Why is Amazonia a ‘source’ of biodiversity? Climate-mediated dispersal and synchronous speciation across the Andes in an avian group (Tityrinae)

Abstract

Amazonia is a ‘source’ of biodiversity for other Neotropical ecosystems, but which conditions trigger in situ speciation and emigration is contentious. Three hypotheses for how communities have assembled include (1) a stochastic model wherein chance dispersal events lead to gradual emigration and species accumulation, (2) diversity-dependence wherein successful dispersal events decline through time due to ecological limits, and (3) barrier displacement wherein environmental change facilitates dispersal to other biomes via transient habitat corridors. We sequenced thousands of molecular markers for the Neotropical Tityrinae (Aves) and applied a novel filtering protocol to identify loci with high utility for dated phylogenomics. We used these loci to estimate divergence times and model Tityrinae's evolutionary history. We detected a prominent role for speciation driven by barriers including synchronous speciation across the Andes and found that dispersal increased toward the present. Because diversification was continuous but dispersal was non-random over time, we show that barrier displacement better explains Tityrinae's history than stochasticity or diversity-dependence. We propose that Amazonia is a source of biodiversity because (1) it is a relic of a biome that was once more extensive, (2) environmentally mediated corridors facilitated emigration and (3) constant diversification is attributed to a spatially heterogeneous landscape that is perpetually dynamic through time.

1. Introduction

A species will extend its range until it is stopped by barriers to dispersal. These barriers are sometimes large and conspicuous, and will stop most species that encounter them, and they are sometimes so small that it is difficult to see why a particular species has not crossed them … Which regions a species occupies depends on when it originated and what barriers to its dispersal it encountered. Although the principle is simple, the possibilities are numerous because species are formed at different times, because barriers appear and disappear, and because barriers differ in importance to different types of organisms.

—M. Slatkin [1, p. 787]

Amazonia is an important source of lineage diversity across the Neotropics and has thus had a profound influence on biotic assembly across other biomes [2]. Specifically, Amazonia has been characterized by high rates of in situ speciation and lineage emigration, yet the mechanisms that are responsible for these processes are debated [3–6]. Thus, the observation that Amazonia is a diversity source exposes two important questions. First, what factors have led to Amazonia's high rates of clade origination and in situ speciation? And second, what are the causes of organismal dispersal (i.e. the acquisition of novel distributions) in the Neotropics, specifically with regard to Amazonian lineage emigration?

In general, Neotropical diversification is widely attributed to allopatric isolation across biogeographic barriers [5–9], even though these barriers are often species-specific [1] and spatio-temporally semi-permeable [6,10,11]. Dispersal across biogeographic barriers is often considered to be a stochastic process [12]. For example, happenstance long-distance dispersal events (i.e. dispersal of a few organisms over an existing barrier) through time may result in constant rates of colonization and diversification, which is analogous to a model wherein regional populations stochastically go extinct and become recolonized [13]. Under this mode of diversification, speciation and extinction rates are constant and diversity is largely a function of the amount of time a lineage has persisted on the landscape [14]. If there is a ‘centre of origin’, such as Amazonia [2], a clade would be expected to emigrate from this source gradually, as time predicts the probability of successful dispersal [6]. If this model dominates in the Neotropics, community assembly would be largely determined by happenstance dispersal events through time [15].

In contrast to stochasticity, continental dispersal (as opposed to dispersal to or among islands) is frequently attributed to the formation of habitat corridors that connect previously isolated biomes due to environmental and climatic change [16,17]. Knowledge of such corridors is based on multiple independent lines of evidence, including coordinated dispersal events noted in the fossil record [18,19], phylogeographical patterns and inference [17,20–23], and vegetation models projected into historical climates [24–26]. Biotic distributional change due to these types of processes has sometimes been termed geodispersal [27,28], which is defined as the movement of organisms in response to the loss of a barrier. Conceptually, geodispersal is equivalent to environmentally mediated dispersal due to corridor formation.

The gain and loss of barriers to dispersal through time is consistent with a model of barrier displacement [29], under which geodispersal is an important parameter. Barrier displacement not only results in geodispersal, but can also cause speciation by vicariance due to the formation of a barrier within a species's distribution. Barrier displacement causes barriers and corridors to form repeatedly through time and may have been especially important during the Pliocene and Pleistocene when climatic cycling increased dramatically [30]. Consequently, if barrier displacement has been a dominant process governing dispersal and speciation in the Neotropics, then biotic assembly may be primarily driven by extrinsic factors such as climate change that drive landscape dynamics.

One additional, though not mutually exclusive model that may influence Neotropical biotic assembly is the idea that diversity-dependent mechanisms dominate diversification dynamics by fostering adaptive speciation early in a clade's history [31] and stymying speciation as time increases due to ecological limits on diversity [32,33]. Under this model, emigration and associated speciation are hampered increasingly through time because of a build-up of biological competitors in neighbouring regions. Here, clades are expected to disperse out of a ‘centre of origin’ decreasingly through time, as interspecific competitors accumulate in adjacent regions.

To test the mechanisms driving biotic assembly in the Neotropics, we investigate the biogeographic history of the Tityrinae, a subfamily of passerine birds that includes four genera (Tityra, Iodopleura, Xenopsaris and Pachyramphus) and 25 recognized species, although species-level diversity is underestimated by current taxonomy [34–36]. The group is primarily restricted to semi-open (e.g. gallery and seasonal forests) and edge (e.g. forest or riverine edge) habitats from Argentina through northern Mexico. Several species also inhabit primary forest interior and canopies [37]. Tityrinae are widespread across the Neotropics with replicated species pairs across many major biogeographic barriers in the region, including the Andes and Isthmus of Panama [36]. As a diverse clade of birds that has repeatedly colonized much of the Neotropics, Tityrinae are well suited to test hypotheses concerning the causes of dispersal, diversification and biotic assembly.

To address these questions, we employed massively parallel-sequencing datasets of genome-wide markers that provide low-uncertainty estimates of divergence times and historical relationships (e.g. [38]). These large datasets, though informative, can represent a computational encumbrance, which can render phylogenomic dating intractable over reasonable time frames [39]. For example, gene tree and other systematic error can increase noise and bias phylogenomic estimation [40,41]. However, filtering and subsampling these datasets offer a potential solution to those difficulties by identifying the most valuable markers that can then be used in a more tractable dataset [39,42,43].

We sequenced thousands of ultraconserved elements (UCEs) [44] and applied a novel filtering protocol to investigate whether Amazonia is a source of Tityrinae diversity and test alternative diversification scenarios proposed to explain the Amazonian source pattern. We reduced mechanistic explanations into three testable models of diversification said to explain biotic assembly with Amazonian origination. The first is a neutral stochastic model (e.g. [13,45]). Here, lineages experience chance successful colonization across pre-existing barriers, a model that predicts constant diversification rates and randomly occurring dispersal and ecological shifts through time. The second model is diversity-dependent diversification (e.g. [32]), which predicts decreasing diversification and dispersal rates, and ecological shifts that occur early during clade evolution. The final model is barrier displacement [29], wherein the waxing and waning of barriers due to environmental change cause geodispersal, speciation and extinction. This model predicts that diversification should be continuous and tied to environmental change, that dispersal should be non-random in time, and that allopatry due to barriers should be the primary mode of speciation.

2. Material and methods

(a). Sampling, genetic locus filtering and phylogenomic divergence dating

We sampled UCEs from 48 individuals across all species with the exception of three (Iodopleura pipra, Tityra leucura and Pachyramphus niger) for which we failed to obtain sufficient sequence data. We sampled 35 of the 37 Pachyramphus lineages identified by Musher & Cracraft [36] under a rigorous species delimitation approach, and an additional 19 lineages across the remaining genera that were consistent with taxa delimited in that study. We additionally sampled 12 outgroups for divergence dating of node calibrations (electronic supplementary material, table S1).

To obtain Bayesian estimates of divergence times, a computationally intensive process, we reduced our dataset by filtering based on two criteria: informativeness and clock-like tendency (henceforth, clock-likeness) [39,40,42]. We wrote custom scripts in R to quantify the proportion of informative sites and clock-likeness for each marker. Markers that were most suitable were identified by modelling clock-likeness as a linear function of informativeness. Outliers were identified as loci that deviated by three times the mean Cook's distance among loci (electronic supplementary material).

To obtain a phylogeny for divergence dating, we concatenated all loci for which 95% of taxa had sequence and estimated the topology using RAxML [46] (electronic supplementary material). Because there are no known fossils closely related to Tityrinae, we estimated temporal calibrations for our UCE phylogeny using data from a previously published study [47]. We incorporated additional sequences for Tityrinae available on GenBank and reran the analysis in BEAST v. 2.4.3 [48] (electronic supplementary material, table S1). We then obtained 95% highest posterior densities (HPD) of node ages for five outgroup nodes from BEAST to apply secondary calibrations to our dataset of most clock-like UCE markers. We used the HPD values to set lower and upper bounds at each of the five nodes and ran our filtered UCE dataset in MCMCTree implemented in PAML v. 4.7 [49]. We summarized nodes as mean ages with corresponding 95% HPD (electronic supplementary material).

(b). Biogeographic and ecological modelling

To understand the history of biogeographic area acquisition and loss, we applied a dispersal extinction cladogenesis (DEC) model implemented in BioGeoBEARs [50]. Because the standard biogeographic model-selection framework [51] may be problematic [51], we chose to apply the DEC model alone, which is robust to complex biogeographic scenarios [52]. We constrained the maximum number of areas in the model to two as no extant taxa occupy more than two regions. We assigned each taxon to areas based on a previous study [36] and inferred ancestral distributions under the DEC model. We finally quantified the number of lineages that speciate in situ, emigrate from and immigrate into each area (electronic supplementary material).

To investigate the ecological history of Tityrinae, we modelled ecological shifts of two types across the phylogeny: (1) habitat affinity (forest interior versus non-forest interior) and (2) elevation (lowland versus montane). To estimate transition rates and map these changes onto the phylogeny, we used RevBayes, a model-specification language for constructing graphical models in a Bayesian framework [53]. We constructed a model that allowed for equal probabilities of state (habitat affinity/elevation) addition and loss, ran this MCMC for 10 000 generations and summarized the results as maximum credibility ancestral states after a burnin of 25% (electronic supplementary material).

(c). Macroevolutionary rates

To estimate diversification rates through time, and to identify potential rate shifts, we used both Bayesian analysis of macroevolutionary mixtures (BAMM) [54] and TESS [55] (electronic supplementary material). We also tested for diversity-dependent diversification. First, using the R package Laser [56], we applied a one-tailed t-test γ-test by estimating γ, which is a measure of the relative branching distances along the phylogeny. This test compares the observed γ to a null distribution of a constant birth model [57]. If significant, we reject the null hypothesis in favour of decreasing diversification rates through time. Then, we applied a model-fitting approach implemented in the R package DDD [58] to test models that explicitly incorporate species carrying capacity (K) (electronic supplementary material).

Finally, we evaluated the effect of palaeoclimate on the net-diversification rate. First, we assessed the correlation between episodic net diversification rate inferred from TESS and palaeoclimate (temperature) by using generalized linear models to evaluate the relative effects of time and temperature on net diversification rates during 50 evenly spaced intervals across Tityrinae's phylogeny. We additionally employed RPANDA [59,60] to compare 10 models where net diversification is either constant, varies exponentially with time or varies exponentially with temperature. For all model-selection approaches, we selected the best-fitting model with the lowest AICc and evaluated relative likelihoods of each model using AICc weights (AICω; electronic supplementary material).

3. Results

We reconstructed a time-calibrated family-level phylogeny of modern birds (Neornithes) using the Claramunt & Cracraft [47] dataset and recovered a temporal history consistent with that study (figure 1; electronic supplementary material, table S2). Our DEC model recovered many more transitions out of Amazonia than out of other regions and higher in situ speciation within Amazonia than within other areas (figure 2). Nearly all maximum-likelihood spatial shifts in the model occurred after the beginning of the Pleistocene (figure 2). RevBayes recovered a history of lowland semi-open habitat affinity with nearly all shifts occurring during the Pleistocene.

Figure 1.

(a) Time-calibrated family-level phylogeny for all birds [45] with Tityridae added (red box), showing a hierarchical series of nodes used for secondary calibrations (black circles) for UCEs. (b) UCE time-calibrated phylogeny for Tityrinae. Orange bars represent 95% HPD for nodes and black bars represent the calibration priors applied. The diversity-through-time plot is shown in the centre with model-fitted (pure birth exponential dependence on time; electronic supplementary material, table S4) (black line) and observed (dark red line) values.

Figure 2.

Modelling results for ancestral distributions (centre) and ecology (top right). Boxes on the nodes of the central tree represent maximum-likelihood ancestral distributions and correspond to the coloured polygons on the map of the Neotropical areas: Amazonia (green), dry diagonal (blue), Atlantic forest (yellow), Andes (teal), west of Andes (orange), and Central America (Purple). The red dashed line represents the Miocene–Pliocene boundary. A key to multi-area colours is at the top left. Bar graphs show the number of spatial transitions associated with the number of in situ speciation, area-specific emigration, area-specific immigration events, and range shifts through time.

We rejected a model of diversity-dependent diversification using the t-test γ-test (γ = 1.743714, p = 0.9495) and model-selection approaches, indicating that rates do not decline through time. TESS speciation and extinction rates tended to increase and decrease, respectively, through time (figure 3; electronic supplementary material, figure S3), resulting in an exponential accumulation of diversity (figure 1). Thus, there was a trend of increasing net diversification toward the present with steeper increases in the mean posterior rate at roughly 17–13 and 5–2 Ma. BAMM similarly showed that the best shift configuration recovered an overall increase in speciation rates through time (electronic supplementary material, figure S4). However, the rate changes observed from TESS and BAMM do not represent statistically supported rate shifts (i.e. sudden rate changes), perhaps due to relatively small sample sizes. Additionally, mean net-diversification rate from TESS is apparently negative prior to approximately 16 Ma. This may also be an artefact of sample size, however, and we therefore hesitate to take this observation at face value. For example, net diversification rates in BAMM (electronic supplementary material, figure S4) and RPANDA (electronic supplementary material, figure S6) do not show negative net diversification early on.

Figure 3.

TESS speciation rates (purple line) and highest posterior density (purple shaded region), extinction rates (red line) and highest posterior density (red shaded region), net diversification rates (blue line) and highest posterior density (blue shaded region), and palaeotemperature (red line) [61].

We additionally found that a model in which net diversification rates are not dependent on carrying capacity (K) fits better than models that explicitly incorporate K (AICω = 0.34; electronic supplementary material, table S3). The next best-fitting model estimated K to be greater than 1.04 million species (AICω = 0.23), which also indicates that the diversity-dependent model does not fit Tityrinae diversification well (electronic supplementary material). Finally, we found palaeoclimate to be negatively correlated with episodic net diversification rates derived from TESS (adj. R²=0.68; electronic supplementary material, figure S5 and table S4), and the best-fitting model was one where net diversification rate varies in relation to interactive effects of time and temperature (AICω = 0.99). Despite this relationship, a pure birth model in which speciation varies exponentially with temperature fit our data similarly (AICω = 0.23) to a model in which speciation rate varies exponentially with time (AICω = 0.26) based on our RPANDA framework, a result indicating that the choice among these models remains statistically uncertain (electronic supplementary material, table S5 and figure S6). Models conditioned on stem and crown ages performed similarly.

We compared ages associated with phylogenetic splits across six barriers for which we had natural replicates (figure 4). The four taxon splits across the Andes ranged in mean age from 2.98 to 2.58 Ma and were strongly synchronous in HPD. Across other barriers, there was more variance in mean ages for taxon pairs, such as between Amazonia and Atlantic forests (3.10–1.08 Ma), across the Isthmus of Panama (2.24–1.14 Ma), and between east and west Amazon (1.70–0.79 Ma). Speciation times across the Isthmus of Tehuantepec in a highland taxon (P. major with a mean of 0.32 Ma) and a lowland taxon (P. aglaiae with a mean of 0.49 Ma) were also notably comparable. Remarkably, taxa within P. polychopterus and Tityra (figure 4, black and green, respectively) speciate synchronously across the Andes, Amazonia and Isthmus of Panama. Taxa within the P. cinnamomeus and P. aglaiae groups (figure 4, orange and red, respectively) speciate synchronously across the Andes, arid diagonal and Isthmus of Panama.

Figure 4.

Posterior densities for node ages derived from MCMCTree for speciation across barriers: the Andes, Amazonia to Atlantic forests, eastern to western Amazonia, Isthmuses of Panama and Tehuantepec, and low- and high-elevation forests. Coloured probability densities correspond to clades at the right: (1) Iodopleura (light blue), (2) Tityra (dark green), (3) P. aglaiae group (dark red), (4) P. rufus group (pink), (5) P. polychopterus group (black), (6) P. albogriseus group (dark blue) and (7) P. cinnamomeus group (orange).

4. Discussion

In accordance with previous studies, we find strong evidence that Amazonia is marked by high in situ speciation and lineage emigration when compared with other Neotropical areas (figure 2) [2]. Specifically, our models indicate that the ancestor of Tityrinae probably existed in semi-open, lowland habitats on the proto-Amazonian landscape beginning in the late Oligocene to early Miocene and diversified in situ, forming the four major lineages (genera) that later colonized other biomes and continued to diversify. Importantly, a recent study found Amazonia to be the historical centre of Neotropical diversification but treated the Chocó and Andean forests as Amazonian, which might bias results in favour of an Amazonian source hypothesis [2]. We found that for Tityrinae, Amazonia is a source even when these regions are treated separately from Amazonia.

We rejected two models of Amazonian-origin biotic assembly in favour of a barrier displacement model [29]. We found that ecological shifts in Tityrinae were rare and occurred late in the clade's history, that dispersal occurred at increased rates during the Pleistocene (figure 2), and that diversification was diversity-independent (electronic supplementary material, table S3). Although speciation and extinction were rather constant through time without statistically supported rate shifts associated with changes in global temperature, we see increases in the net diversification rate during periods of global cooling and increased dispersal during climatic cycling (figures 2 and 3; electronic supplementary material, table S4). We thus find no evidence for purely stochastic or diversity-dependent models of biotic assembly, which predict constant Amazonian emigration and decreasing diversification rates, respectively, over time. Instead, speciation seems to be driven by isolation in habitat fragments due to climate deterioration (figure 3; electronic supplementary material, figures S5 and S6) [47] and biogeographic barriers (figure 4), whereas dispersal between biomes occurs due to geodispersal and is expedited by climatically driven corridors that connect biomes (figure 2) [16,20,23]. Though our results are overall consistent with climate-mediated diversification and dispersal, the relationship between temperature and net diversification rate was not statistically robust, perhaps due in part to disparate degrees of temporal precision in our data (imprecise divergence dates versus precise climate data), a relatively small sample size and additional unmeasured processes that cause barrier displacement such as erosion and tectonics [27]. For example, estimated divergence times are relatively imprecise (precise only to the order of 10⁶ years), but large-scale climatic events often occur on the order of 10⁴ years. Thus, more data are needed in order to robustly determine the relationship between climate and rates of net diversification (electronic supplementary material).

Under a model of barrier displacement, clades accrue diversity and expand their ranges as barriers on the landscape wax and wane, which would also explain increasing rates of net diversification through time (figure 3; electronic supplementary material, figure S6). Barrier displacement is a possible mechanism that drives the pattern of increased species richness toward the centre of a clade's distribution (mid-domain effect) [29,43]. As Amazonia represents the origin and distributional centre for many groups of organisms, barrier displacement has probably played a role in facilitating diversification in many Neotropical groups. Biogeographic simulations suggest that allowing for barrier displacement on a neutral landscape not only results in the mid-domain effect, but also in near-exponential lineage accumulation [29], which is consistent with our observed data (electronic supplementary material, figure S6) and the time-to-speciation effect [14].

The correlation between clade age and clade diversity is well known [61] and has sometimes been attributed to happenstance dispersal events that cause an accumulation of species over time [6]. In Tityrinae, relatively homogeneous speciation and extinction rates through time lead to an increase in diversity that is also consistent with a time-for-speciation model [62]. However, dispersal appears to be temporally non-random in Tityrinae as it has been limited to times of increased climate cycling. Thus, because net diversification is relatively continuous through time but dispersal is non-random, we attribute our diversification pattern to environmentally mediated geodispersal and isolation on a landscape that is continuously dynamic through time [17,29]. Although we cannot reject the simpler model where net diversification rates vary only as a function of time based on our data alone, we suggest that this dynamic landscape is probably in part a consequence of climate change because spatio-temporal shifts in temperature and precipitation profoundly influence the distributions of ecosystems [24,63] and the location of barriers [27,64]. It should be noted that time per se is not causal of diversification; rather, the latter is a function of the rate of barrier formation and degradation.

(a). The biogeographic drivers of speciation in the neotropics

Geographical barriers are well-known drivers of speciation (e.g. [8]), but there is debate over whether these barriers severed populations of organisms in situ upon their formation, or whether allopatry and subsequent speciation post-dates such geological events. Phylogenetic splits across barriers were broadly overlapping in age estimates, indicating that closely related sets of taxa respond similarly to environmental change. Phylogenetic splits across the northern Andes were highly synchronous, which is remarkable given the known variance in divergence times that exists across this barrier [6,36]. The northern Andes were roughly 40% of their current height during the Miocene and earliest Pliocene [65,66], suggesting that their final uplift during the Pliocene (around 5–2 Ma) was instrumental in isolating populations of Tityrinae (but see [67,68]). Similarly, shifts in elevation were detected in three lineages during this period (figure 4), indicating that lineages may have been passively transported to [69] or rapidly colonized [70] higher elevations. Taken together, these results show the profound influence of the Andes on the diversification of Tityrinae, including apparently severing dispersal corridors about 3 Ma. Notably, this time frame also corresponds to rapid climate deterioration following the mid-Pliocene warm period that may have additionally reduced available habitat at higher elevations [30].

Despite synchronous splitting across the Andes, some variance exists among speciation events at other barriers (figure 4). For example, multiple speciation events are likely between Amazonia and Atlantic Forest (including adjacent gallery and seasonal forests). Possible barriers have included receding wet, dry and gallery forest connections, and watershed dissociation due to the emergence of the Fitzcarrald arch [71,72] (electronic supplementary material). Forests, including semi-open gallery and seasonal forests, were connected and disconnected from east to west throughout the Quaternary [21,25,73], which probably drove recurrent periods of isolation and contact between populations of Tityrinae. The majority of phylogenetic splits between eastern and western South America within Tityrinae are associated with more southerly forests, which were repeatedly connected by extensive corridors [26]. Our results therefore support the hypothesis that these southerly connections have generally been more important than corridors in the north [74].

Tityrinae colonization of Central America occurred around 3–2 Ma, which is consistent with the timing of the Isthmus of Panama closure forming a major intercontinental corridor that enabled biotic interchange (e.g. [18,19]). Most speciation across the Isthmus occurred during the Pleistocene, implicating an important role for climatically driven habitat shifts across this area [10]. High variance in the timing of speciation across the Isthmus of Panama is consistent with previous studies [10,75,76]. Although the age of the Isthmus remains contentious, our results of colonization occurring after roughly 3 Ma support a relatively young age of this geographical feature [77].

(b). Why is Amazonia a source of biodiversity?

Amazonia, in a sense, is the vestige of a Palaeogene area of endemism, a vast tropical forest that spanned South America [4]. Many modern lineages of plants were probably established in South America by the beginning of the Eocene [78,79], a warm wet period in Earth's history that allowed tropical forests to exist across many latitudes. As climate cooled after the early Eocene optimum [30], tropical environments contracted and have continued to contract as climate further deteriorated, especially after the mid-Miocene approximately when Tityrinae originated (figures 1 and 2). This Amazonian source pattern might be anticipated particularly for suboscine birds and other groups that are of South American origin [47] or that arrived in South America by the middle Eocene [78].

Continued Andean uplift during the early Miocene increased a rain shadow effect in eastern South America, which further reduced the extent of tropical forest and aided in the formation of an ancient wetland system called Pebas [4]. Fossil and palynological evidence indicate that Pebas contained a mix of relatively open and forested habitats [80]. Thus, ample edge and open-forest environments were probably inhabited by taxa specializing on them, including early Tityrinae, in addition to those specializing on terra-firme. We therefore attribute the observation that Amazonia is a biodiversity source to three interdependent phenomena. First, we suggest that the frequency of clade origination in Amazonia is primarily due to the large geographical extent and diversity of habitats that Neotropical forests encompassed during much of the Palaeogene. Second, a dynamic landscape promoted differentiation in its biota, driving high and relatively constant diversification rates through time. Finally, in this dynamic landscape, a high rate of environmentally mediated connectivity between these habitats and those of other regions later facilitated emigration to other regions, and additional speciation [17].

5. Conclusion

Our densely sampled phylogeny of Tityrinae [36] showed no evidence of diversity dependence, a frequently observed pattern in diversification studies that sample biological species [33,58]. Like other studies, we find that by accounting for all taxonomic diversity within a clade, we recover increased resolution for biogeographic inference [81] and divergence dates across barriers [6]. For example, two of four phylogenetic splits across the Andes in Tityrinae correspond to taxonomic diversity seen as intraspecific [36]. We suggest that estimates of diversification rate and causes of speciation should be evaluated using as many evolutionarily distinct lineages (i.e. taxa) from a group as possible because the factors that drive intraspecific differentiation also initiate speciation [82]. In sum, we find that increased net diversification is tied to barrier displacement and correlated with large-scale patterns of climate deterioration. We thus found no evidence for hypotheses of pure stochasticity or diversity dependence, and attribute diversity-independent diversification to a landscape that is both spatially heterogeneous and perpetually changing [29].

Ethics

All samples for this study were obtained legally from various institutions' genetic resources collections listed in electronic supplementary material, table S1.

Data accessibility

Scripts are found at https://github.com/lukemusher/Tityrinae_biogeography. UCE sequences, alignments and gene trees are found at the Dryad Digital Repository: https://doi.org/10.5061/dryad.43n9j1p [83].

Authors' contributions

L.J.M. and J.C. conceived of the study. L.J.M., M.F. and J.M. carried out bioinformatic data processing. L.J.M. wrote all R scripts, designed and carried out analyses. A.L.A. collected preliminary data that were used in designing the current study. L.J.M. wrote the manuscript. All authors edited and revised the manuscript for intellectual content.

Competing interests

The authors declare no competing interests.

Funding

Funding for this work was provided by NSF 1146248, NSF/NASA 1241066 (Dimensions US-Biota-São Paulo) to J.C. and FAPESP 2012/50260-6 to Lucia Lohmann. Fellowship support to MF was provided by CAPES PDSE (#8881.133440/2016-01).