Drivers of Variation in Avian Community Composition Across a Tropical Island Montane Elevational Gradient

Abstract

Elevational variation in avian community composition can be significantly influenced by habitat degradation, fragmentation and secondary forest growth. Few studies have identified the drivers of changes in bird community composition across disturbed montane elevational gradients of smaller tropical islands. We examined patterns of avian diversity using long-term mist-net data (2008–2014) across three different forest elevations along a tropical montane elevational gradient in the Northern Range of Trinidad, West Indies. All three sites (lowland mature secondary forest, mid-elevation highly disturbed secondary forest, and undisturbed high elevation forest) were found to have distinctive bird communities. Turnover rather than nestedness explained most of the total dissimilarity between sites. Whilst some turnover could be attributed to elevation, changes to diversity at the mid-elevation site result more from local habitat heterogeneity related to human activities and secondary growth, with increased species richness attributable to habitat-generalist species indicative of disturbance. Significant anti-nestedness in species occupancy was observed, underpinned by the loss of ground-dwelling and understory insectivores from the mid-elevation site. Differences in bird community composition, in contrast, were driven by the abundance of specialist nectarivores in the highest elevation undisturbed montane forest, and by generalist nectarivores and frugivores at lower elevations.

1. Introduction

Patterns of avian diversity across different spatiotemporal scales are driven by topographic heterogeneity, and both climatic and biotic assembly processes [1,2,3,4], with additional constraints imposed by extent of habitat [5] and geographic isolation [6]. Human activity is further shaping these patterns, with urbanization and agricultural intensification exacerbating habitat loss and degradation, with a concomitant reduction in bird taxonomic and functional diversity [7]. Across tropical montane elevational gradients, changes in avian community composition reflect degrees of habitat specialization driven by topographical and climatic variation [8,9], compounded by interspecific competition [4,10]. Quantifying changes in diversity along tropical montane elevational gradients can improve our understanding of habitat specialization by birds at different elevational zones and ecotones, and how habitat heterogeneity along these gradients helps to maintain diversity, as well as identifying the impacts of anthropogenic activities on montane bird communities [11].

Much of this above-mentioned research has focused on key areas in the South American Andes (e.g., [12,13]) and in Central America (e.g., [9,14,15]), but broad patterns of avian diversity in elevational gradients have been studied in subtropical and temperate latitudes across the globe ([8], and references therein). Elevational changes in avian diversity have also been studied on large island systems such as Madagascar [16] and New Guinea [17], but less is known regarding patterns of elevational diversity of montane bird communities on smaller islands, where bird diversity is typically lower due to differences in area, climate and isolation [6].

Measures of community composition in any given habitat or region are commonly described in terms of observed species richness and several alpha (α) diversity indices [18]. Beta (β) diversity refers to the difference in diversity between two or more communities, though there are many different concepts and measures of β-diversity [18,19]. This dissimilarity can result from two opposing processes: species turnover—the replacement of some species by others; nestedness—the degree to which species found in sites of low species richness are a subset of those present in richer sites, resulting from non-random loss [20]. Partitioning β-diversity into these two components can help to determine the ecological processes that underpin observed differences [20,21].

Changes in bird community composition can also result from differences in species abundance, with some studies reporting that increased abundance of just a small number of species can drive significant changes in community composition between forest edges and interior habitat [22]. Although several β-diversity metrics have been described and their performance assessed (see [23,24,25]), few studies have used these approaches to examine the drivers of changes to avian community composition across smaller tropical island montane elevational gradients.

Here we begin to address this knowledge gap using bird mist-netting data from seven consecutive breeding seasons, to describe and quantify variation in avian community composition at three different forest sites along an elevational gradient in the Northern Range mountains of the island of Trinidad and identify the primary drivers of these patterns. Specifically, across the elevation gradient, we: (1) describe patterns of alpha diversity (species richness and evenness) and determine how this differs spatially (between sites) and temporally (over the 7-year study period); (2) examine patterns of beta (β) diversity between sites and between different years to quantify the relative contributions of species turnover and nestedness, and identify the bird species unique to each site; (3) determine whether species with specific ecological traits may be driving differences in community composition and whether these patterns correlate with elevation or levels of anthropogenic disturbance.

2. Materials and Methods

2.1. Study Sites

Between 2008 and 2014, staff and students from the University of Dundee (UK) used mist nets to capture and ring birds from three forest sites, differing in elevation and degree of anthropogenic disturbance, in the Northern Range mountains (Figure 1), Trinidad, West Indies. This island is situated 11 km from the northeast coast of Venezuela, having separated from the mainland during the Pliocene, approximately 4 Ma. This proximity, combined with the periodic formation of land bridges, means the island’s flora and fauna resembles most closely that of the South American continent, albeit less species rich [26]. The Northern Range is an 88 km long metamorphic mountain range spanning east to west in the north of the island, with peaks and ridges varying between 456 and 760 m above sea level (the highest peak, Cerro del Aripo, reaches 940 m). The first survey site, Arena, is located approx. 8 km southeast of the town of Arima, in the Arena Forest Reserve, at 30 m above sea level. Throughout much of the 19th and 20th centuries, the lowland forest was managed for timber logging and charcoal production, which ceased to operate in the 1980s. The site has since been managed by the forestry bureau for recreational purposes. The lowland mature secondary forest here shows irregular vegetation stratification, with an upper layer of occasional emergent trees (30–42 m), a continuous middle canopy (12–27 m) and a continuous lower sub-canopy, with abundant palms and liana species.

The second site, Verdant Vale, sits in the Arima Valley in the foothills of the Northern Range, at an elevation of 250 m. The disturbed seasonal deciduous forest at this mid-elevation site was previously partially cultivated with cacao (Theobroma cacao) and citrus (Citrus spp.) trees, with these areas abandoned and more recently dominated by fast-growing species such as mahogany (Swietenia sp.) and bamboo (Bambusa vulgaris) [27]. The semi-open canopy is low (3–9 m) with emergent trees reaching 18 m. The ground layer is sparse, except on steep slopes when shrubby vegetation appears. Industrial-scale quarrying for limestone has been undertaken in Verdant Vale since the 1940s. Although mining operations in the region waned during the 1980s and 90s, open pit limestone mining as well as illegal quarrying and timber harvesting have proliferated over the last 20 years [28].

The final site is located at Morne Bleu, on the ridge of the Northern Range mountains. This higher elevation site is dominated by undisturbed transitional vegetation between lower montane and montane evergreen forest, with a closed canopy (21–30 m in height) and a dense, unstratified understory dominated by tree ferns and small palms. The ridge is relatively exposed, with the northern slope particularly steep, and at an elevation of 725 m above sea level the forest is often shrouded in clouds, creating wet conditions.

2.2. Mist Net Sampling and Data Management

Bird communities at each site were sampled using 5–8 mist nets (18 m length, 2 m height) positioned in freshly cut transects. Nets were opened from just before dawn (0600) until mid-afternoon (approx. 1500). Each bird caught was extracted, identified, and ringed following the standard methodology described by the British Trust for Ornithology [29]. Birds were ringed using metal alloy rings with unique identification numbers obtained from the BTO by the licensed trainer on the expeditions, the exception being hummingbird species (due to the short tarsi which prevents ringing). Thus, for each individual hummingbird three small dots of acrylic paint were painted on the back of the head, enabling identification of recaptured birds. For all other birds caught, the species, ring number, recapture status, age and sex (where identifiable), wing length and weight were recorded. Mist nets were checked every twenty to thirty minutes to reduce the risk of predation and distress to the birds, and nets were temporarily closed together during periods of severe rain. Each site was sampled twice, for 2 to 3 consecutive days each time, between June and August of each sampling year (corresponding to the majority of their known breeding season). Sampling effort was constant across sites and years (32 net days per site per year) except for 2013, where operational constraints reduced the sampling effort at Morne Bleu (24 net days) and Arena (16 net days).

Ringing logs provided by the expedition were converted to a matrix of community data by first grouping the data by species, then counting the number of unique ring numbers that occurred of each species in each location in each of the seven years. Using this method ensured that no individual was counted more than once within each year but included individuals that may have been counted more than once across different years. Records for destroyed or lost rings (28 and 5 records, respectively) were removed from the dataset. The same process was used to create matrices of community data grouped only by year, and by location. Hummingbirds were painted with temporary paint that remained for less than 4 weeks, and so individuals from different years were painted with the same order of paint (e.g., Red White White, RWW). Given the risk of hummingbirds from different years being incorrectly identified as the same individual, the paint codes were amended so that each was counted separately. Catching and ringing of birds was conducted with the annual approval of the Wildlife Section, Ministry of Agriculture, Land and Fisheries Dept., Government of the Republic of Trinidad and Tobago. Permission was kindly granted to Dr. Alan Barclay by the BTO for use of their rings in Trinidad (code: FW: 4/10/3, 17 June 2010; code 5/10/7, 1 June 2013; code 4/10/3, 12 June 2014).

2.3. Statistical Analyses—Alpha Diversity Measures

All analyses were conducted in R v. 4.4.1 [30] using the IDE RStudio version 2024.12.0+467 [31]. Rarefaction curves were used to visually determine whether the number of captures were likely to be sufficient for the observed species richness to approach the actual species richness at each site, and each year.

Alpha diversity was estimated for the bird communities at each site using first, second and third-order Hill numbers, which can be considered as the effective number of equally abundant species and includes a parameter q that determines the extent to which species frequencies are considered. First-order Hill numbers (q = 0), ⁰D, denote the species richness. For values of 0 < q < 1, rare species have a greater weighting, but as q increases beyond 1, the resulting Hill numbers become more sensitive to dominant species [32]. Second-order Hill numbers (q = 1), ¹D, equal to the exponential of Shannon’s diversity index (H), give the effective number of species, whilst third-order Hill numbers (q = 2), ²D, are the effective number of dominant species in a community, equivalent to the inverse of the Simpson’s diversity index. All diversity values were compared between sites using 95% confidence intervals (CI); if no overlap is seen at the 95% CI for diversity values, the differences between sites can be interpreted as being significant at the p < 0.05 level [33,34]. Hill numbers for each site were calculated using functions in the iNEXT package [35].

Multiple linear regression was used to model observed species richness as a function of year, site, and the interaction between the two. The model was simplified using backward stepwise selection based on statistical significance, and model residuals confirmed as satisfactory through model-checking plots. Observed species richness between sites was further analyzed as a linear mixed effects model, with year included as a random factor to account for interannual variation in temperature, precipitation, etc., using the R package nlme [36].

2.4. Beta (β) Diversity and Nestedness

Beta (β) diversity between the three different elevations (all years combined and within each year) was calculated using Sørensen and Jaccard dissimilarity indices, using the betapart package (

Table A1. Formulae for the measures of alpha and beta diversity measures used in this study.

Metric	Notation	Description a	Reference
Hill Number	qD	${}^{q}D={\left({\displaystyle \sum _{i=1}^S}{p}_i^q\right)}^{1/(1-q)}$	[43]
Sørensen dissimilarity index	βsor	${\mathsf{\beta }}_{sor}={\displaystyle \frac{b+c}{2a+b+c}}$	[20,37]
Simpson dissimilarity index (turnover fraction of Sørensen dissimilarity)	βsim	${\mathsf{\beta }}_{sim}={\displaystyle \frac{\mathrm{m}\mathrm{i}\mathrm{n} (b,c)}{a+\mathrm{m}\mathrm{i}\mathrm{n} (b,c)}}$	[20,37]
Nestedness fraction of Sørensen dissimilarity	βsne	${\mathsf{\beta }}_{sne}={\displaystyle \frac{b+c}{2a+b+c}}-{\displaystyle \frac{\mathrm{m}\mathrm{i}\mathrm{n} (b,c)}{a+\mathrm{m}\mathrm{i}\mathrm{n} (b,c)}}$	[20,37]
Jaccard dissimilarity index	βjac	${\mathsf{\beta }}_{jac}={\displaystyle \frac{b+c}{a+b+c}}$	[21,37]
Turnover fraction of Jaccard dissimilarity	βjtu	${\mathsf{\beta }}_{jtu}={\displaystyle \frac{2 \mathrm{m}\mathrm{i}\mathrm{n} (b,c)}{a+2 \mathrm{m}\mathrm{i}\mathrm{n} (b,c)}}$	[21,37]
Nestedness fraction of Jaccard dissimilarity	βjne	${\mathsf{\beta }}_{jne}={\displaystyle \frac{b+c}{a+b+c}}-{\displaystyle \frac{2 \mathrm{m}\mathrm{i}\mathrm{n} (b,c)}{a+2 \mathrm{m}\mathrm{i}\mathrm{n} (b,c)}}$	[21,37]

^a where p_i is the proportion of individuals found in the ith species, S is the total number of species, a is the number of species present in both sites, b is the number of species present in the first site but not in the second, and c is the number of species present in the second site but not in the first. The parameter q determines the sensitivity of Hill number to the relative frequencies.

) [37]. Both indices use species incidence, rather than species abundance, to quantify beta diversity; the Jaccard index calculates the number of unique species as a proportion of the total number of species recorded, whilst the Sørensen index gives double the weight to the shared species [38]. For each dissimilarity index, the β-diversity was partitioned into two components: species turnover and nestedness-resultant dissimilarities (Table A1). β-diversity between successive years, for each site and for all sites combined, was also calculated, and simple linear regression used to model the relationship between each measure of β-diversity (total β-diversity, species turnover or nestedness-resultant dissimilarities, for both Sørensen and Jaccard indices) and year. Nestedness between the three sites was estimated using the Nestedness metric based on Overlap and Decreasing Filling (NODF) metric [39]. In addition to the overall nestedness of the community occupancy matrix, the NODF metric also gives an estimation of the nestedness in species composition between sites (NODF among rows, NODFr), and in species occupancy (NODF among columns, NODFc). The significance of the NODF indices were evaluated through comparisons against 1000 null model simulations, using occupancy matrices generated with the ‘quasiswap’ algorithm that maintains both row and column frequencies [40]. Nestedness between years, both overall and within each site, was investigated in the same manner through occupancy matrices of bird species occurrence against year. NODF analyses were carried out using the nestednodf method in the oecosimu function in the vegan package [41].

2.5. Bird Community Dissimilarity

Differences in species relative abundance (total number caught during mist net sampling) and differences in avian community composition based on Bray–Curtis dissimilarities between the three Trinidad sites were examined using non-metric multidimensional scaling (NMDS). NMDS ordination was also used to examine the dissimilarities in the bird assemblages caught each year, for all sites combined. PERMANOVA was used to test the significance of the distance between the centroids of the locations and years, and thus the differences in composition and/or relative abundances of the bird communities, in the adonis2 function in the vegan package [41]. To better understand which bird species had the greatest contributions to the observed distances between the three sites, similarity percentages for each species between Arena (low elevation)-Verdant Vale (mid-elevation), Arena (low elevation)-Morne Bleu (high elevation) and Verdant Vale (mid-elevation)-Morne Bleu (high elevation) locations were calculated using the simper function in the R package vegan [41]. Bird species were categorized into trophic-behavioral guilds (fruit and/or nectar feeders, insectivores, omnivores and seed eaters) according to [42].

3. Results

3.1. Measures of Bird Alpha Diversity Across Sites and Years

Over the seven-year period, a total of 8584 bird captures were recorded, corresponding to 7054 individuals across 103 species (

Table 1. Total number of mist net captures, individual birds and number of species caught each year and at each forest site. Arena = low-elevation; Verdant Vale = mid-elevation, and Morne Bleu = high elevation.

Year	Site	Number of Captures		Number of Individuals		Number of Species
		By Site	Yearly Total	By Site	Yearly Total	By Site	Yearly Total
2008	Arena	294	1273	263	1157	32	76
	Verdant Vale	365		324		52
	Morne Bleu	614		570		43
2009	Arena	276	1292	245	1180	32	75
	Verdant Vale	457		407		46
	Morne Bleu	559		528		44
2010	Arena	238	1407	202	1249	27	61
	Verdant Vale	610		535		44
	Morne Bleu	559		512		36
2011	Arena	199	1315	188	1221	32	64
	Verdant Vale	470		433		37
	Morne Bleu	646		600		38
2012	Arena	187	1117	165	1006	26	67
	Verdant Vale	406		369		45
	Morne Bleu	524		472		35
2013	Arena *	94	852	89	783	25	62
	Verdant Vale	426		384		44
	Morne Bleu *	332		310		33
2014	Arena	230	1328	197	1186	32	68
	Verdant Vale	565		493		43
	Morne Bleu	533		496		36
		Number of captures		Number of individuals		Number of species
All Years combined	Arena	1518		1126		48
	Verdant Vale	3299		2709		76
	Morne Bleu	3767		3219		58

* indicates full sampling effort was not achieved.

). Rarefaction curves for each site and each year were close to asymptotic but did not fully plateau, suggesting that the sampling effort may not have been sufficient to fully capture the species richness at each site (Figure 2). Whilst for most years the highest number of captures was at the higher elevation Morne Bleu site, observed species richness was higher for the mid-elevation disturbed Verdant Vale site, and lowest for the lowland elevation Arena site (Table 1 and Figure 3). Both Year and Location (elevation) were found to have a significant effect on observed species richness (β_year = −1.095, p = 0.007, β_{Morne Bleu} = 8.57, p < 0.001, β_{Verdant Vale} = 15.14, p < 0.001), and these two covariates explained around 80% of the variation in observed species richness (Adj. R² = 0.806). A further model that included an interaction term showed no significant interaction between Year and Location on observed species richness (F = 0.543, p = 0.592). Year and location remained significant when species richness from 2013, when full sampling effort was not reached, was excluded (β_year = −0.995, p = 0.028, β_{Morne Bleu} = 8.67, p < 0.001, β_{Verdant Vale} = 14.5, p < 0.001). The effect of location on species richness persisted when year was modelled as a random factor (β_0:Arena = 29.29, p < 0.001, β_{Morne Bleu} = 8.57, p < 0.001, β_{Verdant Vale} = 15.14, p < 0.001, between year variance = 2.55).

Verdant Vale, the mid-elevation disturbed site, showed significantly greater observed species richness than either Arena or Morne Bleu (Hill numbers q = 0, Figure 4). However, the steep decline in diversity with increasing Hill number order for Verdant Vale shows this site to be the most uneven in terms of the distribution of relative abundances (Figure 4, [43]). In contrast, although captures were lowest and fewest species were caught at the site at the lowest elevation, Arena, this site displayed the greatest effective number of species and the greatest effective number of dominant species (Hill numbers q = 1 and q = 2), and the least steep decline with increasing Hill number order, demonstrating greater evenness of relative abundances (Figure 4).

3.2. Beta Diversity and Nestedness Between Elevations

Sørensen and Jaccard dissimilarities both showed similar levels of β-diversity between the mid-elevation Verdant Vale site and both the higher-elevation Morne Bleu and lowland Arena sites. Low β-diversity between lowland Arena and high elevation Morne Bleu suggested a higher level of similarity between these two bird communities despite these being the two most widely separated elevational sites (Figure 5a,b). Turnover was a greater contributor to the dissimilarities in species assemblages than nestedness in all between-site comparisons but was greatest between mid-elevation Verdant Vale and high-elevation Morne Bleu (Figure 5a,b). Dissimilarities due to nestedness was greatest between Verdant Vale and Arena, and more similar between Morne Bleu and both other sites. Overall, the bird communities sampled at the three forest sites showed significant anti-nestedness (N_obs = 54.07, N_exp = 56.51, Z-value = −6.25, p = 0.001; Figure 6a).

The observed NODFr between sites (N_obs = 70.11) was significantly higher than expected from the null model (N_exp = 69.53, Z-value = 2.66, p = 0.015), indicating that the extent to which sites with lower species richness were sub-sets of sites with higher species richness was significantly greater than from a random distribution (Figure 6b). This is perhaps not surprising, as generalist forest bird species were present across all three elevations. In contrast, the observed NODFc (N_obs = 54.06) was significantly lower than expected from the null model (N_exp = 56.50, Z-value = −6.25, p = 0.001), meaning that uncommon species were less frequently found at the sites of greater species richness (Figure 6c). The disturbed mid-elevation Verdant Vale site was the most species rich, including 35 species not recorded at the lower and higher elevation sites, but there were more infrequently recorded species found only at the lowland Arena and/or high-elevation Morne Bleu than would be expected by chance (

Table A2. Bird species unique to each site across the study period.

Arena (4 spp. Total)	Verdant Vale (35 spp. Total)			Morne Bleu (13 spp. Total)
Dusky-Capped Flycatcher ■ Myiarchus tuberculifer Lesser Elaenia ● Elaenia chiriquensis Red-Crowned Ant Tanager ■ Habia rubica Yellow-Breasted Flycatcher ■ Tolmomyias flaviventris	Amazonian White-Tailed (Green-backed) Trogon ▲ Trogon viridis Barred Antshrike ■ Thamnophilus doliatus Blue Dacnis ● Dacnis cayana Chestnut-Bellied Seed Finch ♦ Sporophila angolensis Ferruginous Pygmy Owl ■ Glaucidium brasilianum Fuscous Flycatcher Cnemotriccus fuscatus ■ Great Kiskadee ● Pitangus sulphuratus Grey-Breasted Martin ■ Progne chalybea Greyish Saltator ■ Saltator coerulescens Lesson’s Seedeater ♦ Sporophila bouvronides Lineated Woodpecker ■ Dryocopus lineatus Long-Billed Starthroat ▲ Heliomaster longirostris	Mouse-Coloured Tyrannulet ■ Phaeomyias murina Pale-Breasted Spinetail ■ Synallaxis albescens Piratic Flycatcher ▲ Legatus leucophaius Red-Eyed Vireo ■ Vireo olivaceus Ruby Topaz ▲ Chrysolampis mosquitus Ruddy Ground-Dove ♦ Columbina talpacoti Rufous-Shafted Woodstar ▲ Chaetocercus jourdanii Silver-Beaked Tanager ● Ramphocelus carbo Small-Billed Elaenia ■ Elaenia parvirostris Smooth-Billed Ani ● Crotophaga ani Southern House Wren ■ Troglodytes aedon Southern Rough-Winged Swallow ■ Stelgidopteryx ruficollis	Streaked Flycatcher ● Myiodynastes maculatus Streaked Xenops ■ Xenops rutilans Trinidad Euphonia ● Euphonia trinitatis Tropical Kingbird ■ Tyrannus melancholicus Tropical Mockingbird ■ Mimus gilvus Tufted Coquette ● Lophornis ornatus Turquoise Tanager ● Tangara mexicana Variegated Flycatcher ■ Empidomus varius White-Tipped Dove ♦ Leptotila verreauxi Yellow Bellied Elaenia ● Elaenia flavogaster Yellow Oriole ▲ Icterus nigrogularis	Black-Faced Antthrush ■ Formicarius analis Brown Violet-ear ▲ Colibri delphinae Chestnut Woodpecker ■ Celeus elegans Collared Trogon ■ Trogon collaris Great Antshrike ■ Taraba major Grey-Rumped Swift ■ Chaetura cinereiventris Hepatic Tanager ■ Piranga flava Lined Quail Dove ♦ Zentrygon linearis Olive-Striped Flycatcher ▲ Mionectes olivaceus Orange-Billed Nightingale Thrush ● Catharus aurantiirostris Speckled Tanager ▲ Tangara guttata Stripe-Breasted Spinetail ■ Synallaxis cinnamomea Yellow-Legged Thrush Turdus flavipes ▲

▲ = fruit and/or nectar feeders, ■ = invertebrate feeders, ● = omnivores, ♦ = seed eaters (Feeding guilds from [42,49]).

and

Table A3. Species absent from only one site across the study period.

Absent from Arena Only (7 spp.)	Absent from Verdant Vale Only (10 spp.)	Absent from Morne Bleu Only (6 spp.)
Spectacled Thrush ● Turdus nudigenis Black-Throated Mango ▲ Anthracothorax migricollis Blue-Grey Tanager ● Thraupis episcopus Copper-Rumped Hummingbird ▲ Amazilia tobaci Palm Tanager ▲ Thraupis palmarum Red-Legged Honeycreeper ▲ Cyanerpes cyaneus Southern Beardless Tyrannulet ■ Camptostoma obsoletum	Golden-Olive Woodpecker ■ Colaptes rubiginosus Grey-Fronted Dove ♦ Leptotila rufaxilla Grey-Throated Leaftosser ■ Sclerurus albigularis Plain Antvireo ■ Dysithamnus mentalis Red-Rumped Woodpecker ■ Veniliornis kirkii White-Bellied Antbird ■ Myrmeciza longipes White-Flanked Antwren ■ Myrmotherula axillaris White-Shouldered Tanager ■ Tachyphonus luctuosus White-Throated Spadebill ■ Platyrinchus mystaceus Yellow-Olive Flycatcher ■ Tolmomyias sulphurescens	American Pygmy Kingfisher ▌ Chloroceryle aenea Blue-Black Grassquit ● Volatinia jacarina Forest Elaenia ■ Myiopagis gaimardii Rufous-Browed Peppershrike ■ Cyclarhis gujanensis Shiny Cowbird ■ Molothrus bonariensis Tropical Parula ■ Parula pitiayumi

▲ = fruit and/or nectar feeders, ■ = insectivores, ● = omnivores, ♦ = seed eaters, ▌ = fish (feeding guilds from [42,49]).

). Within each year, total β-diversity was lower between Arena and Morne Blue than between Verdant Vale/Morne Blue and Verdant Vale/Arena for both the Sørensen and Jaccard indices, reflecting the same pattern as when sites are compared across years (

Table A4. Partitioning of Sørensen dissimilarity beta diversity between sites in each year.

	2008			2009			2010			2011			2012			2013			2014
	βsim	βsne	βsor	βsim	βsne	βsor	βsim	βsne	βsor	βsim	βsne	βsor	βsim	βsne	βsor	βsim	βsne	βsor	βsim	βsne	βsor
Arena—Verdant Vale	0.406	0.141	0.548	0.500	0.090	0.590	0.333	0.160	0.493	0.563	0.032	0.594	0.462	0.144	0.606	0.360	0.176	0.536	0.500	0.073	0.573
Verdant Vale—Morne Bleu	0.419	0.055	0.474	0.568	0.010	0.578	0.389	0.061	0.450	0.486	0.007	0.493	0.486	0.064	0.550	0.485	0.074	0.558	0.444	0.049	0.494
Arena—Morne Bleu	0.250	0.110	0.360	0.156	0.133	0.133	0.148	0.122	0.270	0.344	0.056	0.400	0.231	0.113	0.344	0.160	0.116	0.276	0.375	0.036	0.412
	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR	βSIM	βSNE	βSOR
Overall beta diversity (multiple-site dissimilarity)	0.433	0.103	0.536	0.495	0.066	0.561	0.370	0.119	0.489	0.522	0.030	0.552	0.473	0.108	0.581	0.420	0.125	0.545	0.506	0.055	0.561

and

Table A5. Partitioning of Jaccard dissimilarity beta diversity between sites in each year.

	2008			2009			2010			2011			2012			2013			2014
	βjtu	βjne	βjac	βjtu	βjne	βjtu	βjne	βjac	βjtu	βjne	βjtu	βjne	βjac	βjtu	βjne	βjtu	βjtu	βjne	βjac	βjne	βjac
Arena—Verdant Vale	0.578	0.130	0.708	0.667	0.752	0.578	0.130	0.708	0.667	0.752	0.578	0.130	0.708	0.667	0.752	0.529	0.169	0.698	0.667	0.062	0.729
Verdant Vale—Morne Bleu	0.590	0.053	0.643	0.725	0.008	0.590	0.053	0.643	0.725	0.008	0.590	0.053	0.643	0.725	0.008	0.653	0.064	0.717	0.615	0.046	0.661
Arena—Morne Bleu	0.400	0.129	0.529	0.270	0.179	0.400	0.129	0.529	0.270	0.179	0.400	0.129	0.529	0.270	0.179	0.276	0.157	0.432	0.545	0.038	0.583
	βJTU	βJNE	βJAC	βJTU	βJNE	βJTU	βJNE	βJAC	βJTU	βJNE	βJTU	βJNE	βJAC	βJTU	βJNE	βJTU	βJNE	βJAC	βJTU	βJNE	βJAC
Overall beta diversity (multiple-site dissimilarity)	0.605	0.094	0.698	0.662	0.057	0.605	0.094	0.698	0.662	0.057	0.605	0.094	0.698	0.662	0.057	0.592	0.114	0.706	0.672	0.047	0.719

). Again turnover, rather than nestedness, formed the greater component of β-diversity in all pairwise comparisons between sites within years, and, apart from 2014, the smallest dissimilarities due to nestedness were between Verdant Vale and Morne Bleu.

3.3. β-Diversity and Nestedness Between Successive Years

β-diversity was smaller between successive years than between sites, both overall and within each site (

Table A6. Range of Sorensen and Jaccard dissimilarity beta diversity indices between sites within each year, and between years within each site.

	βSOR	βJAC
Overall beta diversity between sites	0.487	0.655
Range of beta diversity between sites, within each year	0.489–0.581	0.657–0.735
Overall beta diversity between years	0.411	0.583
Range of beta diversity between years, within each site	0.085–0.364	0.156–0.538

), a consequence of the short time frame in which species could be lost and/or replaced. There was no obvious trends and no significant relationship between year and either total β-diversity, species turnover or nestedness-resultant dissimilarities within each site or overall, for both Sørensen and Jaccard indices. There was a slight increase in total β-diversity over successive years at the lowland Arena site that was close to significant (β_sor coefficient = 0.022, p = 0.050, β_jac coefficient = 0.031, p = 0.050;

Table A7. Coefficients and significance of linear models of year against Sørensen and Jaccard dissimilarity beta diversity indices.

	Sørensen Dissimilarity						Jaccard Dissimilarity
	Turnover βsim		Nestedness βsne		Total Beta Diversity βsor		Turnover βjtu		Nestedness βjne		Total Beta Diversity βjac
	Coeff.	p	Coeff.	p	Coeff.	p	Coeff.	p	Coeff.	p	Coeff.	p
Arena	0.014	0.285	0.008	0.439	0.022	0.0501	0.021	0.280	0.011	0.494	0.031	0.0501
Verdant Vale	−0.004	0.747	−0.006	0.466	−0.010	0.416	−0.005	0.765	−0.009	0.469	−0.014	0.389
Morne Bleu	−0.006	0.667	−0.001	0.874	−0.007	0.476	−0.008	0.706	−0.002	0.858	−0.010	0.489
Sites combined	−0.003	0.689	−0.001	0.897	−0.005	0.145	−0.005	0.725	−0.002	0.889	−0.007	0.147

). Nestedness of species of birds caught between years across the three sites combined in Trinidad was highly significant (N_obs = 68.33, N_exp = 68.01, Z-value = 1.78, p = 0.007). This was not due to the species caught in more depauperate years being a nested subset of those caught in richer years, as the observed NODFr of the birds caught across sites between years did not differ significantly from that of the null models (N_obs = 82.43, N_exp = 82.38, Z-value = 0.35, p = 0.732). However, there was highly significant nestedness in species occupancy, as shown by the greater observed NODFc than the null models (N_obs = 68.27, N_exp = 67.95, Z-value = 1.78, p = 0.007). Within each site, the NODF metric showed that the sampled bird communities were not significantly more or less nested than random between years, either in terms of species composition or species occupancy (

Table 2. Results of NODF nestedness analysis for species by year matrices of Trinidad birds, with each site considered separately.

Site	NODF Index	Nobs	Nexp (SD)	Z-Value	p
Arena	NODF	69.56	69.74 (0.433)	−0.433	0.566
	NODFr	71.35	71.30 (0.508)	0.108	0.916
	NODFc	69.52	69.71 (0.437)	−0.439	0.566
Verdant Vale	NODF	66.84	66.51 (0.268)	1.248	0.105
	NODFr	77.48	77.42 (0.214)	0.286	0.734
	NODFc	66.76	66.43 (0.269)	1.247	0.105
Morne Bleu	NODF	68.50	68.47 (0.299)	0.081	0.932
	NODFr	82.66	82.70 (0.190)	−0.205	0.848
	NODFc	68.32	68.30 (0.296)	0.083	0.924

NODF = nestedness metric based on Overlap and Decreasing Filling; NODF = nestedness metric based on Overlap and Decreasing Filling in species composition between sites (among rows); NODFc = nestedness metric based on Overlap and Decreasing Filling in species occupancy (among columns).

). In contrast to the total yearly catches, there was also no nestedness in species occupancy within sites between years, and no overall nestedness.

3.4. Bird Community Dissimilarity

The NDMS ordination analyses (stress = 0.0452) revealed a clear distinction in the community composition of birds between all three elevations (Figure 7). The overall difference in the community composition between locations was highly significant (F_2,18 = 28.53, p = 0.001), with location explaining 76.0% of the variation observed. In contrast, there was no distinction between the composition of birds sampled across the three sites when grouped by year, with the distances between the centroids of the ellipses for each year not significant (F_6,14 = 0.251, p = 0.999). Between nine and ten bird species were responsible for making the greatest contribution to the patterns of dissimilarity of the avian community composition between each pair of forest sites (

Table 3. Cumulative contributions of the most influential species of bird to the dissimilarity between each pair of Trinidad forest sites. A = Arena; MB = Morne Bleu; VV = Verdant Vale.

Arena-Verdant Vale		Arena—Morne Bleu		Verdant Vale—Morne Bleu
Species	Cumulative Contribution	Species	Cumulative Contribution	Species	Cumulative Contribution
Bananaquit (VV) Coereba flaveola	0.215	Bananaquit (MB) Coereba flaveola	0.210	Golden-headed manakin (MB) Pipra erythrocephala	0.126
Palm tanager (VV) Thraupis palmarum	0.287	Golden-headed manakin (MB) Pipra erythrocephala	0.341	Bananaquit (VV) Coereba flaveola	0.219
Plain brown woodcreeper (A) Dendrocincla fuliginosa	0.348	Rufous-breasted hermit (MB) Glaucis hirsutus	0.409	Palm tanager (VV) Thraupis palmarum	0.281
Blue-black Grassquit (VV) Volatinia jacarina	0.400	Purple honeycreeper (MB) Cyanerpes caeruleus	0.473	Green hermit (MB) Phaethornis guy	0.342
Violaceous euphonia (VV) Euphonia violacea	0.447	Green hermit (MB) Phaethornis guy	0.536	Purple honeycreeper (MB) Cyanerpes caeruleus	0.390
Rufous-breasted hermit (VV) Glaucis hirsutus	0.489	White-chested emerald (MB) Amazilia brevirostris	0.585	Blue-black Grassquit (VV) Volatinia jacarina	0.437
Spectacled thrush (VV) Turdus nudigenis	0.531	Blue-chinned sapphire (MB) Chlorestes notata	0.628	White-necked thrush (MB) Turdus albicollis	0.481
White-bearded manakin (A) Manacus manacus	0.566	Plain brown woodcreeper (A) Dendrocincla fuliginosa	0.671	Spectacled thrush (VV) Turdus nudigenis	0.518
Ruddy ground dove (VV) Columbina talpacoti	0.600	Bay-headed tanager (MB) Tangara gyrola	0.704	Plain brown woodcreeper (MB) Dendrocincla fuliginosa	0.552
Golden-headed manakin (A) Pipra erythrocephala	0.631			Ruddy ground dove (VV) Columbina talpacoti	0.583

).

4. Discussion

4.1. Elevational Variation in Alpha and Beta Diversity

Mist-net sampling conducted over seven consecutive breeding seasons across the elevational gradient revealed significantly higher observed species richness at the disturbed Verdant Vale mid-elevation site (Figure 3 and Figure 4). When species diversity was weighted by abundance, however, the lower number of effective species at Verdant Vale compared with Arena (Hill numbers q = 1), and lower number of effective dominant species than both Arena and Morne Bleu (Hill numbers q = 2) indicates this to be the least even of the three sites, instead being most greatly dominated by a small number of species.

Whilst some montane systems show a monotonic decline in bird species richness with increasing elevation, increases in bird species richness at mid-elevational gradients as observed here are well documented, particularly when differences in surface area of each elevational ‘belt’ are considered [8,44]. However, the bird assemblages at Verdant Vale were more dissimilar (higher β-diversity) to those in the Arena lowland forest and the high elevation Morne Bleu forest than these two sites were to each other (Figure 5 and Figure 7), indicating additional factors are driving the increased avian richness and diversity at the mid-elevation habitats in this montane range.

The ongoing quarrying and timber harvesting around Verdant Vale has resulted in a heterogeneous landscape of bare land, localized development, forest edge habitat, and a patchwork of secondary forest at various stages of succession. Consequently, this landscape mosaic may be influencing the mid-elevation site bird community composition by supporting a greater range of edge specialists and habitat generalists [45,46]. Ref. [47] showed that the alpha diversity of birds from this study region was similar across actively managed cacao shade plantations, secondary forest of different ages and primary forest but community dissimilarity increased with increasing age difference between forest habitats, underpinned by a decrease in the proportion of obligate forest-dependent bird species. More broadly, bird species richness tends to be, on average, around 12% lower in secondary forests compared to primary forests [48]. Whilst this contrasts with the increased species richness observed in Verdant Vale, it is consistent with the lower species richness and number of captures at the more mature Arena secondary lowland forest.

Of the 78 species recorded from Verdant Vale, almost half (35 species) were not found at either Morne Bleu or Arena, including species indicative of more disturbed habitats. Whilst two forest specialist granivores (grey fronted dove Leptotila rufaxilla and lined quail dove Zentrygon linearis) were only present at the higher elevation Morne Bleu and lowland Arena site, granivorous species including Lesson’s seedeater (Sporophila bouvronides), chestnut-bellied seed finch (Sporophila angolensis) and ruddy ground dove (Columbina talpacoti), found predominantly in shrub and cultivated land [49], were captured only at the mid-elevation Verdant Vale site. In addition, the greater abundance of C. talpacoti and the blue-black grassquit (Volatinia jacarina), another shrubland seedeater at Verdant Vale, contributed to the dissimilarity of the three bird communities (Table 3).

4.2. Species Turnover Between Sites and Elevation

Species turnover has been reported to be a greater contributor than nestedness to elevational changes in assemblages of birds and other taxa in both tropical [50,51] and temperate [52,53] montane environments. However, β-diversity in temperate systems tends to be lower, so nestedness may play a stronger role in the reduction in diversity seen at higher elevations [54]. Known elevational species ranges can provide a partial explanation of turnover; however, the environmental gradients of a given montane system may further restrict ecological distributions [9,55]. A sharp decline in bird species richness at 1750 m elevation in the Bolivian Andes, for example, corresponded to many species reaching their upper elevational limit between 1250 and 1700 m, resulting in high turnover, though other peaks in turnover did not correspond to clusters of known elevational species limits [56]. At 725 m, the elevation of Morne Bleu perhaps does not preclude occupancy by bird species found elsewhere based on their upper elevational limits, but lower elevational limits may explain some of the observed turnover. Species found only at the high elevation Morne Bleu site include hepatic tanager (Piranga flava) and orange-billed nightingale thrush (Catharus aurantiirostris), which are typically only found at elevations above 400 m; yellow-legged thrush (Turdus flavipes), commonly found above 600 m; and speckled tanager (Tangara guttata) and stripe-breasted spinetail (Synallaxis cinnamomea), found above 700 m [57].

4.3. Nestedness in Species Composition Between Sites

We found significant nestedness in the species composition between the mid-elevation Verdant Vale and the other two less species-rich sites, reflecting the prevalence of widespread forest species present at all three elevations. There was significant anti-nestedness overall, driven by anti-nestedness in species occupancy; specifically, there were more infrequently recorded species found only in the lowland Arena and/or higher elevation Morne Bleu site than would be expected by chance (Figure 6). Nine of the 10 species absent only from Verdant Vale were understory and terrestrial insectivores. Insectivorous species richness has been shown to decrease along tropical montane elevational gradients (e.g., [50]), but the absence of insectivores from the mid-elevation Verdant Vale site alone is more likely due to the current levels of anthropogenic disturbance and less favorable structure of recent secondary forest growth. Several studies have shown that many terrestrial and understory insectivorous bird species are sensitive to tropical forest disturbance and fragmentation [58,59,60,61]. Whilst the forest at Verdant Vale is not heavily fragmented, the disturbance from areas of quarrying and timber harvesting creates edge habitats, which is known to adversely affect understory insectivore populations [22].

4.4. Dissimilarities in Bird Communities

Most of the species that contributed to the dissimilarities between elevations were frugivores and nectarivores (Table 3). The greater abundance of bananaquits (Coereba flaveola), palm tanagers (Thraupis palmarum) at Verdant Vale strongly contributed to the differences in community composition between sites (Table 3). Both species are abundant and widespread in Trinidad [47], with broad habitat preferences including forest edges, shrubland, open disturbed woodlands, and gardens [62]. In contrast, the high-elevation Morne Blue bird community composition was strongly influenced by the greater abundance of hummingbirds, the nectivorous purple honeycreeper (Cyanerpes caeruleus), and two frugivorous, golden-headed manakin (Pipra erythrocephala) and bay-headed tanager (Tangara gyrola). Nectarivores generally show little variation in species richness across elevational gradients [12,50], or across mature tropical forests and successional vegetation in the Peruvian Andes [63], though habitat loss may reduce species diversity [64] or result in a high species turnover that relates to changes in the plant community composition associated with forest fragmentation [65]. We suspect that the shift from hummingbirds dominating the undisturbed high-elevation Morne Bleu forest, to the increase in bananaquits and palm tanager captures at Verdant Vale, is likely to reflect food source available. Nectarivores and frugivores are common montane elevational migrants [14,66,67]. In our study, sampling occurred during the same two months of each breeding season and may not have adequately captured variation in bird community composition due to elevational migration since no recaptured marked birds were found at any different site to the one in which they were first caught.

4.5. Temporal Changes in Avian Community Composition

There was only a small variation in the number of bird species caught each year across the three sites and a small yet significant decline in the number of species caught each year at each site. Consequently, β-diversity was smaller between successive years than between sites, with no significant nestedness in species composition between years detected. This suggests that the species either not present or not caught in more depauperate sampling years was more likely random; however, given that sampling effort between years was near constant, the occurrence of rarer bird species in more species-rich years, as evident from the significant nestedness in species occupancy, may be attributed to other factors.

Accounting for species’ detectability may be necessary to detect patterns of turnover or nestedness in species loss over time [68]. No nestedness in species composition or occupancy was observed between years within each elevation sampled as with the overall yearly species captures. This, in part, may be due to the use of mist-netting as the sole sampling method, since mist nets capture only a subset of the bird community that utilize mid-story to ground level, and under-sample canopy-dwellers, particularly larger-sized species [15,69]. Whilst some avian β-diversity studies have used only capture data (e.g., [22,70]), most use either auditory and visual counts (e.g., [9,47]), or a combination of both (e.g., [71,72]), with statistical assessments of species detectability [65,68] to produce more comprehensive species lists. A similar approach for small island tropical montane forest ecosystems would enable both alpha and β-diversity to be estimated with greater certainty and help better infer the mechanisms driving any shifts in community composition [73].

Diversity metrics can also be calculated based on the phylogenetic relatedness among species [74] or on the extent of functional differences among the species present [75]. Changes in species richness do not necessarily reflect changes in functional diversity, although increased species richness in fragmented tropical forests can be accompanied by a reduction in functional traits at lower elevations [55]. Along tropical montane elevational gradients, avian functional diversity tends to be overdispersed at lower elevations and clustered at higher elevations, driven by resource availability [76]. Increased land use changes (caused by selective logging and agriculture) can affect the retention of functional traits in bird communities, leading to substantial reductions in the functional traits, particularly for both frugivores and insectivores [77]. Future studies of smaller island tropical montane elevational gradients should focus on quantifying the differences in functional diversity using characteristics beyond broad feeding guilds based on morphological traits [77] or ecological traits [78]. For the Northern Range in Trinidad, such research would give better insights into the range of niches supported along the elevational gradient and the effect of disturbance on the resilience of the montane bird assemblages. Sampling at additional points along the montane gradient, at sites with different levels of disturbance, would enable the direct effect of elevation on species diversity to be more precisely modelled.