a changing climate [66]. While seed germination and seedling survival rates have not been estimated for limber pine at the treeline, whitebark pine germination rates at the treeline in two study areas in the northern Rockies ranged from 42.5 to 64.1% [91] and annual survival rate (estimated over five years) ranged from 0.571 to 0.992 [20]. Seed germination and seedling survival of limber pine at the treeline may also be quite high, compared to other conifers [92]. For example, in the arid White Mountains of eastern California, limber pine has advanced into the ATE above bristlecone pine (Pinus longaeva), possibly aided by mycorrhizal mutualisms as well as directed dispersal by Clark’s nutcrackers [60,61,93]. Our finding that limber pine is significantly more abundant as a solitary tree than other ATE conifer species in RMNP may result from higher seedling survival on harsh sites.

4.2. Role of Limber Pine in Tree Islands in the ATE

While Engelmann spruce, subalpine fir, and other conifers outcompete limber pine under advancing succession in subalpine and lower-elevation forests [40,41,42], the same may not be true at treeline. As a component of multi-tree islands, limber pine has a roughly even proportional abundance with Engelmann spruce and subalpine fir, suggesting that the occurrence of different species of trees within tree islands is stochastic. Furthermore, competitive interactions may not be important among these species in the ATE. In general, studies indicate that facilitative interactions tend to be more prevalent under stressful conditions [13,26,94,95]. Holtmeier (2009) found limber pine to be a component of multi-tree islands with Engelmann spruce and subalpine fir elsewhere in the Front Range [24].

With respect to facilitation, no conifer species had significantly greater odds of being found in the windward position of a tree island than any other species across study sites except at the Ute Trail study site. There, Engelmann spruce had significantly greater odds of occupying the windward position than limber pine. Thus, based on our data, no species is a more frequent tree island initiator than any other species and the species in the initiator role may be stochastically determined at these study sites.

We also found that a species’ relative abundance as a solitary tree does not predict its likelihood of becoming a tree island initiator, contrary to previous findings for whitebark pine [70,71]. There may be fundamental differences in local site characteristics where limber pine establishes as a solitary tree as opposed to local sites that support multi-tree islands. In most cases, limber pine is found less often as a tree island initiator in our study sites than we would expect, given its abundance as a solitary tree, except for the Battle Mountain study site. Based on our hypothesis that tree islands are established by individual trees, which potentially ameliorate conditions downwind for other conifers, these results may mean that limber pine may not provide a sufficiently protective leeward microsite for seed germination or seedling survival. Limber pine does not reproduce by growing clonal stems or through stem-layering [24]. The dense vegetative growth of krummholz subalpine fir and Engelmann spruce may provide better shelter for new trees; temperatures and moisture are more stable within and leeward of spruce-fir tree islands [24]. Dense spruce-fir tree islands also promote soil development, particularly inside the islands where podzolization may occur [24], but soil properties associated with limber pine tree islands have not yet been investigated. However, whitebark pine, which is similar in morphology to limber pine, was found to provide protective leeward microsites that ameliorated soil temperatures and wind [19,70,96]. We suggest that the discrepancy may be explained by the conditions where limber pine grows as a solitary tree, which may be so harsh that seeds of other tree species are unlikely to germinate, and/or that seedlings have low survival probability. In fact, few individuals of Engelmann spruce or subalpine fir were found as solitary trees; Engelmann spruce and subalpine fir tended to be found in multi-tree islands. Further work is necessary to determine which of these hypotheses is best supported; our sampling frame was limited to each of the four study sites selected.

4.3. Differences among ATE Community Structure and Abiotic Characteristics

Four ATE communities in geographic proximity are insufficient to generalize differences in the structure and species composition of krummholz ATE communities on the eastern slope of the Colorado Front Range. Our goal was to determine if and to what extent limber pine occurred in these communities, and its ecological function in these communities. The present study may be used to generate hypotheses about how differences in site conditions may yield differences in species, and thus community structure, and so guide future work.

Where limber pine was found in greater abundance as single-tree islands (the Battle Mountain and Longs Peak study sites), ATE communities are comprised of many individual limber pine trees growing solitarily at low densities across the landscape. In contrast, the Ute Trail and Rainbow Curve study sites were characterized by multi-tree islands that formed large patches (at the Ute Trail study site) or elongated “stringers” oriented with prevailing wind direction (at the Rainbow Curve study site). The proportion of multi-tree islands vs. single tree islands, the tree island metrics, and the solitary tree density may be associated with species composition; a formal analysis with a sufficient sample size could assess this.

Limber pine was most abundant at the Battle Mountain and Longs Peak study sites, which had northerly aspects. The Battle Mountain study site consisted of N/NW aspects and a SW prevailing wind, and the Longs Peak study site had N/NE aspects and a W prevailing wind (Table 1). Limber pine was not very abundant above the timberline at the Ute Trail study site, which had a SE/E aspect. In addition to a greater abundance of spruce and fir, shrubs and forbs were prevalent at the Ute Trail study site, particularly Juniperus communus, Betula glandulosa, and Salix sp., indicating higher snowpack. Fewer solitary conifers were found in general at this study site. It may be that the SE/E aspect of this study site provides more shelter from a WNW prevailing wind, allowing other plant species to predominate. The Ute Trail study site was also at the head of a gulch, and the deeper snowpack and higher moisture may be more amenable to Engelmann spruce and subalpine fir establishment [97,98]. Furthermore, limber pine seedlings are inhibited from establishing in sites with deeper average snowpack due to increased risk of snow fungus infection [24].

The relative importance of competitive and facilitative interactions may be inversely related depending on the harshness of conditions, particularly with respect to wind and snowpack [25]. The Rainbow Curve study site had a NW aspect and a WNW prevailing wind, with local peaks on either end of the study site forming a sort of saddle through which wind appears to be funneled. Longer, multi-tree krummholz islands with an even representation of species were found toward the center of this saddle, while single-tree limber pine islands tended to be found on the steeper slopes directly facing prevailing winds. It is possible that variation in topography over short distances, and especially convex as opposed to concave landforms in conjunction with prevailing winds determines the distribution of solitary trees vs. multi-tree islands.

4.4. Considerations for the Conservation and Management of Limber Pine

White pine blister rust has impacted limber pine on 88% of plots evaluated in this northernmost part of its range and limber pine is listed as endangered in Alberta [79,80]. More than 60% of limber pine trees are infected in many stands in Wyoming [99]. Since blister rust was first identified in RMNP in 2010, the infection has spread. In 2019, more than 150 trees were pruned of blister rust cankers in the Beaver Ponds area of RMNP (B. Verhulst, personal communication). While treeline areas are thought to be at lower risk of infection because of dry, windy conditions unfavorable to spore transmission [100], we found cankers on multiple trees at the Rainbow Curve and Ute Trail study sites in 2019 (Sindewald et al., unpublished data). Whitebark pine, a closely-related species that has experienced substantial decline throughout its range due to white pine blister rust, has also experienced pervasive infection in the ATE [70,71,77,79,100]. Limber pine may therefore also be vulnerable at the treeline.

A proactive conservation plan for limber pine was written for RMNP in 2015 (and expanded and updated in 2019) [74,101]. The principal management approaches are to (1) plant seedlings and sow seeds to increase limber pine populations overall, collecting seeds from across the range to preserve limber pine’s genetic diversity; (2) identify alleles that confer complete or partial resistance to white pine blister rust and balance the frequency of these alleles in limber pine populations (high frequencies of complete resistance may accelerate blister rust virulence to the major resistance allele); and (3) prune branches with sporulating cankers to reduce the spore load that could spread blister rust across the landscape [74].

To implement the proactive management plan effectively, the future of limber pine under a changing climate must be considered. Bioclimatic envelope models project that limber pine will move upslope with increasing temperature [66], and this projection is supported by our findings of a prevalence of solitary limber pine trees in all four communities examined. However, a wider study of ATE communities where limber pine is a component is needed to determine whether our findings hold across ATE communities with limber pine. The occurrence of blister rust infection in these communities raises further concerns that limber pine in the ATE may not be able to survive the spreading pathogen or advance upslope [102]. If blister rust limits limber pine’s expansion into the ATE, in combination with the upslope advance of more competitive Engelmann spruce and subalpine fir at limber pine’s lower elevational limits, limber pine may not be able to persist under novel climate scenarios at all. Our findings suggest that further study of limber pine in the ATE may be important for conservation of the species under changing climate and advancing blister rust.

5. Conclusions

Limber pine was the most abundant solitary tree within each of the four communities, indicating that this species may be a colonizer of the ATE at sites where conditions may be unsuitable for other species. Limber pine was also predominant at two sites as single-tree islands (large, single limber pine), suggesting that once established, it may persist or even thrive at the examined sites without facilitation from other species. None of the conifer species predominated as a tree island initiator. Instead, species’ occurrence in this functional role seems to be stochastic. While all three study sites were directly upslope of subalpine forest communities where limber pine is abundant, some differences were observed in limber pine’s proportional representation.

Here, we provide the first descriptive characterizations of four different limber pine ATE communities in RMNP, showing that under certain conditions, limber pine may be the only tree to establish and form a low-density community of solitary trees. Limber pine is also a component of multi-tree islands with Engelmann spruce and subalpine fir. In all likelihood, limber pine frequently occurs under similar conditions in other Front Range ATE communities and may be more widespread in occurrence and functional role in the ATE than previously known. Limber pine may increase in relative abundance in the ATE with climate warming, due to its drought tolerance and to directed seed dispersal by Clark’s nutcrackers. Further work is indicated to examine the influence of wind and microclimatic conditions in shaping the species composition of ATE krummholz communities in RMNP. Understanding limber pine’s ability to respond to changing climate may have important management considerations as white pine blister rust continues to spread within RMNP.

Understanding community-level variation in species composition in the ATE as well as interactions among tree species may help refine predictions for treeline response to changing climate. Treeline response may be dependent on the availability of suitable microsites, species-specific niche requirements for seedlings, and facilitative interactions among species. Moisture availability based on snow distribution may be an important determinant of the ability for tree species to move upslope with increasing temperature. Future research should also consider differences in seed dispersal mechanisms as an influence in the ability of trees to colonize in the ATE.