The application of phylogenetics has stimulated the field of community ecology (Webb et al., 2002; Cavender-Bares et al., 2009). Beyond informing us on the nature of community assembly in the present, however, phylogenetic community ecology also holds the promise of integrating deep evolutionary history to understand the origin of communities in relation to geographic and climatological changes across tens of millions of years (Emerson and Gillespie, 2008). In this spirit, we envision the eventual merger between the fields of “community ecology” and “biogeography.” More specifically, we imagine a time in the near future where every species within a community can be placed into its broader phylogenetic context, allowing us to pinpoint each species time and place of origin, and its broader pattern of trait evolution and diversification. From an ecological perspective, biogeographic approaches can provide insight into the origin of the larger species pool for a given region. And from a biogeographic perspective, community ecology can provide insight into the ecological processes that structure and maintain diversity within the same region.

In biogeographic studies, a common way to delineate the formation of a species pool is to focus on the larger geographic region or a particular biome within that region. A major topic along these lines, but one that has not been sufficiently treated for most biomes and taxa, involves the role of migration (originating ex situ) vs. diversification (originating in situ) in shaping a regional biota (Emerson and Gillespie, 2008). These processes are not mutually exclusive, of course, but represent ends of a spectrum. Yet, the extent to which either of these processes dominates remains poorly understood. Classic island biogeography predicts that migration will dominate in regions that are both new or near a source species pool, while diversification will dominate in regions that are old or isolated (MacArthur and Wilson, 1967; Emerson and Gillespie, 2008; Losos and Ricklefs, 2009). An example of the former is the origination of Oxalis diversity in the Atacama Desert, where the modern Oxalis species pool was formed through multiple dispersals into the Atacama Desert region from geographically adjacent, likely pre-adapted lineages (Heibl and Renner, 2012). In contrast, the remarkable species diversity in the Andes is often attributed to prolific in situ diversification (Bell and Donoghue, 2005; Hughes and Eastwood, 2006; Antonelli et al., 2009). The Andean uplift and the development of more temperate environments effectively isolated this region from its surrounding tropical species pool (i.e., “continental island” effect) and permitted the colonization of temperate lineages via a small number of dispersals events (often only one from within a single larger clade), which subsequently radiated. Understanding how the balance between dispersal and diversification for a given biome changes with time, however, remains poorly understood.

A major component necessary for understanding the balance between migration and diversification is the degree to which lineages are pre-adapted to newly inhabited regions (Donoghue, 2008). Along these lines, it has been hypothesized that given the prevalence of phylogenetic niche conservatism in plants, species may more frequently migrate into regions to which they are pre-adapted, i.e., “it is easier to move than to evolve” (Donoghue, 2008). For instance, lineages from a clade that share a pre-adaption to a new biome are more likely to establish there, in which case the biome's species pool is likely to be assembled through multiple dispersal events. In contrast, lineages from a clade that are not pre-adapted will be required to adapt subsequent to their arrival to successfully establish. To the extent that in situ adaptation is difficult, lineages that disperse into the biome will fail to establish, thus favoring in situ diversification from a smaller number of dispersals. Perhaps the most striking evidence of the former scenario was presented by Crisp et al. (2009), who identified a broad pattern of phylogenetic bias in the tendency for lineages to disperse and establish across ecologically similar biomes.

Short of a complete phylogeny for both the biome and its resident species, it is most effective at this stage to focus on a representative clade that forms an important component of a biome's diversity. Here, we investigate a biome of high diversity, seasonally dry tropical forest (SDTF) in Mexico, and a clade of flowering plants, Malpighiaceae, that represents an important component of this diversity. Mexico is one of the top ten megadiverse countries in the world and contains a variety of tropical biomes (Williams et al., 2001). Among its most prominent and distinct biomes is the widespread SDTF, which blankets the Pacific slopes of Mexico, ranging from central Sonora and southeastern Chihuahua to the southern state of Chiapas southward into Central America (Figure 1). SDTF is determined by precipitation patterns characterized by total annual precipitation less than 1800 mm yr⁻¹ and a distinct seasonally dry period with less than 100 mm over 5–6 months (Murphy and Lugo, 1986; Gentry, 1995). At 1800 mm yr⁻¹, SDTF falls below the threshold of what defines a tropical rainforest (2000 mm yr⁻¹). What further distinguishes SDTF is their striking seasonality in precipitation, which results in a distinctly green period for half of the year, followed by an equally long dry period during which many species lose their leaves. In terms of their physiognomy, Mexican SDTF is characterized by low to medium trees in which grasses are a minor component (Pennington et al., 2000).

Details on the origin of the Mexican SDTF flora, and the timing of its expansion, have received recent attention (reviewed in De-Nova et al., 2012). The challenge of understanding the origin and expansion of this biome is attributed to scant fossil evidence, as well as, complex geological changes that created the conditions necessary for the biome's characteristic climate (Graham, 2010). Geologically, the establishment of the SDTF in Mexico likely began with the advent of mountain building in southern Mexico during the late Eocene. Today, the north-south Sierra Madre Occidental and the east-west Neovolcanic mountain chains greatly maintain the climatic conditions of this biome, especially by blocking cold fronts from the north. The last uplift of the Sierra Madre Occidental was between 34 and 15 Myr, while the Neovolcanic mountain chain was thought to have been established more recently and in several stages, from west to east, beginning ~23 Myr and ending only recently, 2.5 Myr. Thus, the formation of SDTF in Mexico likely occurred slowly and steadily over a period spanning at least 34 Myr, coincident with these major mountain building events. Becerra (2005) used a time-calibrated phylogeny of the prominent Mexican dry forest clade Bursera to assess the origin and expansion of the SDTF flora. Her findings indicated that the oldest Mexican Busera began to diversify between 30 and 20 Myr in western Mexico, in concert with the origination of the Sierra Madre Occidental. In contrast, younger lineages (>17 Myr) diversified in south-central Mexico, with the expansion of the east-west Neovolcanic mountain chain. These results suggest that SDTF formation was greatly facilitated by mountain building in this region and that this biome first established in western Mexico during the Oligocene and subsequently expanded south and east in Mexico, and eventually to Central America (Graham and Dilcher, 1995; Becerra, 2005).

Malpighiaceae are a pantropical clade and include ~1300 species, 90% of which are found in the New World (Davis et al., 2001; Anderson et al., 2006 onwards; Davis and Anderson, 2010). The family is thought to have originated in South America (Anderson, 1990, 2013; Davis et al., 2002b, 2004), an hypothesis that has been corroborated more recently with greatly expanded phylogenetic sampling (Cai et al., unpublished results). The clade has received broad phylogenetic and biogeographic attention (Cameron et al., 2001; Davis et al., 2001, 2002a,b, 2004; Davis and Anderson, 2010), including more focused investigations in the Old World, especially in Africa and Madagascar (Davis et al., 2002a). Efforts to determine finer scale patterns of Malpighiaceae biogeography in the New World, however, have been hampered by taxonomic sampling deficiencies. Along these lines, one area that is ripe for exploration is the Mexican Malpighiaceae flora, which include ~162 species (Anderson et al., 2006 onwards; Anderson, 2013). Malpighiaceae are especially abundant in Mexico's SDTF (Gentry, 1995), where they are represented by ~60 species, putting them in the top five most diverse families in SDTF (Lott and Atkinson, 2006). Based on taxonomic and phylogenetic grounds, Mexican Malpighiaceae have been hypothesized by Anderson (2013) to represent as many as 42 independent origins from outside of Mexico.

The first goal of our study is to test Anderson's hypothesis on the origin of Mexican Malpighiaceae by greatly expanding current taxon sampling for the family (Davis and Anderson, 2010). This will not only facilitate a rigorous assessment of the number of introductions to Mexico (including those inhabitants of SDTF but more broadly in this geographical region), but will also establish the timing and ancestral origins of the diverse Mexican Malpighiaceae flora. This will enable us to distinguish between more ancient long-distance dispersal events directly from South America versus more recent shorter-distance, “stepping stone” dispersal via Central America. This is especially relevant to understanding when Central America became a major corridor for plant migration (Cody et al., 2010; Gutiérrez-García and Vázquez-Domínguez, 2013; Leigh et al., 2014). A related second goal is to understand how rates of dispersal compare to rates of in situ diversification, and how these processes shaped the species richness vs. their phylogenetic diversity in Mexico. Finally, the third goal is to investigate if lineages were pre-adapted to SDTF, or if they adapted to this novel biome in situ, subsequent to their arrival in Mexico. Relatively few studies have investigated the timing of biome formation using phylogenetic proxies (Becerra, 2005; Davis et al., 2005; Arakaki et al., 2011; Couvreur et al., 2011; De-Nova et al., 2012), yet these approaches hold tremendous promise. It is important to keep in mind, however, using present-day categorizations of biomes as static ecological characters to infer their origin does not fully capture their more dynamic formation over geological time. In particular, different aspects of the abiotic parameters that characterize extant biomes may originate at different times and change at different rates. Here, we seek to elucidate the origination of SDTF in Mexico using two key climate parameters that define this biome today, i.e., overall precipitation (total annual precipitation), and precipitation seasonality (precipitation in the driest quarter). Understanding the dynamics of these interactions is likely to shed key insights into the temporal and spatial nature of the timing of the origin of this biome as it is defined today.

We have structured our discussion below to focus first on the arrival of Malpighiaceae into Mexico as a geographic entity. Here, we detail the broader biogeographic context of these immigrants, including their ancestral areas of origin and timing of arrival. For the second part of our discussion we explore the relative contribution of in situ lineage diversification vs. dispersal-mediated processes in forming the Malpighiaceae species pool in Mexico, which gave rise to SDTF there. And third, we focus on the numerous immigrants to Mexico that have become geographically restricted (i.e., endemic) to this region. Mexican endemic Malpighiaceae are overwhelmingly represented in the SDTF (Figure S7), and thus represent a key to understanding the aridification of Mexico and the formation of this important biome. Here, we focus on characterizing the temporal and spatial nature of the origin of SDTF by investigating if these adaptations arose before or after these lineages became endemic to Mexico.

Mexican Malpighiaceae represent numerous dispersal events from South America. Our analyses indicate that the ~162 species that constitute the Mexican Malpighiaceae flora represent ~33 independent introductions from outside Mexico (Figure 3). These estimates corroborate the large number of introductions to Mexico hypothesized by Anderson (2013) based on his knowledge of the phylogeny and taxonomy of the family. Our ancestral area reconstructions further identify South America as the ultimate source of these many Mexican clades, which is supported by earlier evidence that the origin and early diversification of the family occurred in South America (Anderson, 1990; Davis et al., 2002b, 2004). These findings more generally corroborate the striking floristic similarities between dry-forested regions in northern South America and Mexico (Linares-Palomino et al., 2011).

The migration of South American Malpighiaceae into Mexico falls into two categories. The first involves more ancient northward migrations of South American ancestors directly into Mexico (Figure 3). These six migrations began during the Eocene and continued into the early Miocene (46–19 Myr): the first three occur between the Eocene–mid Oligocene (46–33 Myr) and involve disjunct distributions between South America and Mexico; the remaining three occur between the mid-Oligocene–early Miocene (28–19 Myr) and involve disjunct distributions between South America and Mexico or occupancy of Mexico alone. Based on our ancestral range reconstructions, these migrations do not appear to be facilitated by intervening connections involving Central America or the Caribbean. Taxon sampling bias, while a concern for some lineages, is not a likely explanation for the general pattern we observe (see Materials and Methods). Given the physical distance between these two regions for at least the earliest three dispersal events, long-distance dispersal is likely the predominant mode of migration from South America to Mexico. The three more recent dispersals occur during a time period when we believe Central American corridors were becoming at least partially available for migration, which perhaps explains the increased ambiguity in these ancestral range reconstructions (Figure 3).

The second category of migration from South America to Mexico includes the majority of events, and implicates a “stepping-stone” dispersal scenario, with Central America, and to a lesser extant the Caribbean, serving as a bridge between South America and Mexico (Figure 3). Here, a remarkable 17 of the 33 Mexican Malpighiaceae clades resulting from northward migrations originating in South America had ancestral ranges that expanded to also include Central America. These migratory events begin in the early Miocene ~16 Myr, and continued until as recently as the early Pleistocene, ~2 Myr. Collectively, these findings corroborate a wide range of studies indicating that dispersal plays a major role in tropical forest assembly (Dick et al., 2003; Lavin et al., 2004; Clayton et al., 2009).

A broader geographical framework that more precisely incorporates the timing of these dispersal events into Mexico provides a more nuanced context for interpreting the biogeographic origins of the Mexican Malpighiaceae flora. South American migrations to Mexico occur with striking regularity over an extended period beginning in the Eocene and continuing to the Pliocene (46.4–3.8 Myr). During this period we observe an average rate of 0.8 migrations Myr⁻¹. This average, however, obscures a far more dynamic pattern in the change in the rate of migration. At ~23.9 Myr (Q_95%: 19.5–28.3 Myr) we infer a significant, six-fold increase in the rate of migration between North and South America.

This early Miocene shift in the pattern and rate of migration likely sheds key insights into geological processes that shaped Neotropical biogeography. The Cenozoic paleoland and paleoclimatic reconstructions for Central America, especially involving the Central American Seaway and the rise of the Isthmus of Panama, are contentious (Klocker, 2005; Molnar, 2008). Although numerous lines of geological and biological evidence support a more recent Pliocene shoaling of the Isthmus (Molnar, 2008; Leigh et al., 2014), which established the first direct connection between North and South America, evidence from geology and molecular divergence time estimates in various clades indicate that this barrier was partly permeable to plant migration beginning at least by the Oligocene (Montes et al., 2012; Gutiérrez-García and Vázquez-Domínguez, 2013; Leigh et al., 2014). Our results demonstrate a combination of early long-distance dispersal between South America and Mexico in the Eocene–Oligocene, followed by a shift in the early Miocene, when we see a dramatic increase in the rate of migration, likely due to an increased potential for shorter-distance dispersal. These results suggest that sufficient Central American land corridors were available for plant migration between North and South America, well in advance of when the geological connection between these continents appears to have been fully established ~3.0 Myr (Leigh et al., 2014). Additionally, our analyses of xeric adaption discussed below indicate that these Central American corridors would likely have been characterized by drier forest biomes, in particular, forests with relatively low annual precipitation.

Finally, fruit dispersal syndromes summarized by Anderson (2013) provide ancillary support for these two hypothesized dispersal routes to Mexico–i.e., direct long-distance dispersal to Mexico from South America vs. “stepping-stone” dispersal via Central America. The two clades implicated in older long-distance dispersal events to Mexico from South America, Bunchosia and Byrsonima, are characterized by fleshy bird-dispersed fruits. While the influence of dispersal morphology on the potential for long-distance dispersal remains controversial (Higgins et al., 2003; Nathan, 2006), our results provide ancillary evidence indicating that bird dispersed lineages were more likely to disperse long-distances without an intervening land connection such as Central America. This is corroborated by other empirical studies that have found an increased propensity for long-distance dispersal among bird dispersed plants in the tropics (Hardesty et al., 2006; Jordano et al., 2007). In contrast, the vast majority of Malpighiaceae that successfully made the more recent migrations possess wind-dispersed fruits, including species with winged samaras (Adelphia, Bronwenia, Banisteriopsis, Calcicola, Callaeum, Carolus, Christianella, Cottsia, Diplopterys, Gaudichaudia, Heteropterys, Hiraea, Mascagnia, Psychopterys, Stigmaphyllon, Tetrapterys) and bristly fruits (Echinopterys, Lasiocarpus). These fruits, while not ideal for traversing large bodies of open ocean, were likely sufficient for making shorter step-wise dispersals to Mexico when facilitated by intervening land masses with the rise of Central America. Malpighia, which possesses mostly bird-dispersed fleshy fruits, is an exception in this regard. They are one of the more recent inhabitants of Mexico (expanding into Mexico ~5 Myr) that migrated to that region via Central America. On the basis of fruit morphology, however, Anderson (2013) hypothesized that Malpighia may have colonized Mexico via winged, wind-dispersed ancestors, which is consistent with our broader interpretation of dispersal patterns. A final conundrum is Galphimia, which represents the oldest introduction to Mexico. Although the petals of Galphimia species in Mexico are persistent and thus may facilitate wind-dispersal, their fruits otherwise break apart into dry cocci that exhibit no obvious adaptation for long-distance dispersal (Anderson, 2013).

Dispersal mediated processes dominate initial species pool formation and contribute disproportionately to the phylogenetic diversity of Mexican Malpighiaceae. Our analyses of species diversification rates illuminate the formation of the initial species pool of Malpighiaceae that gave rise to SDTF in Mexico. The vast majority of the ~33 South American lineages that dispersed into Mexico exhibit no evidence of accelerated net diversification rates (Figure S6). Beginning in the Miocene, when rates of migration increased significantly in conjunction with the development of land corridors through Central America, the rate of Malpighiaceae dispersal between South America and Mexico was 1.7 lineages Myr⁻¹, outpacing even the highest rate of net species diversification, 0.3 lineage Myr⁻¹ (Figures 3, 5). This nearly six-fold difference in the rate of dispersal vs. diversification establishes the former as key factor in the initial establishment of the Mexican Malpighiaceae species pool. The lone exceptions to this pattern of relatively low net diversification are the two oldest introductions to Mexico in the Eocene and Oligocene, Galphimia and Bunchosia. These two clades exhibit significantly increased net diversification rates. Although we cannot pinpoint the precise geographic location where these two clades began to diversify, their average diversification rate is on par with the rate of dispersal we observe into Mexico following the development of a Central American migratory corridor.

Our results collectively highlight the balance between in situ diversification vs. dispersal in forming a regional species pool, and how this balance can change with time. Here, the two lineages that demonstrate prolific diversification in the family, Galphimia (22 sp. in Mexico; Anderson, 2013) and Bunchosia (20 sp. in Mexico), constitute roughly 24–28% of the ~162 extant Mexican Malpighiaceae. Beyond these exceptional clades, however, it is remains clear that even the large number of independent dispersal events into Mexico (~33) that formed the initial species pool cannot account entirely for the extant diversity of Mexican Malpighiaceae (~162 sp). Thus, in situ diversification is clearly an important contributor to the extant species richness of this region. At the same time, however, our results indicate that the predominant source of increased phylogenetic diversity (PD) in Mexican Malpighiaceae is attributed largely to dispersal. Using the stem group age of each introduction to Mexico as a conservative assessment of PD generated via in situ diversification (as measured by mean phylogenetic distance; Webb et al., 2002), we found that the mean phylogenetic distance (i.e., the mean branch length between any pair of taxa) between relatives that arose via in situ diversification was 12.5 Myr (Q_95% = 10.4-14.9 Myr). In contrast, mean phylogenetic distance between the Mexican subclades that arose via dispersal into the region was substantially larger, 100.5 Myr (Q_95% = 96.4-104.6 Myr). It is clear from these results that dispersal has contributed substantially to PD in Mexico by effectively “sampling” broadly from the entire Malpighiaceae clade. The contribution of dispersal to PD we identify may be a general pattern facilitating the formation of SDTF in Mexico, and to the formation of regional species pools more broadly. The Oxalis of the Atacama Desert, for instance, represent a broad diversity of Oxalis clades that independently dispersed into this region (Heibl and Renner, 2012).

What limited the majority of introduced lineages from radiating in Mexico? One hypothesis is that newly established lineages did not exhibit a competitive advantage to more recently introduced lineages. Thus, newly arriving lineages were able to occupy SDTF in Mexico with a frequency commensurate with the apparent richness of their source pool outside of Mexico (especially from South America). Support for this hypothesis comes from the fact that we find the vast majority of immigrant lineages arrived pre-adapted to at least one of the two traits that characterizes SDTF, annual precipitation. However, most lineages were not adapted to seasonally dry periods. A second hypothesis is that SDTF is more generally dispersal-limited, thus restricting the ability of lineages to radiate subsequent to their arrival into expanding habitat. Such a scenario might offer newly available niche space for lineages to occupy via dispersal from outside the regional source pool. Previous studies have found a significant degree of genetic differentiation among geographically isolated populations across dry forest habitats, supporting the idea that these dry-forests are dispersal-limited (Pennington et al., 2006a, 2009). Furthermore, at least one of the two lineages (Bunchosia [20 sp.], and less significantly so Malpighia [19 sp.]) that diversified substantially in Mexico are bird dispersed. Such lineages are likely to more easily migrate and radiate across a landscape via allopatric speciation. This is compounded by the fact that the formation of SDTF was not likely contiguous, both spatially and temporally (Graham and Dilcher, 1995). New fragments of SDTF likely arose at different times and in different regions of Mexico, which might have greatly facilitated the continuous establishment of new Malpighiaceae lineages via dispersal from external source pools. In combination, these processes might act to limit the overall diversification of single Mexican introductions, and help explain the pattern of frequent and continued dispersal of new lineages into this biome.

The predominance of dispersal in the formation of a regional species pool has been shown elsewhere, including in the California chaparral (Ackerly, 2004), the Atacama Desert (Heibl and Renner, 2012), and in temperate, post-glacial habitats (Williams et al., 2004). However, these and related studies are commonly restricted to a narrow window of time or to a single, relatively young clade. In our case, the overwhelming pattern of dispersal we identify spans a wide period of time as well as a large, diverse plant clade. This raises the possibility that the characterization of species pools being dominated by single lineages that moved into a region and subsequently radiate, such as in the Andes (Hughes and Eastwood, 2006) or on remote islands (Baldwin and Sanderson, 1998; Emerson, 2002), may be exceptions rather than the rule of species pool assembly. A more general pattern, especially with regard to the phylogenetic diversity of a species pool, might instead be the steady recruitment of independent lineages via dispersal over tens of millions of years, most of which exhibit modest rates of in situ diversification.

Endemic Mexican Malpighiaceae exhibit both ex situ pre-adaptation and in situ adaption to xeric environments in Mexican SDTF. Our analyses of trait adaptation to precipitation provide direct insights for more clearly interpreting the spatial and temporal origins of xeric adaptations by Mexican Malpighiaceae. An emerging question in the field of biogeography is when did traits relevant to community assembly originate (Ackerly, 2004; Cavender-Bares et al., 2009; Simon et al., 2009; Edwards and Donoghue, 2013). Specifically, did Mexican Malpighiaceae evolve adaptations to xeric habitats elsewhere (ex situ, i.e., were they pre-adapted?) followed by movement into Mexico or did they evolve these adaptations in situ in Mexico. This question has been more succinctly phrased as “is it easier to move, than to evolve?” (Donoghue, 2008). To explore this question in Mexican Malpighiaceae, we focused on two key traits central to the definition of contemporary Mexican SDTF: lower total annual precipitation (≤1800 mm) and extreme seasonality in precipitation (≤50 mm over the driest 3 months). Here, we specifically examined the ~22 Mexican lineages that have become endemic to Mexico (Figure 5), since these lineages are overwhelmingly represented in SDTF of this region (Figure S4).

We found that the timing of adaptation to these two key climate traits contrast sharply in relation to when lineages become restricted to Mexico (Figures 4, S4, S5). Adaptation to drier annual precipitation evolved well before Malpighiaceae become endemic to Mexico by an average of ~15 Myr (Q_95%: -1.3 to 64.8 Myr) (Figures 4, S4). Thus, lineages that subsequently became geographically restricted to Mexico were likely pre-adapted for living in relatively drier environments. This climatic adaptation appears to have arisen largely in South America (Figure S4). In contrast, adaptation to extreme precipitation seasonality evolved largely concurrent with or after Malpighiaceae became restricted to Mexico (Figures 4, S5). We interpret these results to indicate that adaptations to precipitation seasonality largely arose in situ in Mexico as the abiotic conditions of SDTF were developing in this region (Figure S5). On average, the lag time for this in situ adaptation to precipitation seasonality was ~0.7 Myr. This pattern is supported by recent findings that identified a lag time in the adaptation of lineages to dry environments following their arrival into the Atacama Desert in South America (Guerrero et al., 2013).

Our results, however, provide a more nuanced perspective of in situ adaptation to dry forest environments and illuminate the timing of the origin and expansion of modern SDTF in Mexico. The first lineage to inhabit Mexico evolved dry forests adaptations ~26.0 Myr. For the remaining lineages, however, adaptation to precipitation seasonality did not arise until much later, around ~13.7 Myr on average. This suggests that although SDTF appears to have arisen in Mexico during the late Oligocene, it was likely geographically restricted at the time. It did not expand greatly until the mid-Miocene. This pattern is consistent with slow and steady north-south mountain building coinciding with the final uplifit of the Sierra Madre Occidental (34–15 Myr), followed by east-west orogeny of the Neovolcanic mountain chain (~23–2.5 Myr), which is hypothesized to have initiated the geological and abiotic conditions that gave rise to SDTF in Mexico (Moran-Zenteno, 1994; Becerra, 2005). Furthermore, this pattern more broadly coincides with global aridification that began during the Miocene, which marked the worldwide expansion of grasslands and succulent biomes (Cerling et al., 1993, 1997; Arakaki et al., 2011). Our results for Malpighiaceae further indicate that the expansion and widespread establishment of SDTF in the mid-Miocene marks a transition in the temporal pattern of adaptation to precipitation seasonality. While older lineages that become restricted to Mexico adapted to precipitation seasonality in situ, younger lineages that become restricted to Mexico (≤13.7 Myr) tended to be pre-adapted to precipitation seasonality (Figures 4, S5). These younger lineages appear to have arisen from ancestors that adapted to precipitation seasonality in Mexico and Central America, but subsequently became geographically restricted to Mexico. These patterns collectively illuminate the timing of the origin and expansion of modern SDTF in Mexico, and indicate that this biome more recently served as an important species pool for other SDTFs across the Neotropics.

It is important to keep in mind that the formation of the SDTF biome in Mexico occurred over millions of years, which appears to have afforded the earlier inhabitants of this region the time necessary to adapt to the novel abiotic conditions that define this biome today. If our hypothesis is correct, we should observe a pattern of more gradual adaptation to precipitation seasonality for dry periods that are consistent with intermediate historic levels (i.e., twice the level of current precipitation seasonality, 200 mm). When we perform this exercise, we observe a clear pattern of earlier adaption to these historic levels as expected under our hypothesis (Figures 4, S5). The extent to which this in situ adaptation occurred within the geographic footprint of the biome itself, vs. geographically adjacent regions in Mexico cannot be resolved with our data. However, we would expect in situ adaptation within the biome for many of these lineages. Further investigations using additional taxon sampling and far better geographic reconstructions of this biome throughout the Cenozoic would be required to determine this more confidently. Additionally, studies of recent range expansions and adaptions to novel environments could shed insight into this scenario. Nevertheless, the lag times we identify raise concerns about using modern biome categorizations as static characters for phylogenetic character state reconstruction (Davis et al., 2005; Crisp et al., 2009). In particular, our study indicates that ancestral reconstructions of biomes as static characters are problematic if the abiotic parameters that define these biomes today are much more recently evolved than the clades that inhabit them. Instead, our results suggest that inferring the evolution of key parameters that characterize these biomes (e.g., annual precipitation and precipitation seasonality) provide far more insight into the evolution of the lineages that inhabit the biome, as well as the biome itself.

The assembly of a biome's flora depends on the complex interplay of dispersal, adaptation, and in situ diversification (Emerson and Gillespie, 2008). These processes in turn depend on both the geographic proximity and ecological similarity between the source pool and the biome in question (Emerson and Gillespie, 2008; Edwards and Donoghue, 2013). Our results additionally demonstrate that these processes are not static and change over evolutionary time. This has important implications for the composition of SDTF in Mexico to the formation of regional species pools and biomes more broadly. In the case of the formation of the Mexican Malpighiaceae flora, we found that changes in the geographic connectivity between South America and Mexico in the Miocene lead to dramatic changes in dispersal patterns from South America. Initially, lineages were limited to less frequent, longer-distance dispersal from South America. But increasingly the availability of Central America permitted more frequent, shorter-distance dispersal, facilitating vastly greater migration to Mexico starting in the mid-Miocene. The increased rate of dispersal, in turn, significantly influenced the composition of Mexican Malpighiaceae. While much of the extant species diversity in Mexico derives from in situ diversification, the overwhelming majority of phylogenetic diversity derives from multiple, independent introductions from South American ancestors spanning the phylogenetic breadth of the Malpighiaceae clade. Furthermore, climatic conditions that define the extant SDTF biome appear to have changed through time. While most lineages arrived in Mexico pre-adapted to one key factor of SDTF, total annual precipitation, these lineages subsequently adapted to another major axis, seasonal dry periods in situ. In this case, in situ adaption appears to have occurred gradually, over many millions of years, coincident with mountain building that established the geological conditions that maintain the biome today. Moreover, we demonstrate that once SDTF becomes widely established in Mexico, it increasingly becomes a source pool for other more recently developed SDTF, especially in Central America.