SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Separated throughout most of the Cenozoic era, North and South America were joined during the mid-Pliocene when the uplift of Panama formed a land bridge between these two continents. The fossil record indicates that this connection allowed an unprecedented degree of inter-continental exchange to occur between unique, previously isolated biotic assemblages, a phenomenon now recognized as the “Great American Biotic Interchange”. However, a relatively poor avian fossil record has prevented our understanding the role of the land bridge in shaping New World avian communities. To address the question of avian participation in the GABI, we compiled 64 avian phylogenetic studies and applied a relaxed molecular clock to estimate the timing of trans-isthmus diversification events. Here, we show that a significant pulse of avian interchange occurred in concert with the isthmus uplift. The avian exchange was temporally consistent with the well understood mammalian interchange, despite the presumed greater vagility of birds. Birds inhabiting a variety of habitats and elevational zones responded to the newly available corridor. Within the tropics, exchange was equal in both directions although between extratropical and tropical regions it was not. Avian lineages with Nearctic origins have repeatedly invaded the tropics and radiated throughout South America; whereas, lineages with South American tropical origins remain largely restricted to the confines of the Neotropical region. This previously unrecognized pattern of asymmetric niche conservatism may represent an important and underappreciated contributor to the latitude diversity gradient.

The continents of North and South America have been isolated from one another throughout most of their histories. The fossil record indicates that a transient Beringian connection between North America and eastern Asia existed throughout much of the Cenozoic (Simpson 1947, Tiffney and Manchester 2001). Unlike North America, South America was an island continent during this time, allowing for the evolution of an ancient and largely endemic biota (Patterson and Pascual 1972, Simpson 1980, Vuilleumier 1984, Ricklefs 2002). Species assemblages unique to each continent met abruptly some 3 million years ago (mya, Coates and Obando 1996) upon the final uplift of southern Central America and the formation of a new terrestrial corridor, the Panamanian land bridge. This uplift initiated a process that led to unprecedented ecological and evolutionary consequences for previously isolated biotas (Simpson 1980), an event now referred to as the Great American Biotic Interchange (GABI).

The mammalian response to the land bridge formation was written from the rich fossil deposits on both continents. Among the many evolutionary inferences gleaned from the mammalian fossil record are: 1) immigrant taxa appeared on each continent soon after the formation of the Panamanian Isthmus (Webb 1976); 2) an early wave of xeric-adapted species was followed by a second wave of mesic-adapted species (Webb and Rancy 1996); 3) there was an equal exchange of taxa between continents (Marshall et al. 1982, Webb and Marshall 1982); and 4) most exchange had ceased by the onset of the mid-Pleistocene (Webb 1976). Over time, lineages moving into South America were more “successful” than those moving in the opposite direction. Today, >50% of modern South American genera have North American origins whereas only about 10% of North American mammal genera are derived from southern immigrants (Webb and Marshall 1982).

Avian bones are thin and light relative to those of mammals and as a consequence fossilize poorly. The depauperate nature of the avian fossil record precludes a direct fossil-based comparison of the responses of birds and mammals to the land bridge formation (Mayr 1964, Olson 1985, Vuilleumier 1985). The few avian fossils available do suggest some interchange during the late Cenozoic, although the timing is unclear (Vuilleumier 1985). Furthermore, there is the perception that because birds can fly, their dispersal abilities differ from non-volant organisms such as terrestrial mammals (Lomolino et al. 2006). The colonization of distant islands by birds provides ample evidence that they can cross water barriers. The over-water dispersal abilities of some groups of birds are well known (“tramps” sensu Mayr and Diamond 2001; Turdus thrushes, Voelker et al. 2009). However, other species of birds, such as those found in the Neotropical forest understory, do not readily disperse across water gaps (Hayes and Sewlal, 2004Moore et al. 2008) and the propensity for such dispersal is simply unknown for most birds. Thus, one cannot determine from the evidence at hand whether avian lineages regularly crossed between North and South America prior to the land bridge completion.

Intercontinental dispersal may be approximated by identifying trans-isthmus diversification events in well resolved molecular phylogenies and then applying a molecular clock. Although, molecular clocks are not without controversy (see Methods) they provide a previously unavailable perspective on how birds responded to the presence of a terrestrial corridor connecting these previously isolated continents. We approach this question in two ways. First, we used molecular data and a phylogenetic tree-based approach to summarize the timing and patterns of recent avian trans-isthmus exchange. We then compared discernible patterns to those inferred from the fossil record for mammals and we evaluated dispersal direction and the influence of elevation and habitat preference on exchange. Second, we used a taxonomic approach in which we summarized and contrasted the historical and present-day distributions of Nearctic and Neotropical families to understand how the deep history of avian groups has affected present-day diversity patterns of New World birds. When combined, our analyses allowed us to determine the relative role of the Panama land bridge formation on avifaunal exchange at different temporal and taxonomic scales.

Materials and methods

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We performed a literature search for molecular phylogenies on New World avifauna. Studies were selected if they had complete or near complete taxon sampling and if the constituent taxa included both North and South American lineages. Suitable datasets were downloaded from Genbank using the program Geneious v. 2.5.4 (Drummond et al. 2006). In all, we compiled 64 molecular phylogenetic studies of New World birds, summarizing those events in which sister lineages have diverged after intercontinental dispersal. The birds that we studied are an ecologically and taxonomically diverse group that includes representation from 11 orders, 34 families, and over 100 genera (complete taxon list provided in Supplementary material).

Range evolution

In order to identify trans-isthmus diversification events we reconstructed phylogenetic trees for each dataset in order to perform range evolution analyses. To select a model of sequence evolution for each dataset we used the program MrModeltest v. 2.3 (Nylander 2004) and the Akaike information criterion (AIC). We constructed Maximum Likelihood phylogenetic trees in the program PAUP* 4.0b 10 (Swofford 2002) using the heuristic tree search, (TBR branch swapping and 10 random addition replicates). Tree topology was verified by comparison to the original published tree. We then performed a likelihood ratio test for clocklike evolution. For trees that were not clocklike, we used the program BEAST v. 1.4.8 (Drummond and Rambaut 2007) and applied a relaxed molecular clock (molecular rates are discussed below) with a Yule process speciation prior. Each analysis was initially run for 10 000 000 generations and sampled every 1000. We used the program Tracer v. 1.4 (Rambaut and Drummond 2007) to assess the mixing of MCMC chains in the analyses and to determine the burn-in. If the runs did not converge or we did not achieve reliable ESS values (>200) we re-ran the analysis for more generations. All trees generated prior to the point of stationarity were discarded. In cases where the greater topology was unresolved, we reduced the dataset to the clade of interest.

To estimate the direction of dispersal we employed the dispersal-extinction-cladogenesis (DEC) model (Ree and Smith 2008) in the program Lagrange 2.0.1 (Ree et al. 2005). We defined six geographically relevant New World regions; North America (from Panama, north of the Darien, to Alaska), South America, the Greater Antilles, the Lesser Antilles, the Hawaiian Archipelago, and the Galapagos Archipelago. For each dataset, we set the basal divergence time to 1.0 because we were only interested in the relative timing of events for this particular analysis. We chose not to place dispersal time constraints on the analyses in order to optimize the data and allow all practical range evolution scenarios to occur. On trees that had an unresolved pre-isthmus ancestral area, we placed an ancestral area constraint to reduce uncertainty concerning the geographic origin of the clade. To justify the use of such a constraint, we performed ancestral state reconstruction using parsimony in Mesquite v. 2.6 (Maddison and Maddison 2009). Constraints were applied only if ancestral state reconstruction inferred that the basal node in a phylogeny was unambiguously assigned to one continent, and that node was older than 4 million years. This procedure eliminated a discontiguous ancestral area in both North and South America prior to the isthmus uplift.

Diversification times

For each phylogeny, we identified diversification (cladogenetic) events between North and South American taxa that represent a dispersal event across the isthmus. We recognize that a lag time between dispersal and diversification events exists, and that the duration of this time varies across species depending on such factors as ecology and dispersal ability. For this study, we assumed that the lag time between dispersal and lineage diversification was negligible with respect to the evolutionary time scale of the interchange. We reasoned that although that some intercontinental gene flow likely persisted for a short time after dispersal occurred, the relatively linear configuration of southern Central America, the geographic bottleneck at its contact point with South America, and the climatic and geologic turmoil that occurred at the time of the isthmus uplift led to relatively rapid genetic isolation. To improve precision, mitochondrial (mtDNA) sequences of all the descendant lineages arising from each dispersal/diversification event (called diversification events hereafter) were re-run through MrModeltest v. 2.3 (Nylander 2004) to obtain a sequence model that was unaffected by the phylogeny it was nested in. We re-ran each individual dataset through BEAST following the same methodology as described above. To evaluate the robustness of our results, we additionally performed analyses with alternative clock calibration times, a uniform clock, strict clocks, as well as uncorrected genetic distances.

Four different mitochondrial genes were used in the trans-isthmus dataset: cytochrome b (cyt-b), cytochrome oxidase I (COI), NADH dehydrogenase subunit II (ND2) and ATP-synthase 6 and 8 (ATPase6&8). We assessed the variability of rates among mtDNA loci by using pairwise comparisons of average intrageneric genetic distances for each gene used (Fig. 1). These analyses indicated that cyt-b and COI evolve at similar rates and for these we applied the widely used cyt-b rate of 2.0% sequence divergence per million years (0.010 substitutions/site/lineage/million years, [s/s/l/m] Lovette 2004). Our comparisons indicated that ND2 and ATPase6&8 evolve at approximately 1.25 times the rate of cyt-b (slope of regression line, Fig. 1), thus we applied a rate of 2.5% per million years (0.0125 s/s/l/m) for these genes.

image

Figure 1. Plot of relative rates of cyt-b and ND2 genes for 23 genera of New World birds. The dotted line indicates equal rates.

Download figure to PowerPoint

Rate heterogeneity among taxa and across genes has led some researchers to question the general utility of a molecular clock (Pereira and Baker 2006, Nabholz et al. 2009) however, other studies have shown that when analyzed carefully (as we have done here) divergence time estimates can be robust and informative (Weir and Schluter 2008).

We acknowledge other sources of bias in estimates based on a single locus such as mtDNA. For example, we did not correct our data for ancestral polymorphism (Arbogast and Slowinski 1998); nor, did we address time dependent rate variation (Ho et al. 2005). Both issues potentially affect a small proportion of our estimates in the tail of our distribution. Had we addressed these potential errors, a subset of points would be moved toward the present, strengthening our interpretation of the results. We did not account for coalescent variance (Edwards and Beerli 2000). Simulation studies have shown that a single vicariant event can generate a wide range of genetic distances among geminate species pairs (Hickerson et al. 2006). Discrete diversification pulses that were undetected with our analyses may have occurred during the GABI. Finally, any lag time between gene and population divergence (Edwards and Beerli 2000) would cause some of our estimates, particularly those associated with more recent diversification events, to be younger than we estimated.

Isthmus of Panama completion time estimates

The estimated time period of the final isthmus uplift and the subsequent formation of a terrestrial corridor have been inferred from multiple lines of evidence. Although most researchers agree that the Panamanian land bridge was present by the mid-Pliocene, there is no consensus on precisely when this event occurred. Evidence based on marine deposits suggests that the connection between the Caribbean and Pacific was cut off by ca 3.1 mya (Coates and Obando 1996). Radio-isotopic dating from mammalian fossil beds indicates an approximate isthmus closure at 3 mya (Marshall et al. 1979); however, these same mammalian data have been interpreted as evidence for a more recent isthmus closure at 2.5 mya (Stehli and Webb 1985). The most recent geological estimate indicated that North and South America may have been connected as early as 4 mya (Kirby et al. 2008). To account for the uncertainty in isthmus closure times, we used two different estimates for the beginning of the post-land bridge period, the widely used date of 3.1 mya (Coates and Obando 1996) and the most recently published estimate of 4 mya (Kirby et al. 2008).

Phylogenetic tree-based comparative diversification analyses

To test the hypothesis that the land bridge was critical for avian exchange we compared observed diversification time estimates with appropriate null models (Zink et al. 2004). We used two complementary approaches. First we simulated distributions of diversification events using the program Phyl-O-Gen v. 1.2 (Rambaut 2002). These distributions were generated with parameters that would reflect a range of possible diversification scenarios. We used two different constant growth scenarios, moderate speciation (birth rate=0.2, extinction rate=0.2); and high speciation (birth rate=0.8, extinction rate=0.2). We also generated distributions affected by extinction pulses, using a birth rate of 0.2 and episodic extinction events that removed 20% (low extinction pulse) and 80% (high extinction pulse) of the lineages in a simulation. We converted branching events in simulated phylogenies into time using the program LASER (Rabosky 2006) by setting the basal divergence of each simulation to 12 million years. Diversification events were periodically sampled to generate a simulated distribution. To make the distribution comparable to the distribution of empirically derived diversification events, all points less than 0.04 million years were omitted.

We realized that simulated distributions may not capture the complexity of diversification occurring within and among multiple avian lineages. To address this issue, we employed a second approach that used the results from our BEAST analysis to generate a distribution of all diversifications occurring within 28 completely sampled avian genera that are represented in both North and South America. Temporal estimates of all diversifications (n=413) occurring within the 28 selected genera were then compiled into a single distribution that represents an empirically derived estimate of background diversification over time. To allow direct comparison with our distribution of isthmus related diversifications (n=135), we generated a distribution that represents a temporal baseline of cladogenetic events occurring over the last 8 million years. From this distribution we took 100 random draws of 135 diversification events. To compare the simulated and empirical distributions of diversification events with distribution of trans-isthmus events, we used the nonparametric two-sample Kolmolgorov–Smirnov test (K-S test) across the entire temporal distribution (0–8 mya), as well as to pre-land bridge (4–8 and 3.1–8 mya) and post-land bridge (0–4 and 0–3.1 mya) time intervals.

Historical and ecological diversification patterns

We further explored the timing and pattern of dispersal by performing additional K-S tests on direction of dispersal and the potential role of elevation and habitat preference. Following the designations of Stotz et al. (1996), we separated lineages into two broad elevation categories, highland and lowland. The transitional zone between montane and lowland habitats varies across mountain ranges and regions in the Neotropical region (Stotz et al. 1996). To account for this complexity, we defined lowland birds as those never occurring above 2000 m and highland birds as those never occurring lower than 1000 m (fide Stotz et al. 1996). This analysis was used to test the hypothesis that birds occupying montane habitat “islands” may be predisposed to disperse with greater regularity than lowland forest birds. Lineages were also divided into two broad ecological categories based on predefined habitat and foraging strata categories (Stotz et al. 1996). These included birds that are obligately dependent on the vertical structure occurring in true forest habitats (hereafter: forest birds) and “non-obligates,” birds occurring predominantly in successional, edge, and scrub habitats (hereafter: edge/scrub birds). This approach allowed us to use birds to evaluate the ecogeographic model (e.g. alternating periods of rainforest and drier, more open habitat corridors) that has been proposed to explain the mammalian interchange (Webb 1991, Vrba 1992).

Taxonomy-based comparative biogeographic analysis

The timing of avian exchange is not only important for discerning the impact of the Panamanian land bridge but it also provides critical insight into how the ancestral origin of avian groups affected the present-day distribution of New World avian diversity. To examine the contribution of historical assemblages to present-day patterns of diversity we identified New World avian families for which an ancestral origin had been inferred using molecular data. For each family, we summarized the number of species now inhabiting either Nearctic or Neotropical regions (source: Sibley and Monroe 1990). To test whether families with northern ancestral origins have a greater number of species in the Neotropical region than families with southern ancestral origins have in the Nearctic (Mayr 1964), we used a generalized linear model with a quasibinomial error term to account for over dispersion.

Results

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We identified 135 trans-isthmus diversification events that occurred at multiple taxonomic scales including intra-specific, inter-specific, inter-generic, and inter-familial branching events (Supplementary material). Overall, our results indicate that the land bridge was critical in facilitating intercontinental exchange and subsequent diversification; although, a number of geologic, geographic, and climatic factors that are known to have occurred during this time period certainly played a role. We obtained qualitatively similar results irrespective of the clock approach used (results not shown), suggesting that considerable signal is present in these data. Diversification between North and South America occurred from the mid-Miocene through the Late Pleistocene, with 76% of the trans-isthmus diversification events we identified occurring within the last 4 million years. A histogram of trans-isthmus diversification times (Fig. 2) suggests a pulse from 4 to 2 mya, coincident with the land bridge formation. This pulse was followed by a pronounced decrease in the frequency of diversification events during the Pleistocene.

image

Figure 2. Histogram of trans-isthmus diversification events. Diversification times were estimated by employing a gene-specific relaxed molecular clock. The shaded area corresponds to the post-land bridge time period. Taxonomic rank is indicated by color: intra specific (red); inter specific (blue); inter generic (gold); and inter familial (grey).

Download figure to PowerPoint

The distribution of trans-isthmus diversifications differed significantly from all simulations (Fig. 3; Table 1) but we could not reject the background distribution (i.e. all diversifications accruing within 28 genera). However, when our data are partitioned into biologically relevant time intervals, more precise patterns of diversification are recovered. All comparisons for the earliest time interval (4–8 or 3.1–8 mya) indicate that early diversification events between North and South America followed the expected null diversification rate (Table 1). Importantly, this trend strongly shifts in the recent interval (regardless of which isthmus time closure estimate is used), a time attributed to the GABI. During this period, the frequency of trans-isthmus diversifications exceeds those derived from both the simulated and background distributions (Table 1). Within the post-land bridge period, a decline in diversification is evident in the isthmus-related and background distributions, a phenomenon shown to be common in birds (Weir 2006, Phillimore and Price 2008). We suggest that the notably steeper decline apparent in the trans-isthmus distribution reflects a density dependent effect.

image

Figure 3. Relative frequency of simulated, empirical and trans-isthmus diversification events. Moderate speciation (birth rate=0.2, extinction rate=0.2); high speciation (birth rate=0.8, extinction rate=0.2); low extinction pulse (birth rate=0.2) with episodic extinction events that removed 20% of lineages; high extinction pulse (birth rate=0.2) with episodic extinction events that removed 80% of the lineages. “Background” diversification times were estimated for all cladogenetic events occurring within 28 widely distributed (i.e. with elements occurring on both continents) New World avian genera.

Download figure to PowerPoint

Table 1.  Results of comparisons between the distribution of trans-isthmus diversifications and alternative diversification scenarios. To compare the simulated and empirical distributions of diversification events with the trans-isthmus distribution, we used the nonparametric two-sample Kolmolgorov–Smirnov test (K-S test). The test statistic shown is the D statistic, a measure of difference between two distributions.
Distribution modeled0–8 (mya)4–8 (mya)3.1–8 (mya)0–4 (mya)0–3.1 (mya)
  1. *p<0.05.

  2. p<0.001.

Moderate speciation0.61660.25750.17720.58400.5445
High speciation0.36230.22560.12000.36260.3078
Low extinction pulse0.58240.23630.15960.55330.5078
High extinction pulse0.33330.22800.13480.31320.2560*
“Background”0.14950.23360.22720.2129*0.2449*

Within the tropics, the direction of exchange (Fig. 4a) was symmetrical and reciprocal (north to south, n=52; south to north, n=50; K-S test, D=0.195, p=0.259) but different trends emerge following the onset of the Pleistocene. Dispersals into South America slowed during this time while dispersal into North America reached its highest frequency. The role of elevation on trans-isthmus dispersal appears to be negligible. Our comparison between lowland (n=72) and highland (n=21) birds (Fig. 4b) indicates no difference between these groups (K-S test, D=0.2480, p=0.233). We did, however, detect potentially different temporal trends for forest and scrub/edge birds (Fig. 4c). The mean diversification time of scrub/edge birds (n=25; 3.33 mya, 2.55–4.11) was older than the obligate forest birds (n=49; 2.67 mya, 2.07–3.27) although the distributions were not significantly different (K-S test, D=0.248, p=0.224).

image

Figure 4. The frequency of diversification events relative to historical and ecological variables. (a) dispersal direction (K-S statistic: D=0.1946, p=0.259); (b) elevation (D=0.2480, p=0.233); and (c) habitat group (D=0.2482, p=0.224).

Download figure to PowerPoint

When family-level biogeography was examined, we detected contrasting evolutionary histories (Table 2). Of the 12 families we identified as having northern origins, three are constrained today to the tropical region while the remaining nine occur widely across both Nearctic and Neotropical regions. Conversely, of the 16 families we identified as having southern (i.e. South American) origins, most invaded tropical Central America but only two were able to push further north to colonize Nearctic regions. The families with northern origins had a significantly greater proportion of species in the Neotropical region (83.4% of species from northern families now occur in the Neotropical region) than families with southern origins in the Nearctic (2.7% of species from southern families occur in the Nearctic; 2×P [t1>−9.627]<0.001).

Table 2.  Breeding distributions of New World avian families for which ancestral origins were inferred from molecular data.
Ancestral northNo. of species
 NeotropicalNearctic
  1. *Old World members of family are excluded from this total.

  2. Thraupidae is secondarily South American.

Trogonidae*250
Momotidae90
Mimidae348
Vireonidae3813
Corvidae*1817
Polioptilidae114
Troglodytidae669
Thraupidae4020
Parulidae6549
Icteridae7620
Cardinalidae3315
Emberizidae*8135
Ancestral southNeotropicalNearctic
Psittacidae*1500
Trochillidae33118
Ramphastidae350
Semnornithidae20
Capitonidae110
Bucconidae320
Galbulidae170
Furnariidae2350
Thamnophilidae2090
Tityridae310
Tyrannidae40027
Pipridae510
Melanopareidae40
Rhinocryptidae440
Conopophagidae100
Grallaridae490

Discussion

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Birds and mammals, a shared history

Molecular and fossil based evolutionary reconstructions can be affected by forms of bias unique to each approach, making direct comparisons between such studies difficult. Nevertheless, the avian response to the Isthmus of Panama formation that we describe shares a close resemblance to the late Pliocene interchange described for mammals. The temporal consistency between the avian and mammalian histories underscores the dramatic impact that the newly formed terrestrial corridor had on the two previously isolated biotas. For both birds and mammals, some lineages were able to disperse across the water gap prior to isthmus completion; although, in both groups intercontinental exchange for the vast majority was facilitated by the presence of the land bridge. While birds can fly, our result suggests that most birds are subject to the same dispersal constraints as their mammalian counterparts.

The similarities between the avian and mammalian responses to the isthmus completion are also seen in the patterns of interchange. The shared decline in trans-isthmus dispersal over time likely reflects a real pattern that may be driven by the unavailability of ecological space. We suggest that this pattern is the result of a density dependent effect (the priority effect, MacArthur 1972), such that exchange across the isthmus would have decreased over time due to competition with congeners (or diversifying conspecifics) that had already crossed the isthmus. Other previously identified factors that may lead to a genetic signature of density-dependent diversification, such as extinction and inadequate taxon sampling (Rabosky and Lovette 2008), are not easily explained by the data. An increase in extinction rates that would cause a decline in diversification is not likely. There is no a priori prediction that would allow extinction to differentially affect recent colonists (last 2 million years) at a much higher rate than that experienced by earlier colonists. Inadequate taxon sampling is also not likely, as 35% of the points in the data are intra-specific genetic breaks identified by widespread geographic sampling.

Our data suggest that by the onset of the Pleistocene, the majority of species in geographic proximity to the land bridge had already crossed, even though source pools continued to receive new immigrant species from the interchange. Once these source taxa immigrated, their descendants only rarely dispersed back into their continent of origin. This is evident by the infrequency of backcrosses between continents, with 85% of the points represented by single dispersal events. Nearly half of the lineages that did backcross represented older events at higher taxonomic levels. It is unlikely that our results are confounded by sampling effects such as over-emphasizing highland taxa (at the expense of more continuously distributed lowland species) or by selective study of only those taxa presumed a priori to possess complex population structure. Our data include intraspecific divergences for 34 lowland taxa that are presumed to be continuously distributed across the Isthmus of Panama. In the course of summarizing the relevant available data we encountered only one instance of haplotype sharing across the isthmus, suggesting that most lowland taxa do differ across this biogeographic break and that diversification occurs soon after trans-isthmus dispersal.

The similarity between the highland and lowland distributions is somewhat surprising given that the distributions of many highland birds are fragmented by intervening lowland valleys. Our result may indicate that highland birds are able to readily disperse through lowland corridors during interglacial habitat shifts. Regardless, additional data may be required to further clarify the role of elevation on trans-isthmus dispersal. We identified only 21 highland diversification events compared to seventy-two for lowland birds. The disparity in sample size between highland and lowland birds suggests that the land bridge corridor had a lesser impact in facilitating dispersal for highland species because the corridor only directly connected lowland habitats.

The mammalian GABI is known for the exchange of savanna species despite the present-day lack of a dry habitat corridor between the Americas. Our dataset contains several non-rainforest species (xeric, scrub, and edge taxa) that would have difficulty dispersing through the extensive rainforest on the isthmus today. The existence of historically more continuous dry habitats within the Neotropical region (Pennington et al. 2000) and a drier isthmus habitat corridor (Webb 1991) coincide with a known Pliocene warming period (Zachos et al. 2001). Thus, it seems reasonable that large mammalian herbivores and birds adapted to more xeric habitats may have dispersed at this time with true forest birds crossing sometime later.

Although our analysis lacked statistical significance, differing temporal trends for these two ecological groups are apparent (Fig. 4c). Scrub/edge birds have an older mean diversification time and the frequency of dispersal declines in the Pleistocene which is consistent with the early wave of large mammalian herbivores. In contrast, forest birds show a pattern similar to that seen for a second wave of mesic phase mammals (Webb and Rancy 1996), with both groups reaching their highest frequency in the Pleistocene. As rainforest continued to expand on the land bridge, many potential immigrants (xeric, scrub, and edge species) from the north would have been cut off, likely contributing to the slowdown in birds dispersing southward. Assuming our interpretations are correct, the decline in the rate of diversification events was driven in part by density dependent effects and also by habitat changes that may have prevented edge/scrub taxa from continuing to participate in the GABI. Despite the overall slowdown in avian exchange our results indicate that for tropical South American birds, the GABI does not appear to be over.

Intercontinental avian exchange and the latitudinal diversity gradient

Overall, the exchange of lineages in response to the Isthmus of Panama closure played a major role in assembling the present-day Neotropical avifauna. The molecular record shows that within the tropics, the number of birds dispersing between the continents in each direction was similar; however, from a broader geographic perspective this symmetry disappears. For both birds and mammals, the GABI had relatively little affect on diversity in temperate North America. We suggest that this phenomenon is best explained by an asymmetry in niche conservatism and this asymmetry has contributed to the long-recognized latitudinal diversity gradient (Mittelbach et al. 2007).

Avian families with northern origins occur widely in both the Nearctic and Neotropical regions, thus northern lineages have successfully diversified throughout South America (Table 2). Conversely, the southern families that participated in the GABI remain almost entirely restricted to the Neotropical region, as they have been for tens of millions of years. As a specific example, the oscine and suboscine passerines are a large and diverse group representing over one half of all extant avian species (Sibley and Monroe 1990). Based on molecular dating, the oscines arrived in the New World between 34 and 14 mya, via multiple independent dispersal events, likely through Beringia (Barker et al. 2004). The suboscines have occupied South America some 65 million years, prior to its drifting from Antarctica and Australia (Barker et al. 2004). Despite having northern ancestral origins, 55% of New World oscine species now breed in South America, many of them in tropical habitats. In contrast, only 2.4% of suboscines have secondarily adapted to North American temperate zone habitats (Table 2).

The difficulty tropical organisms have in colonizing seasonal temperate zone environments is widely recognized. Niche conservatism at this scale, is an important component of the Tropical Niche Conservatism hypothesis (TNC), a model invoked to help explain the latitudinal diversity gradient (Wiens and Donoghue 2004). The observed asymmetry is also consistent with another component of the TNC model, the “time for speciation” effect (Stephens and Wiens 2003). Most ancestrally southern bird lineages now occurring at the northern limits of the Neotropical region, have arrived since the time of isthmus uplift and may not have had sufficient time to adapt to temperate conditions. Conversely, the ability of the northern birds to repeatedly invade the tropics may be linked to their long exposure to tropical conditions in Central America well before the uplift of the isthmus.

One part of the TNC hypothesis that does not entirely explain the distribution of New World avian biodiversity is the role of shifting tropical habitats. During the early Tertiary, tropical forests were once much more widespread across the globe and have since contracted to equatorial latitudes (Behrensmeyer et al. 1992). This contraction along with the inability of tropical organisms to adapt to extratropical habitats has been argued be to a major driver of the latitudinal diversity gradient in New World birds (Hawkins et al. 2006). Support for this pattern in the Old World is seen in fossil evidence that indicates relatives of the present-day Ethiopian (Afrotropical) avifauna were once distributed in northern Europe and the United States (reviewed in Feduccia 1999, Ksepka and Clarke 2009). But, in the New World proper, neither fossil evidence (Vuilleumier 1985) nor molecular data support a unified contraction of the Neotropical avifauna, because the North and South American avifauna remained largely isolated from one another until the Panamanian land bridge formation (Fig. 2). The early Cenozoic North American tropical avifauna is largely extinct and has contributed very little to the extant Neotropical avifauna. We were able to identify only two Neotropical families (Trogonidae and Motmotidae) which are presumed to have northern tropical origins (Feduccia 1999, Witt 2004, DaCosta and Klicka 2008). The contrasting biotic histories of the Old and New World are reflected in their geology. The Old World continents have a long history of connectivity which is in sharp contrast to the prolonged isolation of South America from North America.

Conclusion

Our study suggests the formation of the Isthmus of Panama was critical for facilitating avian exchange in a manner similar to that known for mammals. This exchange, contributed to a reconfiguration of the taxonomic composition of New World avian communities. The GABI was the process that united the isolated North and South American tropical avifaunas into a single biogeographic unit, the Neotropical region. However, the GABI had little impact on avian diversity in the Nearctic. This asymmetric exchange between the Nearctic and Neotropical regions provides an additional insight into why species diversity is higher towards the equator. The interchange of birds from ancestrally northern families increased avian diversity in the tropics, whereas, birds from the ancient South America avifauna contributed little to extant diversity in the Nearctic. We suggest that the ability of organisms with Nearctic origins, to colonize and radiate within the tropics may represent an important and underappreciated contributor to the latitudinal diversity gradient.

Download the Supplementary material as file E6335 from <www.oikos.ekol.lu.se/appendix>.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We thank Jef Jaeger, Robert Zink, Gary Voelker, Garth Spellman, Matt Miller, Brett Riddle, the UNLV Systematics group, and four anonymous reviewers for their comments on this manuscript. We acknowledge Richard Ree, who helped us with our Lagrange analyses, and Jason Weckstein for the use of his unpublished data. Jorge Perez and Daniel Cadena provided critical specimen material and Cheryl Vanier provided assistance with statistical tests. Some of the analyses were performed on the Computational Biology Service Unit at Cornell Univ., an entity funded in part by the Microsoft Corporation. We thank the field collectors, specimen preparators, and curators whose efforts contributed to this study. This work was supported in part by an NSF grant (DEB 0315469, to J.K.) and research funds provided by the UNLV Museum Foundation.

References

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References