Aim We examine latitudinal effects of breeding bird taxonomic, phylogenetic and functional β-diversity (Tβ, Pβ and Fβ, respectively) along elevational gradients to test predictions derived from Janzen’s (1967) classic ideas that tropical mountains represent stronger dispersal barriers than temperate mountains.

Location Global

Taxon Birds

Methods Using a global dataset from 46 mountains, we examine latitudinal patterns of Tβ, Pβ, and Fβ, and their components: β _rich and β _repl. For each mountain and each dimension of diversity we calculated (a) total β-diversity, (b) the steepness of distance decay patterns, and (c) within-mountain variability in pairwise β-diversity and regressed each one of these response variables against absolute latitude. We predicted that tropical montane biotas would have (1) overall higher Tβ, Pβ, and Fβ, (2) faster distance decay patterns and (3) higher within-mountain variability in pairwise β-diversity. Furthermore, we expected the richness component β _rich to be more important in temperate mountains, and the replacement component β _repl in tropical mountains.

Results Latitude had a negative effect on total β-diversity for all dimensions of diversity. Similarly, metrics of Tβ and Pβ mostly agree with our expectations, whereas Fβ showed contrasting results with steeper distance decay curves and higher within-mountain variability in temperate mountains. Overall, β _rich was a more important component at high elevations in the tropics and across elevations in temperate mountains, and β _repl more important in tropical low and mid-elevations.

Main Conclusions Our findings are consistent with tropical mountain assemblages containing species with narrow elevational distributions, low dispersal ability and potentially high allopatric speciation, resulting in high β-diversity across elevations. Contrasting results for Fβ indicate high niche packing in tropical assemblages, with most changes in functional diversity among assemblages involving species redundant in trait space. --

Elevational data for bird assemblages were extracted from published articles; for details see Montaño-Centellas et al. (2019) [See Appendix A, in main text]. We assumed that a species was present between its highest and lowest reported elevation (range interpolation) and used that range for each species in the analyses. Each gradient was then partitioned into 200 m wide elevational bands and all species occurring in each band were considered as an assemblage.

Phylogenetic information was extracted from an updated version of the avian phylogeny (Jetz, Thomas, Joy, Hartmann, & Mooers, 2012), based on the backbone tree by Hackett et al. (2008). Functional traits were compiled from Wilman et al. (2014). We used three subsets of traits. Two subsets describe specific ecological strategies thought to be important in mediating biotic interactions: diet and foraging strata (Marra & Remsen, 1997; Naoki & Stouffer, 2007), and one, body mass, considered to be a surrogate of environmental tolerance and caloric requirements.

We calculated overall values of Tβ, Pβ and Fβ in each mountain using the framework proposed by Chao et al. (2014) and extended by Chiu & Chao (2014) and Chiu et al. (2014) to calculate Tβ, Pβ and Fβ. This framework takes advantage of the relationship between local diversity (alpha diversity, α), diversity turnover among localities (beta diversity, β) and regional diversity (gamma diversity, γ), and results in one single value of β-diversity per mountain (as opposed to a set of pairwise values across elevations) (Hill, 1973; Jost, 2007).

Then, we calculated the relationship between assemblage dissimilarity (Tβ, Pβ and Fβ) and elevational distance between two elevational bands (i.e. distance decay patterns). for this we calculated pairwise Tβ, Fβ and Pβ among all pairs of assemblages with Sorensen-family indices using the frameworks proposed by Podani and Schmera (2011) and Carvalho et al. (2012) and fitted fitted power-law models with dissimilarity (pairwise β-diversity) as response variable and elevational distance between two elevations as predictor, for each mountain separately. Distance decay models were calculated with function decay.model in package ‘betapart’ (Baselga, Orme, Villeger, De Bortoli, & Leprieur, 2018). In these models, higher slopes represent curves with faster rates of change in diversity. We extracted the slope, its significance and overall model fit (pseudo r²) for each distance decay model fitted. 

Finally, we calculated the variability of pairwise Tβ, Pβ and Fβ within each mountain. To do this, we used our calculated pair-wise Tβ_sor, Pβ_sor and FB_sor matrices and extracted only pair-wise distances between consecutive elevational bands to represent change in diversity for a 200 m increase in elevation. For each mountain, we summarized overall variability in every 200 m increase in elevation as the SD of Tβ, Pβ and Fβ values.