1. Introduction
Ecologists have long been interested in the distribution and assembly of plant communities. Broad trends and patterns have been described from local to global scales including well-known trends of decreasing species richness with increasing latitude and elevation [
1,
2,
3]. Many of these trends and patterns have been documented in the temperate zone, but less so in the tropics, and even less in tropical montane forests due to their limited accessibility [
4,
5,
6,
7]. Recently, interest in investigating distribution and assembly patterns of plants in montane forests along elevational gradients has grown as these areas are home to at least a third of all terrestrial plant species and will likely show large effects from global warming [
5,
7,
8,
9,
10].
Montane forests are an ideal system to study gradients in species composition because they eliminate confounding regional scale effects. These forests are also among the most species rich in the world, but consistently remain understudied compared to lowland tropical forests [
7,
8,
11,
12,
13,
14,
15]. Montane forests are frequently immersed with clouds and are recognized for their low canopy height, multi-stemmed trees, and high epiphyte abundance [
16,
17,
18]. They are also known to possess high levels of endemism, due in part to the unique environmental conditions and topography where they are found [
7,
12,
15,
19].
Traditionally, montane forest vegetation has been described using floristic inventories that calculate taxonomic diversity metrics such as species richness, Shannon–Weaver diversity [
20], and Simpson’s Index [
5,
21,
22,
23,
24]. These metrics have illuminated many globally consistent patterns of species distributions along elevation that have been useful in understanding the composition and diversity of plant species [
25,
26,
27]. For example, alpha diversity has been shown to decrease with increasing elevation [
19,
28,
29,
30], but has also been found to have a hump-shaped pattern [
6,
11,
31]. Interestingly, a number of abiotic and biotic factors such as temperature, precipitation, cloud cover, soil nutrients, light availability, and competition have been correlated with shifts in community composition along elevation [
16,
28,
29,
30,
32]. Despite the success of these metrics, they alone struggle to describe montane forests’ biodiversity much less community assembly and species co-occurrence patterns along elevational gradients [
10,
33].
Community phylogenies have become increasingly important for providing additional information regarding the diversity and community assembly of forests beyond that which can be gained from analyzing species diversity and composition [
33,
34,
35,
36,
37,
38,
39]. By merging understandings of ecology, evolution, and biogeography in plant communities, these phylogenies can reveal aspects of biodiversity that are not normally observable like linking phylogenetic diversity and dispersion to determine assembly mechanisms of forests [
38,
39,
40,
41,
42,
43]. Unlike taxonomic diversity metrics that use a nomenclatural approach, phylogenies allow an understanding of how communities evolved through time and offer further insights into historical and current processes contributing to diversity [
10,
33,
44,
45], hence, two communities could have the same species diversity, but different phylogenetic diversity [
36]. To gain a better understanding of how montane forest communities are assembled, DNA barcodes can be used to construct phylogenies for community-level analyses. Phylogenies built using DNA barcodes are able to provide estimates of evolutionary distances and relationships between species within the phylogenies [
24,
42,
46]. DNA barcodes used in the construction of plant phylogenies commonly include the phylogenetically conserved coding region,
rbcL, combined with the more rapidly evolving gene region,
matK [
47,
48,
49]. Given the success of using DNA barcodes to build tropical forest community phylogenies [
37,
47,
50,
51], they can be established for less studied montane forests to highlight patterns of community assembly and structure not previously seen using taxonomic diversity metric values.
Community phylogenies can be used to investigate the assembly of plant biodiversity, which is thought to be due to a diverse array of abiotic and biotic mechanisms that filter species composition at both regional and local scales [
52,
53]. Community assembly patterns of phylogenetic relatedness typically fall into three categories: Random, clustered, or overdispersed based upon some attributes (e.g., relatedness, traits) of the species [
34]. These categories focus on the rationale that some community assembly mechanisms favor co-occurrence of closely related species (phylogenetically clustered), whereas others favor co-occurrence of distantly related species (phylogenetically overdispersed) [
35,
42,
49,
51,
54]. These categories can be used as a proxy to suggest underlying community assembly mechanisms [
35,
37,
44]. Phylogenetic clustering has been hypothesized as evidence for the influence of habitat filtering (abiotic-driven processes) [
37,
46,
47,
54,
55] or performance differences [
56]. For phylogenetically overdispersed communities, biotic interactions (e.g., niche differences) are often hypothesized as important for local community assembly (e.g., competition) [
33,
37,
45]. Often, however, the phylogenetic patterns underlying ecological and evolutionary mechanisms associated with composition of many plant communities remain unknown or in dispute [
57]. For example, along elevation, some studies have found patterns of phylogenetic overdispersion [
10,
22,
38,
58], while others have found contrasting patterns of phylogenetic clustering [
24,
57]. Together, community phylogenetics and complementary species diversity metrics have the ability to detect important patterns of distribution, assembly, and structure of tree species along elevation in montane forests.
The goal of this study was to quantify patterns of tree species diversity and phylogenetic community assembly along an elevation gradient in montane forest and investigate potential ecological and evolutionary processes that underlie tree species co-occurrence. Specifically, we asked, 1) how do patterns of tree diversity and community structure change along an elevation gradient? 2) is there a relationship between diversity and structure trends found across elevation? and 3) how does the presence of tree ferns affect the phylogenetic community structure?
3. Discussion
In this study, we investigated patterns of species distribution and structure across an elevational gradient using both taxonomic and phylogenetic metrics. Using multiple taxonomic metrics, we found evidence for species diversity decreasing with elevation (
Figure 2a,b). We also found this pattern with observed phylogenetic diversity (PD), but not with observed mean pairwise distance (MPD; except when tree ferns were excluded) or observed mean nearest taxon distance (MNTD) metrics (
Figure 2c–e). When considering the phylogenetic structure of the tree community, given the gradient of species diversity, we found a non-random structure that was contingent upon the presence of tree ferns. Without tree ferns, lower elevation communities exhibited similar phylogenetic structure as when tree ferns were included (
Table 2;
Tables S3–S8). However, at higher elevations, communities switched from patterns of phylogenetic overdispersion with tree ferns to phylogenetic clustering without tree ferns (
Table 2;
Tables S3–S8). Standardized effects sizes (SES) of MPD and MNTD were not significantly related with elevation, except when two off-trend plots were removed that showed evidence of colliding upper and lower elevation floras (
Figure 4;
Figure S1). Combined, this evidence supports the idea that tree ferns have converged with angiosperms to occupy the same habitat along with an increased filtering of clades at higher elevations.
3.1. Patterns of Taxonomic Distribution
This elevational transect comprised 595 individuals including 36 families, 53 genera, and 88 species (
Figure 1,
Table S1). The forest composition found at Siempre Verde is in general agreement with comparable studies conducted in other montane forest habitats. For example, a previous study found that South American montane forests are typically dominated by species of
Weinmannia, Schefflera, Miconia, and
Myrcianthes [
59]. At the study site, we found that each of these genera, excluding
Schefflera, were found to be among the most diverse within the transect (
Table S1). Actually,
Weinmannia rollottii (
n = 75) is the most common species in the transect. In addition, the number of species (35) and families (19) in the high elevation plots were in-line with similar studies conducted along elevational gradients in the forests at Pasochoa volcano, Ecuador, where the number of species and families in high elevation plots were 32 and 21, respectively [
60]. Furthermore, [
11,
28] found that Aquifoliaceae and Theaceae become more abundant at high elevations, while Melastomataceae is dominant at mid-elevations and Rubiaceae is common at lower elevations, results that match this study’s findings (
Table S1).
Four separate analyses of community composition were performed: Richness, Shannon–Weaver diversity (H’), Simpson’s dominance (D
2), and Simpson’s evenness. Only richness and H’ showed significant correlations with elevation (
Figure 2a,b,
Table 1). Both richness and H’ decreased as elevation increased. This tendency of decreasing diversity has been shown along elevational gradients in different forest types around the world [
7,
28,
29,
30,
61,
62,
63], although, regardless of the trend, we found the highest values for both metrics at mid-elevation (plot six, 2820 m a.s.l.). Other studies along elevational gradients have found similar findings where a plot, not located at the lowest elevation, exceeds all others in diversity [
6,
11,
19,
31]. Cloud cover may be one potential cause for this mid-elevation increase in diversity. Cloud cover is known to saturate montane forests causing a decrease in temperature and an increase in precipitation and overall moisture [
6,
16,
18,
32]. This has led many to refer to plots located where clouds move into the forest as “mid-elevation bulges” as the highest diversity is often seen at these intermediate elevation sites [
6,
7,
64]. It has been hypothesized that at these mid-elevations, a mixture of species from low and high elevations have reached the maximum and minimum, respectively, of their elevational range and have converged on a particular niche that combines the effects of the environment and competition, increasing diversity [
9,
11,
24,
31].
3.2. Patterns of Phylogenetic Distribution
Sequence recovery rates at our study site were relatively high, where we obtained a genetic sequence for ~80% of species located within the elevational transect, and with 72.9% of those having both
rbcL and
matK sequences and 27.1% missing the
matK sequence (
Table S2). This recovery rate is slightly lower compared to similar studies, where in tropical and temperate forests, other studies have successfully sequenced between 85–93% of samples for
rbcL and between 69–75% of samples for
matK [
47,
50,
51,
65]. The higher recovery rate for
rbcL over that of
matK has been shown to be attributable to its shorter length and better capability of sequencing across all angiosperms making it easier to obtain [
37]. In our study, DNA samples were taken from herbarium specimens at Herbario QCA at Pontificia Universidad Católica del Ecuador that had been preserved in alcohol. It is known that alcohol quickly degrades the quality of DNA [
66,
67], which also may have led to the slight reduction in sequence recovery seen here compared to other studies. DNA vouchers should be taken from fresh collections and dried in silica gel until DNA extraction. Currently there is a lack of publicly available barcode sequences for montane plant species. Our collection represents a substantial contribution to public reference databases, as the majority of the species in our study were not accessible for research.
Comparisons between patterns of phylogenetic community structure with and without tree fern species revealed the impact the abundance of these tree ferns had on driving the phylogenetic patterns. Observed values of each of the phylogenetic diversity metrics were tested for correlation with elevation as a proxy for environmental variables known to change with elevation. With and without tree ferns present, PD was significantly negatively correlated with elevation (
Figure 2c), which is expected as this metric is the sum of all the branch lengths in the phylogeny and thus as species richness decreases as elevation increases there are fewer branches in the phylogeny [
38,
68,
69]. However, when tree ferns were excluded, observed values of MPD were also negatively correlated with elevation (
Figure 2d). This suggests that if we exclude tree ferns, communities at higher elevations are made up of less diverse, more closely related species, a commonly found pattern [
24,
70,
71].
With tree ferns included, standardized effect sizes of phylogenetic distance (PD), mean pairwise distance (MPD), and mean nearest taxon distance (MNTD), revealed phylogenetic overdispersion in 14 instances, phylogenetic clustering in three instances, and phylogenetic randomness in all other cases indicating a lack of uniform phylogenetic structure across the elevation gradient (
Table 2). However, without tree ferns, these patterns were drastically altered with 11 instances of phylogenetic clustering and two instances of phylogenetic overdispersion (
Table 2). For example, plot 6 (2820 m a.s.l.) showed no significant phylogenetic pattern when tree ferns were included in the analyses, but showed significant phylogenetic clustering across all three metrics (SES.PD, SES.MPD, and SES.MNTD) when tree ferns were excluded (
Table 2). Few plots remained consistent in their phylogenetic pattern with and without tree ferns. Plot 14 (3320 m a.s.l.) did remain consistent and showed significant clustering for SES.MPD and SES.MNTD with and without tree ferns in the analyses. This suggests that co-occurring species are more closely related than expected by chance. This result is not surprising given that 53 of the 100 stems in this plot are from the genus
Weinmannia within the family Cunoniaceae. In general, there is a non-random phylogenetic structure along the elevation gradient, with discrepancies between low and high elevation plots that are largely influenced by the presence or absence of tree fern species.
Both with and without tree ferns, standardized effects sizes (SES) of MPD and MNTD were not significantly related to elevation. However, without tree ferns, a visible negative trend for both metrics with elevation was obvious (
Figure 4). Our results agreed with prior research findings of increased phylogenetic clustering at higher elevations; hypothesized to be evidence for the influence of abiotic driven processes on phylogenetic community structure [
24,
57,
72,
73]. The relationships between SES.MPD, SES.MNTD, and elevation were significant when two off-trend plots were removed (
Figure 4;
Figure S1). We found that the floras of these two plots were distinct from neighboring floras with multiple species found here exclusively or not exceeding this elevation along the transect. This could be evidence of colliding floras from upper and lower elevations, but more extensive sampling is needed for further investigation. In total, our results support ideas of habitat convergence by tree ferns with angiosperms and an increased filtering of clades leading to phylogenetic clustering at higher elevations.