Historical Signal - Phylogeographic Structure
Recent historical events, particularly Pleistocene glacial-interglacial cycles, have had a critical impact on generating phylogeographic structure in species . Typically, phylogeographic studies have examined how species responded to these events within a single biome or biogeographic region [16–18]. Here, we report how a species distributed across multiple biomes and biogeographic regions of North America responded to the climatic oscillations of the Pleistocene. In the Northern Cardinal, we found six lineages that have been likely differentiating since the Pliocene. Each of the four continental clades is found in a well-established biogeographic area: the Mexican Pacific coast, the Sonoran and Peninsular deserts, eastern North America and the Yucatán Peninsula. Additionally, two insular lineages on Cozumel and the Tres Marías Islands were monophyletic.
Major geological and climatic barriers appear to define the ranges of the mainland clades. The distribution of carneus is constrained by the Sierra Madre del Sur to the east and habitat turnover to the north and south. Igenus shares a range with many co-distributed desert species that border the Sierra Madre Occidental to the east and extend to the northern and southern limits of the Sonoran Desert. This clade does not exhibit the deep genetic structure between the Sonoran and Peninsular desert populations found in other co-distributed species . However, low-levels of genetic structure were detected in igneus, which is likely due to recent differentiation of the putative subspecies found in the Sonoran Desert, Baja California Peninsula, and Tiburón Island . Present-day gene flow among cardinals in these areas is likely limited by their geographic isolation or by ecological constraints , but gene exchange among these groups could have been more regular during the Last Glacial Maximum when sea levels were lower  and populations were in closer proximity.
Igneus is replaced in the Chihuahuan Desert by cardinalis and haplotypes of both clades are found east and west of the Chihuahuan-Sonoran transition zone. This region has been identified as an area where many species exhibit genetic breaks and phylogroups often come into secondary contact . Although our sampling was limited in the transition zone, we found no evidence of geographic overlap in these haplotypes and our genetic data and ecological niche model support an earlier conclusion that these forms do not hybridize . The present-day ecological niche model found the geographic area between these clades unsuitable for either group, but the paleoecological niche model indicated a more continuous distribution across this gap during the Last Glacial Maximum.
The evolution of the cardinalis clade in eastern North America has been a history of population connectivity. Remarkably, cardinalis, with a range of over 3,000,000 km2, has no appreciable geographic structure in mtDNA, an unexpected pattern for a widely distributed clade that has several described subspecies. Comparative phylogeographic work on other widely distributed avian species has shown a similar pattern of unstructured mtDNA in eastern North American birds . Alternatively, a diverse array of taxa including birds distributed in the southeast U.S.A. exhibit distinct phylogeographic structure that was shaped by a diverse array of historical factors .
The range limits of cardinalis and coccineus in eastern México are not well understood, but the mtDNA break coincides with the biogeographic barrier separating the Yucatán Peninsula biotic province from surrounding regions . Morphological work has identified an area of turnover where cardinalis grades into coccineus in Tamaulipas . Based on genetic data presented, however, specimens from Tamaulipas and areas farther south in Querétaro and Veracruz, México all possessed cardinalis haplotypes. Much of the Gulf Coast along México is unsuitable for cardinals because it has been converted to agriculture or encompasses low-lying areas that periodically flood. The ecological niche model also showed that habitat along the Gulf Coast to be sub-optimal. Additionally, coccineus appears to be a recently diverged lineage and is paraphyletic with respect to the monophyletic saturatus, a pattern consistent with incomplete lineage sorting between young groups .
The two island lineages, mariae and saturatus, are currently classified as subspecies in the igneus and coccineus groups, respectively, and both are monophyletic, suggesting that they are independently evolving lineages. Despite Cozumel being only 20 kms from the mainland and remerging 120,000 years ago , its biota has become distinct, with eight endemic bird and mammal species. Mariae occurs on the Tres Marías Islands, a small island chain 100 km off the Pacific coast of México that was likely submerged in the Late Pleistocene  and has several endemic subspecies . Molecular dating indicated that the mean mtDNA estimates of these island lineages are "older" than the islands. The divergence of the Tres Marías Islands lineage from the mainland was not congruent with the island's geology; the 95% HPD (0.31-2.86 million years ago) covers a time period before the island was presumably under water. But, the low end of 95% HPD for the divergence of the Cozumel Island (0.07-1.06 million years ago) is consistent with geology. The incongruence between the ND2 gene tree and geology may be due to an elevated island substitution rate , an undetected time-dependent substitution rate , or coalescent variance . Understanding the age of theses island lineages will require more thorough molecular dating using multilocus data.
Mitochondrial DNA as a Phylogeographic Marker
Mitochondrial DNA is a powerful marker for identifying independently evolving populations within species because of its fast mutation rate and maternal inheritance . However, incomplete lineage sorting , rate variation , and selection  can limit the utility of mtDNA as an evolutionary marker. Overall, the phylogeographic structure presented in this study is largely congruent with the morphological groups previously described for the species. These mtDNA lineages reflect a long history of isolation and independent evolutionary trajectories that merit further evaluation using multilocus coalescent methods .
Genetic diversity estimated from mtDNA is impacted by a number of factors, such as population size, age, genetic drift and mutation rate . By using range size and genetic distances as proxies for population size and time of isolation, we were able to show that nucleotide diversity in C. cardinalis was more strongly coupled to range size than genetic distance. This relationship between range size and genetic diversity was expected because larger ranges have the potential for larger populations and thus are able to accrue higher levels of genetic diversity . But, this finding is important because it suggests that mtDNA genetic diversity is at least partially coupled with cardinal population size, and indicates that the estimates provided here have not been strongly affected by selection and linkage to the W sex chromosome  or genetic draft .
Post-glacial Expansion: Historical to Contemporary Demography
The leading edge model of post-glacial expansion predicts there will be lower genetic diversity in recently colonized regions . Despite strong evidence for this model in some taxa and some geographic areas, there has not been compelling evidence for this pattern in North American birds [, , ; but see ]. These results are often attributed to other factors such as mixing of separate refugial populations, un-sampled populations, or rapid expansions. Both our genetic data and our ecological niche modeling suggest a population expansion for the cardinalis.
Given that the cardinalis clade likely expanded out of a glacial refugium, it is surprising that we found nucleotide diversity, haplotype diversity and the frequency of private haplotypes to be uniform across cardinalis sampling localities. Intuitively, homogeneous genetic diversity would seem attributable to birds being more vagile than other vertebrates, but this is unlikely the case here. Many birds are constrained by the same ecological barriers as non-volant organisms . This ecological limitation of dispersal is evident in igneus, which has also undergone a recorded northern range expansion. Our igneus samples from Arizona/New Mexico had lower haplotype and genetic diversity than samples from Sinaloa. A critical difference between cardinalis and igneus is the amount of habitat connectivity across their ranges. There are few major dispersal barriers for cardinals in eastern North America, allowing the species to freely disperse in any direction. On the other hand, in the areas of the arid Sonoran Desert where cardinals have recently colonized, they are more restricted to habitats with suitable vegetation such as riparian zones.
The high genetic diversity in recently established populations of the eastern cardinalis contradicts the leading edge model and begs explanation. One suggested alternative is the Phalanx model, which posits that large populations expanded slowly into newly available habitats with no loss of genetic diversity in newly colonized areas . Although the Phalanx predicts homogenous genetic diversity as seen in cardinalis, we see no reason for cardinals to have expanded their range slowly. Indeed, the rapid northward range expansion documented in the last 150 years suggests cardinals have high dispersal rates and can rapidly populate suitable habitat. Why may this be the case?
We suggest that cardinals in eastern North America were held at low numbers by a shortage of suitable habitat prior to the European settlement of North America. Throughout their range cardinals inhabit brushy open habitat, and prior to European settlement this ecotonal species lived in a sea of eastern hardwood forest where suitable habitat was patchy and ephemeral, created by water courses, tree falls, and fires. Throughout most of the evolutionary history of eastern cardinals, suitable habitat must have been ephemeral, leading to selection for high dispersal ability. Once European settlement converted most of Eastern North America into a vast ecotone of excellent cardinal habitat, their evolved propensity for high dispersal would homogenize genetic diversity across vast areas.
The hypothesis that cardinals exhibit high dispersal abilities is borne out by an analysis of banding records for cardinals . Although conducted before root mean square dispersal was being estimated, Dow and Scott  found that 190 of 1523 recovered cardinals were found outside the 10-minute block of latitude and longitude where they were banded; among those that moved beyond their block, first years moved an average distance of 60 km and adults moved an average of 130 km . These are large distances for a resident bird , and comparing them with results for North American blackbirds analyzed in modern ways  suggests that root mean square dispersal for Northern Cardinals is very high, possibly exceeding 100 km. Dispersal distances of this magnitude would be sufficient to homogenize genetic variation in a short period of time.
What these data suggest is that during the lag between historical expansion and the present-day, contemporary demography has played a role in shaping the observed genetic diversity pattern. Population densities are lowest at higher latitudes and population growth increases with latitude. We suggest that this population growth is not just from offspring staying in one area and increasing the density of individuals. Instead, populations are also growing as new individuals with different haplotypes are moving into new areas. Given that genetic diversity is uniform even in the youngest cardinal populations, at the highest latitudes, the lag between expansion and genetic homogenization may be on the order of decades instead of hundreds or thousands of years. As new areas became inhabitable, dispersal would likely not have been just in a northern and linear fashion, but east and west. If cardinals expanded out of southern glacial refuges rapidly and in large numbers, the signature of decreased genetic diversity may have been rapidly erased or alternatively, potentially never existed at all.