Concordant community similarity patterns across functional groups in subarctic plant assemblages

Abstract

In light of increasing threats to global biodiversity, identifying the patterns in and drivers of variation in species composition along the environmental gradients in understudied regions is crucial for understanding ecosystem vulnerability and resilience. Terrestrial plant communities of interior Alaska are dominated by three major groups of primary producers—vascular plants, mosses, and macrolichens. Recent work has demonstrated broad‐scale positive correlations in species richness patterns among these three functional groups across scales in this region. However, the conspicuous and fundamental differences in reproductive strategies, dispersal ability, and physiological adaptations among these functional groups prompted us to investigate how community composition and species turnover vary across environmental gradients among these disparate groups. We hypothesized that species turnover would be greater in vascular plants than for spore‐producing cryptogam functional groups across gradients of elevation, vegetation type, and distance due to underlying differences in dispersal abilities and the temporal and spatial resolution of habitat preferences among these disparate groups. To address these issues, we compiled a uniquely comprehensive species composition dataset in interior Alaska utilizing a multi‐stage systematic design. We analyzed community similarity using Morisita‐Horn index of multi‐community similarity among all pairs of vegetation types and elevation bands for each functional group and used regression analysis to quantify the rate at which compositional similarity decayed with geographical distance among groups. Our study reveals consistently higher compositional variation (lower similarity) among vascular plants as compared to mosses and macrolichens across landscape gradients and spatial scales in interior Alaska. We also found that the decay of similarity with distance was less for both of the cryptogam groups than for vascular plants. Taken together, our results suggest that differences in functional group dispersal abilities, in combination with varying abilities to utilize fine‐scale temporal and spatial habitat variation, result in reduced turnover among mosses and macrolichens as compared to vascular plants both along landscape gradients and across distance in interior Alaska.

Introduction

Understanding the drivers of variation in species composition along gradients is crucial for predicting ecosystem response, given what appear to be ever‐increasing threats to plant community composition and structure globally (Pereira et al. ). Vascular plants, mosses, and macrolichens are major components of most subarctic vegetation communities and often exhibit divergent responses to environmental conditions based on their members' varying responses to resource limitation, disturbance regimes, herbivores and pathogens, and other factors (e.g., Tilman and Lehman , Flinn et al. ). However, several studies have nonetheless shown positive correlations in compositional patterns and/or richness among functional groups along the major environmental gradients (e.g., Pharo et al. , Dynesius and Zinko , Roland et al. ).

Dispersal ability is a crucial factor influencing spatial patterns of community composition and species turnover in many groups of organisms at extensive spatial scales (Qian and Ricklefs , , König et al. ), particularly in high latitudes (Freestone and Inouye ). For example, the high degree of similarity of lichen communities across the Eurasian Arctic relative to mosses and vascular seed plants was consistent with highly effective wind‐dispersal in lichens (Lenoir et al. ). The generally greater dispersal ability of spore‐producing primary producer groups as compared to seed plants is one factor that may act to decrease the degree of spatial turnover among macrolichen communities (Lenoir et al. ). Other studies at global, continental, and regional scales have reported analogous findings of beta diversity being negatively correlated with dispersal ability (e.g., Nekola and White , Qian , König et al. ).

Additionally, patterns in environmental variation and continuity interact with dispersal constraints and influence species establishment success and longevity (e.g., Nekola and White , Barbé et al. ). For instance, several studies have noted that the greater dispersal abilities of cryptogam species may be counteracted by relatively strict environmental tolerance requirements during establishment, the fine‐scale of dispersal events, or highly specific habitat requirements of cryptogam species (Pharo and Zartman , Schei et al. , Barbé et al. ).

The landscape of interior Alaska contains conspicuous and sometimes surprising patterns in the diversity of vascular plants, mosses, and macrolichens across abiotic and biotic gradients at multiple scales (e.g., Hollingsworth et al. , Roland and Schmidt , Roland et al. ). For example, species richness of all three functional groups is generally highest in the topographically heterogeneous alpine areas (and thus in sparse, open alpine tundra plant communities), and lowest in the more productive, low elevation boreal forest areas (Roland and Schmidt , Roland et al. ). Positive correlations in richness across these functional groups have also been found with increasing elevation, although the influence of other environmental covariates associated with richness revealed differences both among functional groups and across spatial scales (Roland et al. ). For example, although increasing soil pH was positively associated with vascular plant and moss richness, and negatively associated with macrolichen richness at the plot‐scale, all three groups were nonetheless positively inter‐correlated at the meso‐ and regional scales (Roland et al. ). Positive inter‐correlation among functional group richness along the gradient of topographic elevation with concomitant divergent responses to site‐level covariates suggests that the community composition and turnover patterns at multiple scales are likely to be complex.

Discerning whether patterns in community composition and similarity among different functional groups are consistent or variable across environmental gradients is important as natural resource and land managers seek to learn how stressors such as climate warming may affect compositional patterns in biological communities with cascading ecosystem consequences (e.g., Wookey et al. ). Additionally, the identification of areas most at risk to biodiversity loss is essential to conservation area managers (Bestelmeyer et al. ). To better inform our understanding of these issues, we conducted a multi‐scale evaluation of ground‐layer community similarity patterns for vascular plants, mosses, and macrolichens utilizing a spatially extensive plot dataset encompassing a wide cross section of site types within a diverse, naturally regulated subarctic landscape that yields unbiased statistical inference over a large region (e.g., Roland et al. , , Roland and Schmidt ). The goals of this research were to evaluate whether patterns in community composition and species turnover along gradients in three disparate functional groups were consistent or different by comparing and contrasting variation in community similarity across the landscape, and by comparing the rate at which compositional similarity decays with distance at the regional scale among these groups. Specifically, based on previous work elucidating concordant species richness patterns in our study area (Roland et al. ), we hypothesized there would be lower species turnover along the environmental gradients and thus greater compositional similarity across scales for the cryptogam groups relative to vascular plants. Additionally, we hypothesized that the greater dispersal ability of spore‐producing moss and macrolichen groups would lead to lower rates of decay in compositional similarity in relation to increasing distance at the landscape scale as compared to vascular plants, which have greater dispersal limitation due to generally heavier propagules.

Methods

Study area

The study area is located within Denali National Park and Preserve (henceforth Denali) in south‐central interior Alaska, mostly north of the crest of the Alaska Range, and includes 1.3 million‐hectares with a center near 63°20.3′ N, 150°49.6′ W, spanning 202 km of longitude and 190 km of latitude (Fig. ). The area contains isolated Alaska Range peaks >2000 m above sea level along the southern margin, grading into foothills and uplands and then into the lowland basin of the Tanana River to the north. The study area is described in detail by Roland et al. () and Roland and Schmidt ().

Overview map of the Denali National Park and Preserve study area in interior Alaska, and the location of the 43 minigrids sampled. The systematic layout of the maximum of 25 plots sampled per minigrid, and variation in terrain elevation (m) is also shown. White spaces within the northwest corner of the study area are lakes.

Field data

Field data come from 976 vegetation monitoring plots surveyed according to a two‐stage systematic grid design where we sampled 43 minigrids consisting of 25 plots each (five rows of five plots, 500 m apart) arranged on a larger grid of 10–20 km spacing (Fig. ). Wetlands (including pond margins and marshes) or other specific habitat types were not excluded from our study, occurring in the plot network in proportion to their existence on the landscape in keeping with a randomized systematic grid sample. Some minigrids had fewer than 25 plots sampled because particular plot locations were inaccessible (e.g., in a pond or on a cliff).

At each plot, we recorded species composition for all vascular plants, mosses, and terricolous macrolichens in a set of four 1‐m² quadrats arranged on bisecting meter tapes. Our estimate of site‐level species composition was the count of unique species occurring in this 4‐m² area. We included only terrestrial taxa native to Alaska in our analyses given we did not encounter non‐native species in our sampling. Taxa were recorded only to the species‐level, except for six vascular species for which we accepted reports of two subspecific taxa each. Occasionally generic determinations were necessary due to lack of sufficient voucher specimen or extreme difficulty in identification of particular groups (i.e., generic determinations make up 1% of our nonvascular species diversity dataset, with genera pooled at the plot level). For data analyses, we excluded strictly crustose, leprose, medi‐, and pin lichens from our species lists, retaining only terrestrial macrolichens found growing on the ground, rocks, and decaying woody material <50 cm aboveground level (hereafter macrolichens). More details on the taxonomy, nomenclature, and classification of species used to develop species lists can be found in Roland et al. ().

Because we were most interested in the variation in general similarity patterns among the three primary producer functional groups at multiple scales, we chose to employ an inclusive grouping method. Thus, the relatively few spore‐producing ferns and fern allies in our vascular dataset (n = 28 out of 418 species) were grouped with the vascular seed plants (Nekola and White , Qian ). Similarly, many individual species in our moss and macrolichen species pool may or may not engage in reproduction primarily via dispersed spores. The need for consideration of multiple methods of dispersal has been recognized (Ozinga et al. ), but we chose to maintain the general functional groupings for this broad‐based comparative study, acknowledging that there is considerable variation within each of the functional categories.

Data analysis

Estimating multi‐scale compositional similarity

To assess overall compositional similarity of the total flora along the landscape gradients and various scales, we first approximated the gradients into landscape segments derived as the ten primary vegetation types and seven elevation bands in our sample (Appendix S1: Table S1). We then used the SpadeR package (Chao et al. ) in R (R Core Team ) to compute the Morisita‐Horn index of multi‐community similarity among all pairs of vegetation types and elevation bands for each functional group independently. The Morisita‐Horn similarity index is based on a two‐way probabilistic approach that takes into account species present but not detected in the samples, thereby reducing bias (Chao et al. ). As compared to the more commonly employed Jaccard and Sørensen indices, the Morisita‐Horn index is capable of handling the multiple assemblages of varied sample sizes represented in our data (see Appendix S1: Table S1). The Morisita‐Horn index is also sensitive to dominant species (Chao et al. ), and thus, we felt it prudent to retain even relatively rare species in our dataset. Our input data matrices were of raw species frequencies from all plots within each vegetation type or elevation band, with a separate matrix constructed for each functional group. We employed the SimilarityMult function to compute the estimated pairwise Morisita‐Horn index based on those species' relative frequencies among the multiple groups compared, utilizing a bootstrap approach of 500 simulations to measure compositional similarity.

To determine vegetation type, we classified plots in the field to level 4 of the Alaska Vegetation Classification (Viereck et al. ) and then condensed the 100 types obtained from field determinations into 11 broad types approximating vascular growth form dominance (tree, shrub, dwarf shrub, open; Appendix S1: Table S1) and taxonomic composition of dominant species. We did not calculate similarity for the graminoid‐herbaceous vegetation type due to a small sample size of plots in that type. For help in visualizing similarity among vegetation types, we input a matrix of pairwise Morisita‐Horn similarities converted to dissimilarities (one minus the pairwise similarity divided by two) into a hierarchical cluster analysis using the complete linkage method in the statistical program R (R Core Team ) to iteratively join the two most similar vegetation types.

We applied parallel methods as above to assess overall compositional similarity at the meso‐spatial scale, using minigrids as discrete community units to organize the data. In these analyses, the similarity index was a function of how many species were unique to one minigrid in relation to how many species were shared by both grids combined. For these analyses, our input data matrix was of species frequencies (raw incidence data) for all measured plots in each of 43 minigrids, with a separate matrix constructed for each functional group. We again used a bootstrap approach to measure Morisita‐Horn similarity, utilizing 250 simulations of this larger dataset. The sample sizes varied slightly per functional group (Table ) due to some plots containing zero group members (i.e., occasionally no macrolichens were present).

Overall means of Morisita‐Horn similarity index values across categories of analyses (scales) and regression statistics for linear models estimating distance decay of similarity for each functional group between 43 minigrid study areas in Denali National Park and Preserve, Alaska

Functional group	Species richness	N	Average pairwise similarities	Regression coefficients
Elevation bands	Vegetation types	Minigrids	Intercept	Slope	R 2	P
Vascular plants	418	975	0.66	0.46	0.53	0.694	−0.003	0.11	<0.001
Mosses	271	952	0.82	0.67	0.67	0.769	−0.002	0.09	<0.001
Macrolichens	209	917	0.81	0.66	0.68	0.785	−0.002	0.07	<0.001

²Total species richness measured in four 1‐m² quadrats per n 200‐m² plots.

For both landscape (vegetation type and elevation bands) and meso (minigrid)‐ scales, we sought to compare overall similarity among the functional groups. To avoid treating similarity indices derived from pairwise samples as independent observations (e.g., Qian ), we used an exact permutation test estimated by 5000 Monte Carlo replications in the perm package (Fay and Shaw ) in R (R Core Team ) to derive these estimates. We determined a difference in mean similarity between functional groups existed only if the upper limit of the 95 percent confidence interval on the P‐value was below 0.05. To further assess functional group similarity across our scales of inquiry, we applied the Mantel test from the ecodist package (Goslee and Urban ) in R (R Core Team ) to derive a Pearson's correlation coefficient between each set of functional group similarities for elevation bands, vegetation types, and minigrids both accounting for geographic distance (partial Mantel) and irrespective of it.

Estimating meso‐scale distance decay of similarity

We compared similarity of the species composition among all pairs of 250‐ha minigrids, which often encompass a diversity of vegetation types and elevation bands, to test our hypothesis of the differing distance‐decay relationships among functional groups at the meso‐scale. Using the latitude and longitude of the central point of each pair of minigrids to calculate the Euclidian distance between them, we then estimated distance decay with a linear regression model of the pairwise comparisons, plotting similarity as a function of distance (Nekola and White ). The distance decay of similarity or rate of community similarity decay with geographic distance (Nekola and White ) has been used extensively as a comparative measure of the influence of dispersal on vegetation community turnover across space. Although some studies employing this method have included climatic or topographic variables in their distance‐decay model (i.e., Qian and Ricklefs , Lenoir et al. ), we chose to omit them here because we were interested in comparing the relative variation among the three functional groups, and our species lists were generated at the exact same plots within one sampling design, where members of each functional group experienced identical environmental conditions. Additionally, those studies concluded that patterns of distance decay were similar whether controlling for environment or not (Lenoir et al. ), and especially so in the far north (Qian and Ricklefs ).

To test the hypothesis that the slope of the distance decay of similarity for functional groups dominated by spore‐producing dispersers (i.e., mosses and macrolichens) would be less steep than for vascular plants, we compiled the similarity and distance values from all groups into one dataset and introduced an interaction term of functional group into our linear model of distance decay. We then used nonparametric bootstrapping in the “boot” package (Canty and Ripley ) in R (R Core Team ) to obtain 10,000 estimates of the regression coefficients, generating confidence intervals for the slope of the interaction term. When the confidence interval did not contain zero, we concluded a difference in the relationship between functional group similarity and distance existed.

Results

Similarity in species composition among elevation bands

There were conspicuous differences among the three functional groups in the relative similarity of the flora among elevation bands (Fig. ). Specifically, similarity index values for the vascular plant composition matrices between the highest elevation band (>1100 m) and each of the three lowest elevation bands (<350 m, 350 to 500 m, and 500 to 650 m) were 0.17, 0.20, and 0.27, respectively—roughly half of similarity index values for mosses (0.48, 0.48, 0.58) and macrolichens (0.45, 0.45, 0.55), respectively, for these same comparisons (Fig. ; Appendix S1: Tables S2–S4). The elevation bands with the consistently highest mean similarity index values across all comparisons were the three mid‐elevation bands for all functional groups, demonstrating highest overlap in species composition in the middle of the elevation gradient vs. the ends (Fig. ; Appendix S1: Tables S2–S4). The elevation band with the lowest mean similarity index value across all comparisons was the >1100 m band for all three functional groups, highlighting compositional discontinuity of the rich alpine flora with the more depauperate flora of lower elevations. The high overall mean similarity across all pairs of elevation bands for mosses (0.82) and macrolichens (0.81) contrasted significantly with the relatively low value for vascular plants (0.66; P < 0.05; Table ). Mantel's test revealed highly significant correlations in similarity among all pairs of functional groups (Table ). Covariation in similarity of analogous elevation band moss and macrolichen communities was highest of all the pairs (Mantel r = 0.992).

Matrix of estimated Morisita‐Horn similarity indices for (A) vascular plant, (B) moss, and (C) macrolichen floras of the six elevation bands shown as compared to the flora of those groups in other elevation bands across Denali National Park and Preserve, Alaska.

Pairwise matrices of Mantel's r correlations of functional group similarity across categories of analyses (scales)

Functional group	Elevation bands	Vegetation types	Minigrids
Vascular plants	Mosses	Macrolichens	Vascular plants	Mosses	Macrolichens	Vascular plants	Mosses	Macrolichens
Vascular plants	1	0.981	0.979	1	0.937	0.914	1	0.828	0.73
Mosses		1	0.992		1	0.813	0.809	1	0.67
Macrolichens			1			1	0.704	0.64	1

Notes

All correlations had P < 0.001. The lower diagonal in the minigrid matrix shows the partial Mantel's r when accounting for the linear effects of geographic distance between minigrid centers. We did not compute partial Mantel's r for elevation bands or vegetation types since those groupings were not inherently related to geographic distance.

Similarity in species composition among vegetation types

In a pattern similar to, but more pronounced than that observed for the elevation strata, the minimum similarity index values of the vascular plant composition among all pairs of vegetation types (0.04, between barren sites and sedge scrub or black spruce) were an order of magnitude less than that observed for the macrolichens (0.28, between barren sites and broadleaf) and mosses (0.18, between barren sites and black spruce; Appendix S1: Tables S5–S7). Similarly, the mean Morisita‐Horn similarity index value across all possible vegetation type comparisons (n = 45 pairs) was 0.67 for mosses, 0.66 for macrolichens, and 0.46 for vascular plants (Table ) indicating significantly greater compositional variation among vegetation types for vascular plants relative to the other two functional groups (P < 0.05). The highest similarity coefficients were observed between vegetation types that share ecotones and intergrade on the landscape, such as black spruce woodlands and sedge‐scrub areas, which are both underlain by deep organic soils on shallow permafrost (similarity coefficients of 0.86, 0.90, and 0.89 for vascular plants, mosses, and macrolichens, respectively; Appendix S1: Tables S5–S7). Similarly, the two alpine tundra types (Dryas and heath‐Salix) that intergrade along the moisture gradients shared similarity coefficients of 0.86, 0.87, and 0.93 for vascular plants, mosses, and macrolichens, respectively. The lowest mean similarity coefficients for all functional groups were for the barren and Dryas‐graminoid tundra types (Appendix S1: Tables S5–S7), indicating substantial compositional discontinuity between these well‐drained primarily alpine communities and other vegetation types across functional groups. Mantel's test revealed highly significant correlations between all pairs of functional groups, though the correlations were weaker than when plots were stratified by elevation band (Table ). In contrast to comparing across elevation bands, similarity in analogous vegetation type moss and macrolichen communities covaried the least strongly of the pairs (Mantel r = 0.813).

Ordering of the different functional group responses by cluster dendrogram helped clarify the relationships in floristic similarity among the major vegetation types (Fig. ). Remarkably, similarity of all three functional groups clustered in the same first‐order pairings of dwarf shrub‐dominated (heath‐Salix and Dryas) and barren vegetative types, types generally occurring in cold, acidic, organic‐rich soils in permafrost situations (sedge, birch‐ericaceous, black spruce), and more productive, closed‐canopy forest and tall shrub‐dominated types in warmer topographic positions with mineral soil substrates (alder, white spruce, broadleaf, and willow). Vascular plant communities displayed the greatest dissimilarity among related clusters, particularly considering the barren and willow vegetation types. For mosses, the willow dominated vegetation type had decidedly greater similarity to the white spruce forest type than found for macrolichens or vascular plants (Fig. ).

Cluster dendrogram illustrating the dissimilarity of vegetation type pairs when considering the estimated Morisita‐Horn similarity index for (A) vascular plant, (B) moss, and (C) macrolichen floras of the ten vegetation types occurring within 984 plots in Denali National Park and Preserve, Alaska. The dissimilarity, equal to 1 minus the similarity of a given pair divided by 2, is given on the x‐axis such that branches of the dendrogram appearing farther left on the x‐axis divide communities that are more dissimilar (and thus less similar).

Similarity in species composition among minigrids

Patterns in the relative similarity of the flora at the meso‐scale varied by functional group in a similar manner as for elevation bands and vegetation types. Across all possible minigrid comparisons (n = 903 pairs), vascular plants exhibited the lowest degree of similarity with a mean Morisita‐Horn index of 0.53, significantly lower (P < 0.001) than the mean similarity index value for mosses and macrolichens (0.76 and 0.68, respectively), which did not differ significantly (P = 0.06; Table ). Mantel test for covariation in the similarities between functional groups across minigrids revealed significant correlations between all pairs of functional groups (Table ). In a pattern mirroring vegetation types, similarity in analogous minigrids showed the greatest covariation between vascular plant and moss communities (r = 0.828) and the least correlation between moss and macrolichen communities (r = 0.67). Accounting for the linear effects of geographic distance slightly lowered all correlations (Table ).

Meso‐scale distance decay of similarity

Euclidean distance between minigrid centers ranged from 9.8 km to 152.5 km. All three functional groups exhibited a significant decline in compositional similarity with increasing distance, although much of the variance was left unexplained (r² < 0.12; Table ). The rate of distance decay varied by functional group with vascular plants exhibiting a rate of decline in similarity with distance 1.6 times greater than the functional groups dominated by spore‐producing dispersers (mosses and macrolichens; Fig. ). The rate of distance decay of similarity did not differ significantly between mosses and macrolichens at this scale (Table ). Thus, on average, minigrids that are 100 km apart decline in similarity by 0.28 for vascular plants and 0.17 for mosses and macrolichens.

Estimated Morisita‐Horn similarity index for (A) vascular plant, (B) mosses, and (C) macrolichen floras for the pairwise comparison of 43 minigrid study areas shown as a function of distance between the center points of those pairs in Denali National Park and Preserve, Alaska. Regression lines show a linear model of the relationship with the shaded area representing a 95% confidence interval.

Discussion

Our study reveals consistently higher compositional variation among vascular plants as compared to macrolichens and mosses across multiple landscape gradients and spatial scales in interior Alaska. For example, similarity indices of both the macrolichen and moss floras in the lowest vs. the highest elevation band in our study area were about twice that of similarity indices of the vascular plant flora. Likewise, the minimum similarity value between all pairs of vegetation types for either macrolichens or mosses was over twice the minimum value for vascular plants. Additionally, mean similarity values across all pairs of elevation bands, vegetation types, and minigrid study areas indicated greater homogeneity in the species composition of moss and macrolichen communities as compared to vascular plant communities from data collected using an identical sampling design. We also found that the decay of similarity with distance was less for both of the primarily spore‐producing groups than for vascular plants, suggesting there is less turnover among the groups with geographic distance as well. Taken together, these results confirm our initial hypothesis of reduced turnover among mosses and macrolichens as compared to vascular plants both along the landscape gradients of habitat and elevation, and across scales in our extensive study area.

Several inter‐related factors likely contribute to the patterns in relative similarity among functional groups we observed. These causal factors include variation among the functional groups with respect to dispersal abilities, differences in the spatial and temporal grain sizes of habitat requirements among functional groups, and in their ability to utilize dormancy during unfavorable periods.

Relative differences in dispersal ability influence community assembly

The greater dispersal ability of spore‐producing cryptogam groups in relation to the vascular plants contributes to the higher compositional similarity among moss and macrolichen communities as compared to vascular plant communities across the multiple scales of our inquiry. Cryptogams' ability to produce spores likely reduces barriers to colonization of suitable sites for these taxa relative to vascular plants, allowing them to more effectively find and fully utilize potential habitat, having a homogenizing influence on community composition across scales. That is, the species pools of moss and macrolichen communities may be more similarly distributed among landscape segments due, in part, to their dispersal capability (e.g., Hylander , Lonnell et al. ).

Further, the rate of distance decay of community similarity for the spore‐producing cryptogam groups in Denali was less than for vascular plants, as has been observed elsewhere across global (König et al. ), continental (Nekola and White , Qian and Ricklefs , Qian , Lenoir et al. ), and regional scales (in part Nekola and White , 2000 km). Our results confirm these findings for interior Alaska, also demonstrating this effect at a substantially shorter set of geographical distances (maximum of 150 km) than previously tested. Specifically, vascular plants across our 150 km study area exhibited a distance decay of similarity 1.6 times greater than mosses and macrolichens, a difference similar to the ratio detected for those same species groups across 2000 km of boreal North America (Nekola and White ). This finding supports the hypothesis that the spore‐producing functional groups contribute to a greater perfusion of these species throughout available habitats, and that dispersal limitation may contribute to greater variation in vascular plant community composition along a geographical gradient relative to the cryptogam groups.

Distance decay of similarity may be particularly important in high‐latitude study areas, where the cumulative effect of such dispersal filtering over time may be magnified due to the region's ecological history and relatively frequent resetting of species composition patterns (i.e., repeated widespread disturbance and subsequent recolonization caused by glacial cycles at multiple scales; Qian and Ricklefs , Roland and Schmidt , Roland et al. ). This cumulative dispersal filtering likely has a differential effect when considering smaller‐scale segments of the landscape that also experience disturbance events at shorter‐time scales. For example, black spruce and mixed forests common at the low elevations in our study area have developed in the presence of relatively frequent fire disturbance, which depending on severity, may or may not expose the mineral soils often necessary for successful seed or spore colonization. Consequently, many of the vascular plants common in these areas rely primarily on vegetative resprouting to persist (i.e., Populus tremuloides, Rubus chamaemorus; Reilly et al. ), therefore minimizing the impact of relative differences in dispersal ability between functional groups (and even within functional groups; Nelson et al. ) in this vegetation type. In contrast, the most frequent disturbance event in the high elevations dominated by tundra vegetation is erosion on steep slopes, where dispersal ability is more likely to play a leading role in determining species composition.

Functional group differences in spatial and temporal habitat utilization

In addition to the influence of variable dispersal abilities among groups, the generally smaller size of cryptogams and finer spatial scale at which they meet their habitat requirements relative to vascular plants (e.g., Alpert and Oechel ) likely contributes to the relative differences in similarity patterns we observed. The smaller cryptogams can inhabit suitable microsites in smaller patches within a given vegetation mosaic relative to vascular plants, which generally have coarser‐scaled habitat requirements (e.g., Vittoz et al. ). In other words, the ability to utilize habitat at a fine grain size could result in less differentiation among the moss and macrolichen communities that develop in differing sites due to the relatively larger amount of habitat variation within a site as compared to variation between sites.

Indeed, numerous studies have found that microenvironment and micro‐topography are important factors governing moss and macrolichen distribution and richness (Alpert and Oechel , Holt et al. , Benscoter and Vitt , Økland et al. ). For example, in Denali, moss and macrolichen richness was greatest in the highest elevation band (>1100 m), where varied topography and substrate serve to expand potential fine‐scale habitat (Roland et al. ). Although heterogeneity of niche‐space may also aid in vascular species establishment, the lower similarity of high elevation vascular communities as compared to moss and macrolichen communities also indicates that the number of niches available potentially favor cryptogam species (i.e., for macrolichens in Denali, xeric habitats comprised of minimally decomposed rock; Roland et al. ). Additionally, along the elevational gradient in particular, vascular plants as a group are reaching their physiological limits due to a range of factors not as limiting for cryptogams. Specifically, mosses and lichens possess a greater capacity to tolerate habitat heterogeneity over time via poikilohydry, the ability to quickly utilize dormancy during unfavorable periods within the growing season, and proficiency at performing carbon acquisition and other life functions during much shorter favorable periods (such as prior to leaf‐out under deciduous woodland canopy) than vascular plants.

Poikilohydry (lacking complex water‐controlling mechanisms) contributes to lichen and moss species' ability to achieve, as well as to recover from, dormancy more quickly (and repeatedly) than even the most desiccation‐tolerant vascular plants (Proctor , Proctor and Tuba ). The unique ability of cryptogams to attain temporary physiological dormancy when conditions are not suitable for carbon acquisition, photosynthesis, or respiration likely acts to increase the habitat range for these organisms relative to more physiologically active vascular plant species (e.g., Proctor and Tuba , Street et al. ).

Cryptogams also possess the capacity to more quickly initiate photosynthesis, as compared to vascular plants, when light and moisture become available at very short intervals (i.e., utilization of sunflecks; Kubásek et al. ). Thus, frequent and abundant cryptogam species of the middle elevations and/or mesic vegetation types (i.e., generalist species such as Hylocomium splendens) are more likely to successfully occupy even marginally suitable microsites available on the extreme ends of those environmental gradients of moisture and light (Lee and La Roi , Bruun et al. , Vittoz et al. ), thereby increasing similarity of those groups across landscape segments (i.e., vegetation type, elevation band). The ability to respond more rapidly to the advent of favorable conditions also confers on cryptogams the ability to utilize periods such as prior to leaf‐out in spring or after leaf senescence in fall, when vascular plants are typically dormant, to grow and reproduce, thereby diminishing the effects of temporal competition for light (e.g., Johansson and Linder , Kershaw and MacFarlane , Street et al. ). Even during mid‐winter warming episodes, which are becoming more frequent with climate change (e.g., Callaghan et al. ), both mosses and lichens may capitalize on growth opportunities not available to vascular plants (Kappen , Bjerke et al. ), although the two groups exhibit differential resistance to subsequent freeze damage (Bjerke et al. ).

Utility of vegetation types for approximating cross‐taxon community similarity

Our analyses of overall compositional variation among vegetation types (see Fig. ) illustrate how patterns in community composition for the entire flora (vascular plants, mosses, and macrolichens) are rather remarkably well‐represented by the major structural vegetation types of tundra, shrub, and forest. For example, the vegetation types associated with the highest elevations (the tundra types of barren, heath‐Salix, and Dryas) formed the primary bifurcation in similarity for all functional groups. Thus, as expected, strong filters on community assembly operating on all three functional groups separate the major habitat types in our area. It is notable that, even in vegetation types defined by variation in vascular plant species (Viereck et al. ), there was still less turnover in the flora of the moss and macrolichen functional groups relative to the vascular plant community. Further, both moss and macrolichen community similarity covaried more strongly with vascular plant similarity across vegetation types than with each other, suggesting that beyond abiotic site conditions, species‐species relationships may also play an important role in community development (e.g., Carlson et al. ).

Overall, the ordering of vegetation types by similarity is broadly concordant among functional groups, perhaps suggesting that one may use the species turnover of one functional group across a gradient of interest as surrogate for estimating the species turnover of another (e.g., Pharo et al. , Su et al. ). However, although correlations in similarity patterns between functional groups may exist, a heterogeneous distribution of specific habitats within landscape segments of the same type is imperative for particular functional groups (e.g., woody debris; Negi and Gadgil ). For example, in our dataset, there was a subtle but unique clustering of moss compositional similarity (as distinct from vascular plants or macro lichen results), wherein the most similar communities occurred in the two broadleaf deciduous‐dominated vascular vegetation types (alder shrub and broadleaf/mixed forest). We interpret this as evidence for the critical importance of litter type and abundance in influencing moss community development (Startsev et al. ), whereas this particular habitat limitation likely has less effect on the community assembly of the generally taller (vascular) or already substrate‐restricted (macrolichen) functional groups in those types.

Our results demonstrating functional group community similarity patterns broadly correspond to and amplify the patterns in functional group species richness described for the same study area in Roland et al. (). Indeed, these results are closely linked and thus reinforce each other and reveal fundamental attributes of functional group community assembly at several scales that allow for greater understanding of the ordering of primary producer diversity across interior Alaska. This information will be crucial as land managers assess how stressors such as climate warming and resulting vegetation change may affect compositional patterns in biological communities with cascading ecosystem consequences. For example, identification of critical transitions among communities where species turnover is high gives land managers a goal for targeted preservation and management (e.g., limiting impacts) of such areas. Further, as certain site types undergo transition from one vegetation type to another in response to changing conditions (e.g., alpine tundra is invaded by taller shrubs and trees; see Roland and Stehn ), this enriched understanding of components of composition and diversity will assist land managers in assessing attendant changes to multiple groups of organisms and ecosystem pattern that will be occurring simultaneously. For Denali, we conclude that the tundra habitats in our area are the most compositionally distinctive communities for vascular plants, mosses, and macrolichens, in addition to being the most species‐rich for each group (Roland et al. ).

Acknowledgments

We thank dozens of field staff for their exemplary data collection efforts, but especially J. Walton and P. Nelson for their longevity with the project. We acknowledge E. Debevec, A. Southwould, and D. Wilder for their expertise in assisting with data infrastructure. We thank M. Carlson and two anonymous reviewers for helpful comments on an earlier version of this manuscript. The US National Park Service funded this work, with programmatic support provided by Denali National Park and Preserve and the Central Alaska Inventory and Monitoring Network. The conclusions presented are the authors' own and do not necessarily represent those of the US government.