Increased diversification rates are coupled with higher rates of climate space exploration in Australian Acacia (Caesalpinioideae)

Introduction

Australia's climate was much warmer, wetter, and more stable during the early Eocene (Macphail et al., 1994; Martin, 2006; Zachos et al., 2001; Crisp & Cook, 2013). Mesic biomes and rainforest were more widespread (Martin, 2006). At that time only the north-west displayed aridity, though central Australia likely had seasonal dryness (Martin, 2006; Huber & Goldner, 2012). Since the Eocene, Australia's climate has progressively dried, presenting ecological opportunities and resulting in spectacular radiations within one or more of the Australian biomes (Crayn et al., 2006; Rabosky et al., 2007; Crisp et al., 2009; Sanders et al., 2008; Watanabe et al., 2006; Catullo & Keogh, 2014; Toon et al., 2015; Lamont et al., 2016; Brennan & Oliver, 2017). For example, Australian rainforests now occupy a small fraction of their former extent, and many rainforest lineages are now extinct (Hill et al., 1999; Martin, 2006; Byrne et al., 2008, 2011). In many areas, rainforest taxa have been replaced by sclerophyll and xeromorphic vegetation that now dominates the Australian landscape (Crisp & Cook, 2013). As a result, Australian floristic diversity is unevenly distributed. Many old rainforest lineages comprise a handful of species or less, whereas many sclerophyll or sclerophyll-inhabiting lineages comprise tens or hundreds of species, including Eucalyptus, Leptospermum, Eremophila, Grevillea, Allocasaurina, Pultenaea, and Acacia.

Macroevolutionary patterns have been linked to historical trends in global and regional climate in many Australian lineages, including elevated speciation in arid-zone pygopodoid geckos (Brennan & Oliver, 2017), macropods (Couzens & Prideaux, 2018), cockroaches (Beasley-Hall et al., 2018), and cicadas (Owen et al., 2017). In plants, the eucalypts provide evidence that climate and historical events are important in explaining species distribution and turnover at the continental scale (Ladiges et al., 2011; Bui et al., 2017), and Hakea has expanded coincident with a transition to more open, drought and fire-prone habitats promoted by aridification (Lamont et al., 2016; Cardillo et al., 2017). Aridity also drove the evolution of embolism resistance in Australian Callitris (Larter et al., 2017) and xeromorphy in other lineages (Hill & Brodribb, 2001). By contrast, climate shifting during the Miocene has been linked with a period of depressed phenotypic diversification in several Australian vertebrate lineages (Brennan & Keogh, 2018). These studies, and many others, as summarized by Byrne et al. (2018), emphasize climatic imprinting on patterns of both lineage and trait diversification across Australia's biotic diversity. Generally, but not always, climate has induced disparity in traits and trait diversity, and in diversification rates and lineage diversity.

The relationship between climate and phylogeny can be tested by subdividing the climate–vegetation matrix into two or more discrete biomes (Cardillo et al., 2017). By treating climate as discrete subunits, transition models can be inferred from phylogenies, which inform the history of climate exploration by a lineage from the set of instances when biome transitions occur. Multiple biome shifts have been inferred in the Australian lineage Hakea (Cardillo et al., 2017), suggesting a dynamic interplay between diversification and environmental trait evolution. However, climate is both multidimensional and continuous; lineages can potentially occupy any realized part of it, and may track continuously through climate space over time, even without traversing an imposed subdivision. Predictive distribution models that treat climate as a multidimensional whole have been employed to infer species distribution changes through time (Guisan & Zimmermann, 2000). Recent analytical advances mean that phylogenetic comparative methods can be applied within multidimensional space, including spaces defined by climate variables (Kozak & Wiens, 2010; Adams, 2014; Adams & Collyer, 2018; Adams et al., 2009).

Acacia is Australia's most speciose vascular plant genus, with c. 1045 species (Maslin, 2008; Brown et al., 2012), all but 17 of which are endemic to Australia (Maslin, 2008; Brown et al., 2012; Kyalangalilwa et al. 2013). Acacia occurs throughout Australia, and is a characteristic component of sclerophyll, monsoon, and arid biomes (Murphy et al., 2010). The earliest pseudocolpate fossil pollen assigned to extant Acacia lineages dates from the late Oligocene, c. 23 Ma, and this represents the minimum crown-node age for the lineage (Miller et al., 2013). Where Acacia arose, on or off the continent, has not been resolved (Crisp & Cook, 2013). Nevertheless, an increase in the diversification rate of Australian Acacia c. 15 Ma was identified by Crisp & Cook (2013), coinciding with an inferred global cooling and aridification event following the Middle Miocene Climatic Optimum (Martin, 2006). Overall, Acacia species turnover patterns are correlated more with climate than geochemistry, except in the most species-rich areas; however, aridification is an important process in the underlying geochemistry evolution (Bui et al., 2014a).

In this study, we ask several questions pertaining to speciation dynamics of the Acacia radiation within a continental climatic context. Do all lineages share the same diversification rate dynamics, and, if not, do lineages with different diversification dynamics have the same rate of climate space exploration? Did climate cooling and drying during the mid-Miocene trigger a lineage-wide response, or heterogeneous responses among different lineages? Are diversification rates correlated with climate space exploration rates? To address these questions, we first analyse molecular sequence data for 503 Acacia species to estimate a time-calibrated phylogeny. We then carry out ancestral state reconstruction and test climate space exploration using previously calculated central climatic tendencies of these species based on BioClim data layers and 132 000 georeferenced herbarium specimens (González-Orozco et al., 2011). We use species environmental means to represent the average of the realized niche occupied by each species. We use analytically derived subtrees of the Acacia phylogeny as experimental units to reduce any dependency between estimates of diversification and taxonomy (Rabosky et al., 2013). Ultimately, we characterize the historical interplay between climate and the spectacular diversification of this dominant component of Australia's contemporary landscapes.

Materials and Methods

To achieve our aims, we explored patterns in the relationship between phylogeny and average realized climatic niche for each species within an evolutionary context. We employed previously published (González-Orozco et al., 2011) estimates of central tendencies of responses (Saupe et al., 2018) for each environmental variable from georeferenced herbarium vouchers. Similar approaches have been used in a number of comparative phylogenetic analyses (Ackerly et al., 2006; Rabosky & Adams, 2012; Kozak & Wiens, 2010, 2016; Cooper et al., 2011).

Results

Diversification models

Log lineage through time plots returned a nearly straight line, with a slight dip below a linear relationship close to the present (Fig. 1). Bayes factor comparison returned decisive support for the constant-rates birth–death process model (Table S5). Diversification and extinction rates were similar across models, with inferred extinction rates around one-tenth the diversification rate. Posterior predictive simulation suggested the constant-rates birth–death model could simulate trees indistinguishable from the Acacia phylogeny (Fig. 1). The CoMET analysis returned low posterior probability for a decrease in speciation rate close to the present, despite a decline in average lineage-wide speciation rate between 1 and 2 Ma (Fig. S4). No evidence for mass extinction events was returned, consistent with the TESS stepping-stone analysis.

Bamm

Effective sample sizes for replicate runs with unbalanced sampling of bipinnate species were higher than with balanced sampling, but were > 200 for both log likelihood and number of rate shifts (Table S6). Balanced and unbalanced sampling returned similar posterior probabilities for the number of rate shifts. A diversification rate regime comprising a single lineage-wide rate and zero shifts was never sampled in any run. The posterior probability comprised rate regimes with between one and seven rate shifts, with two rate shifts having the highest posterior probability (Table S6). The probability of two or three rate shifts was slightly higher in analyses of the unbalanced tree, and the probability of one rate shift slightly higher in analyses of the balanced tree.

Bayes factor comparison of models with more than one shift and the single shift model returned decisive support in favour of models with two or three shifts for the unbalanced tree. By contrast, though support for the two or three-shift models was higher for the balanced tree, Bayes factor support was suggestive but not decisive in comparison with models containing only a single shift (Table S6). The best shift configuration for both balanced and unbalanced data sets contained two shifts (Fig. 2a).

All replicate runs returned PPDs containing four or five distinct shift configurations, most of which contained two core shifts. In all of the credible shift sets from the balanced and unbalanced analyses, a shift on the branch leading to ES 1 was present, which formed more than 95% of the samples (Fig. 2b). In the balanced tree, the same shift was present in four of the five credible configurations, which formed 92–95% of the samples. We use the term ES rather than clade or lineage, because both ES 0 and ES 1 were paraphyletic. The second diversification rate shift varied in position among the configurations but was always embedded within ES 1 and in proximity to a lineage containing most of the bipinnate species and their close relatives (Fig. 2a,b). The most credible shift configurations illustrate the concordance among runs in the localization of the first of these two inferred shifts (Fig. 2b), as does the marginal probabilities associated with shift events (Fig. 2c), and three macroevolutionary cohorts were indicated the cohort matrix (Fig. 2d).

Lineage-wide diversification rates for all Acacia declined slightly before remaining relatively constant until around 4 Ma, when they commenced a shallow decline again (Fig. 3a). The first ES (not including ES 2) initially had a diversification rate twice that of the other lineages, but the diversification rate declined through time (Fig. 3b). A similarly high initial speciation rate and decline was inferred for ES 2. The diversification rate across the basal grade declined steadily through time. Lineage-wide extinction rates were inferred to have remained lower than diversification rates throughout Acacia history (not shown).

Rates of environmental space exploration

ESs differed in their rates of climate space exploration (Fig. 4a). ES 2 explored climate space 1.6–2.5 times faster than the slowest (ES 0), depending on the data set; this difference was significant (P = 0.001) across all six data manipulations. The rate of climate space exploration for ES 1 was significantly faster than ES 0; and though slightly slower than in ES 2, this difference was significant in four of six data manipulations (Table 1).

Table 1. Results of evolutionary rates analysis, with climate data variously transformed and reduced.

Data set	Rates			Ratio	P	Pairwise comparisons
	ES 0	ES 1	ES 2			0 < 1	0 < 2	1 < 2
z-scaled	0.090	0.127	0.157	1.747	0.001	0.001	0.001	0.001
z-scaled VIF	0.088	0.130	0.150	1.699	0.001	0.001	0.001	0.108
PCoA first 10	0.243	0.352	0.407	1.677	0.001	0.001	0.001	0.023
PCoA all	0.088	0.130	0.150	1.699	0.001	0.001	0.001	0.001
Log	0.004	0.005	0.007	1.871	0.001	0.001	0.001	0.001
Bivariate	0.066	0.138	0.168	2.546	0.001	0.001	0.001	0.143

In the Acacia phylogeny, climate space distances between descendant nodes are on average greater per unit phylogenetic branch length in two nested subtrees than they are in the lineages comprising the basal grade outside the first diversification rate increase. P-values of 0.05 or less were returned for 990 of 1000 PPD trees for the principal components scores, and 970 of 1000 PPD trees for the raw bivariate data, indicating the result obtained from the MCC tree was robust to phylogenetic uncertainty (Fig. 4b,c).

Disparity through time

For the Acacia lineage as a whole, inferred environmental disparity was within bounds of expectation under a Brownian null model for most of its history (Fig. 5a). However, the inferred disparity increases at c. 7.5 Ma, and becomes significantly greater than expectation, remaining so to the present, with a sharp increase in inferred disparity at c. 2 Ma. Mean disparity index values were positive in all replicates of this analysis, ranging between 0.089 and 0.093, with probabilities of between 95% and 96% that these values were significantly more positive than expected under time-dependent Brownian null models. The Acacia lineage, as a whole, exhibits high levels of subclade disparity, indicating Acacia subclades overlap in climate space (Fig. 5a). This result suggests most variation in climate occupancy is partitioned within subclades.

ES 0 has significantly higher disparity values than expected under a Brownian null model from very early in its history (Fig. 5b). Environmental disparity shows a positive spike close to the present, starting at c. 2 Ma. By contrast, ES 1 has disparities compatible with the null model until relatively recently, when it, too, exhibits a sharp positive spike in environmental disparities (Fig. 5c). ES 2 has higher than expected environmental disparity for most of its history but, unlike the other two subtrees, does not exhibit a near-present spike (Fig. 5d).

Ancestral state reconstructions for climate variables

Ancestral state reconstructions for climate variables revealed different historical patterns for different parts of the Acacia phylogeny. ES 0 shows climate reconstructions consistent with historical occupancy of warm and fairly dry environments (Fig. 6). However, consistent with the high disparity values recovered for this lineage is the substantial among-lineage variation in some key climatic variables, including coldest quarter soil moisture, warmest quarter and mean temperatures, and driest period precipitation (Fig. 6). Finally, reconstructed environmental variables on the Acacia stem lineage suggest occupancy of moderately warm, seasonally dry climates with moderate to high seasonal change in soil moisture (Fig. 6). Climate state reconstructions on the ES 1 stem lineage suggest their diversification occurred in relatively dry environments, with many daughter lineages then diversifying along a number of environmental variables, and others remaining in dry habitats. Notable within ES 1 are the multiple origins for lineages in extremely dry coldest quarter soils (Fig. 6b). The second ES contains species primarily occurring in the sclerophyll biome in south-east Australia, all of which occupy soils with higher average annual moisture and higher driest quarter moisture than most of the rest of Acacia.

Discussion

In the Acacia phylogeny, ES 1 and ES 2 had significantly higher diversification rates and significantly higher climate space exploration rates than ES 0. Further, ES 2 may explore climate space faster than ES 1, though this was sensitive to data transformation. Phylogenetic localization of the diversification rate increase defining ES 1 (Fig. 2) means phylogenetic comparative methods can seek potentially causative differences between ES 1 and ES 0. Species of ES 1 are distributed among all of the disparate biomes recognized for Australia (Crisp et al., 2004), and the geographic distribution of the lineage encompasses all of Australia. ES 0, which did not experience a diversification rate shift, has the same broad geographic distribution (Fig. S6), suggesting geography cannot explain the difference. Throughout the past 25 Myr, salinity and alkalinity tolerance have evolved repeatedly in Acacia, and range-restricted species are often tolerant of extreme geochemistry (Bui et al., 2014b). This capacity for physiological adaptation may explain differences in diversification rate if salinity and alkalinity tolerance evolved more frequently in ES 1 and ES 2 than in ES 0. Tests for phylogenetic localization of salinity and alkalinity tolerance would be worthwhile. Within ES 1, rhizosphere adaptations may be a key trait to cope with the annual extremes of cyclically waterlogged and parched tropical monsoon soils (Bui et al., 2017). Soil pH tends to decrease with rainfall: tropical monsoon soils tend to be acidic, whereas those to the south are often more alkaline (Ahren et al., 1994; Henderson et al., 2005). The concentration of tropical monsoon species into a few diverse clades (Fig. 6) hints that the requisite adaptations were repeatedly acquired, and resulted in minor bursts of radiation in lineages that acquired them. This provides a hypothesis for physiologists to explore, as the nature of key innovations and the patterns of diversification within the monsoon tropical species of Acacia are poorly known and warrant investigation. Though many Acacia species from the seasonally dry tropics have been highlighted for their economic potential, relatively little is known about their ecophysiology or interactions (Maslin & McDonald, 1996), though one of the parents of the commercially valuable hybrid Acacia × mangiiformis (Maslin et al., 2019) is adapted to acidic soils and resistant to high aluminium ion concentrations (Endo et al., 2011).

Interestingly, four of six data sets returned significantly higher rates of climate space exploration for ES 2 than ES 1, suggesting that the second diversification rate increase might also be correlated with an increase in climate exploration rate. ES 2 occurs in the mesic south-east of Australia (Fig. S6), in contrast to ES 1. This suggests an ability to withstand the stresses of arid and semi-arid environments, and the ability to transition into and exploit mesic environments have both contributed to the current dominance of Acacia across the Australian continent. The south-east mesic radiation of ES 2 includes lineages that have reverted to bipinnate adult foliage and occur in climate spaces with low soil moisture seasonality and higher average soil moisture levels, as expected from studies of water use efficiencies in leaves vs phyllodes (Brodribb & Hill, 1993).

Taxonomic work is unevenly distributed across Australia and Acacia. Although our sample was made without knowledge of phylogeny it is not a random subset of Acacia; it likely overrepresents species from south-eastern Australia. It contains 49 of the 70 (70%) bipinnate species and 82% of south-eastern bipinnate species, a bias we corrected for when estimating lineage diversification rates. What impact on temporal patterns of variation might overrepresentation of south-east Australian Acacia have? Perhaps if related species from the same or similar environments were overrepresented we should expect shorter branch lengths separating species in climate space and lower rates of climate space exploration.

However, other biases in the species sampling may remain, because less than half of the phyllodinous species were represented. Taxonomic work in Australia is not tightly linked to the amount of phylogenetic diversity in individual clades (Miller et al., 2018), with some angiosperm clades having much less phylogenetic diversity than expected given their taxonomic richness. This is relevant to the south-eastern radiation of ES 2 because a south-east Australian sampling bias is a potential limitation of our study. Having acknowledged that bias, we note that south-eastern Acacia species are dispersed throughout the phylogeny, and intuitively we expect adaptive radiations to be geographically biased. The rate shifts, and their environmental correlates, are trends in the data that can be further explored on the basis of more even sampling throughout the Acacia phylogeny.

Node ages within south-east Australian Acacia suggest recent and rapid radiation in the south-east Australian mesic zone after the Pliocene climatic remission. Age estimates are associated with reasonably broad credibility intervals, but the crown node age for ES 2 is 5.77 Ma (7.97–4.05 Ma). This recent and rapid radiation stands in contrast to the diversity reduction and extinction in rainforest and sclerophyll communities documented in the Pliocene–Pleistocene fossil record (Jordan, 1997; Jordan et al., 2007; Sniderman et al., 2013). ES 0 and ES 1 both show a spike in climate disparities close to present (< 2 Ma), corresponding with the onset of Pleistocene glacial–interglacial cycling.

Glacial–interglacial climate cycling might have enhanced the permeability of climatic or biome boundaries as populations and species were repeatedly perturbated through climate cycles. Climate cycling may have enforced rapid climatic adaptation as the only alternative to extinction and may have promoted climate disparification during this time period. Either hypothesis may explain the large number of sister species younger than Pliocene age recovered in the Acacia phylogeny. A key question is how the Acacia lineage invaded and diversified within habitats and communities that may have had higher historical species richness, and which experienced diversity reduction contemporaneous with the Acacia radiation. Perhaps the susceptibility of the mesic zone to prolonged drought is a relatively recent phenomenon, one to which physiologically tolerant lineages, like Acacia, have responded (Brodribb & Hill, 1998). Phyllodes use water more efficiently and are generally more tolerant of drought than leaves (Brodribb & Hill, 1993). However, this cannot be the whole story, because a substantial number of species comprising the Acacia radiation are bipinnate. A predominance of some Acacia lineages in sclerophyll communities today may reflect their capacity to cope with and respond to cyclically fluctuating climate regimes. Radiation and diversification of isolated populations within communities left depauperate by local extinction during glacial cycles may have contributed to extant diversity. Acacia may, too, have been subject to Pleistocene extinctions; Acacia was present in New Zealand until the early Pleistocene (Mildenhall, 1980).

Ecological saturation models, which predict that younger lineages should have ecological opportunities limited by older established lineages (Burbrink & Pyron, 2010; Mahler et al., 2010; Jønsson et al., 2012), do not describe the climate–phylogeny relationship well in Acacia, whose younger lineages seem to have had more opportunity for climate space exploration rather than less. If speciation involved shifts in climate space occupancy, this could explain the correlation between diversity and climate disparity in Acacia (Ricklefs, 2006). Since our analysis used groups of lineages with different diversification dynamics, time dependence is less likely to confound the results. Under time dependence, total climate disparity is proportional to lineage age, and older lineages have more disparity than younger lineages (Pagel, 1999; Purvis, 2004; Claramunt, 2010; Múrria et al., 2018). The Acacia phylogeny exhibits the opposite pattern, with a negative correlation between disparity and clade age, and highest disparity values among the youngest sister lineages (Fig. S5). Acacia appears consistent with other radiations, in which disparification in climate space occurs later in the radiation (Rolshausen et al., 2018), and the mode of Acacia's disparification in climate space should be examined further using more sophisticated methods (Puttick, 2018; Rolshausen et al., 2018). Both the node-height test and ancestral climate reconstructions suggest an Ornstein–Uhlenbeck process with many local optima. ES 0, which contains the oldest Acacia lineages, has higher climate disparity than expected under a Brownian null model (Fig. 5b), which indicates positive time dependence may apply in the basal clades of Acacia, even though the absolute rate of climate space exploration is lower on average than in lineages comprising ES 1 and ES 2. This pattern mismatch across the phylogeny and the incongruence between the rates of climate space exploration and disparities in ES 0 suggest different parts of the Acacia phylogeny have been subject to different macroevolutionary processes.

Whatever historical processes contributed to these opportunities, they did not impact Acacia-wide diversification rates, which remained constant for 20 Myr from the origin of Acacia, before declining during the early Pleistocene. The nearly linear lineage through time plots until the Pleistocene are remarkable, suggesting monotonic diversification resulted in extant Acacia diversity. The shape of the Acacia phylogeny and lineage through time plots are compatible with a constant-rates birth–death process and exhibit no evidence for time compression in node distribution. However, this lineage-wide perspective is not influenced by any asymmetry in the distribution of nodes among lineages, and it overlooks among-lineage diversification rate heterogeneity. Diversification rate increases in ES 1 and ES 2 have been balanced by overall declining speciation rates.

There is considerable evidence from various sources for sea-level rise, mean temperature increase, and expansion of rainforests associated with a remission to wetter climates during the early Pliocene (McGowran et al., 2004). Speleothem pollen records from the Nullarbor region reveal a turnover of vegetation assemblage at 5 Ma from pre-existing xeric shrub-woodlands, including the wind-pollinated Gyrostemaceae, Casuarinaceae, and Haloragaceae, to mesic forests of Eucalyptus, Banksia, and Doryanthaceae (Sniderman et al., 2016). These changes in vegetation were associated with increases in mean annual precipitation from 320–875 mm to 635–2100 mm (Sniderman et al., 2016). The pollen record indicates two floristic turnover events: one with the remission to a wetter climate, and the other by the subsequent return to semi-arid conditions. The pre-existing semi-arid flora was extirpated, and when semi-arid conditions returned the region was colonized by a different flora.

Following the Pliocene climatic remission, global climates entered a forcing phase dominated by orbital precession and obliquity extending from the late Pliocene to early Pleistocene (Sniderman et al., 2007). Sediments from the Stony Creek Basin in Victoria preserve high-amplitude cyclical fluctuations of rainforest pollen types, including Dacrydium, Dacrycarpus, Dilwynites, and Lophosoria. This turnover occurred over a 20 000–25 000 yr period that resembles the Earth's orbital precession in both duration and fluctuating amplitude of cycles, suggesting the rapid changes in abundance of rainforest angiosperms and podocarps were a response to insolation forcing (Sniderman et al., 2007). Floristic turnover at the end of the Pliocene remission is well documented in south-east Australia, where the widespread tropical araucarian rainforest retreated and was replaced by a modern sclerophyll mosaic with pockets of cool-temperate rainforest (McGowran et al., 2004). Whether this modern sclerophyll mosaic was the same as that which presumably occurred before the expansion of rainforests is not known.

Ancestral state reconstructions suggest the Acacia ancestral climate was warm on average, but not extremely hot, and relatively dry but not highly seasonal, perhaps equivalent to climates characteristic of modern semi-arid environments. This is partly consistent with previous studies of Acacia that identified multiple origins for arid-inhabiting species (Murphy et al., 2003; Ariati et al., 2006), although our ancestral state reconstructions suggest that arid inhabitance is possibly symplesiomorphic within Acacia and the multiple lineages in mesic habitats are derived. An endogenous radiation of Acacia into newly formed xeric environments on the recently isolated Australian continent is consistent with our ancestral state reconstructions, and perhaps with the timing of divergence from its mesic-inhabiting Australian–Malesian sister lineage. There are rainforest-inhabiting Acacia species, including Acacia elata, Acacia hylonoma, and Acacia melanoxylon, but these are nested within the Acacia crown lineage. It is important to note that our state reconstruction occurs in the absence of an outgroup, so the reconstructions at the crown node may not reflect environmental variables associated with the most recent common ancestor. The sister lineage to Acacia is Paraserianthes (Miller et al., 2013), a monotypic genus with two subspecies, one of which (Paraserianthes lophantha ssp. lophantha) is the only Australian taxon, and is found in mesic habitats in south-west Western Australia. The other subspecies, P. lophantha ssp. montana, is found in Malesia in Sumatra and Java, and the Lesser Sunda Islands of Bali and Flores (Brown et al., 2011). The other outgroup used in the current study is Pararchidendron, whose only species, Pararchidendron pruinosum, is also a rainforest inhabitant. The habitats occupied by this sister group may, within a broader phylogenetic context, yield ancestral states at the stem node consistent with mesic biomes, but an ancestral transition into xeric environments somewhere along the Acacia stem lineage should still be anticipated.

By definition, the ESs identified in this study are historical entities; we have sought to understand extant Acacia diversity, as distributed among those subtrees, within a historical context. The ESs are characterized by, and have been identified using, their history. In Acacia, increases in diversification rate are coupled with increased rates of climate space exploration once, and possibly twice. The mode of climate space exploration associated with ESs defined by these historical changes in diversification dynamics would be worth pursuing; our results suggest an Ornstein–Uhlenbeck process with attainment of many local optima. Whether any traits underlie differences in diversification dynamics or climate space occupancy within Acacia is not known. Changing climate profiles may have interacted with diversification processes at a continental scale to promote radiation across the lineage. Climate fluctuations during glacial–interglacial cycles may have contributed to the many young species pairs within Acacia, sometimes with rather disparate climate occupancy.

Author contributions

MR: study design, data analysis, results interpretation manuscript writing. CF: data analysis, results interpretation, manuscript writing. JM: study design, manuscript writing. DM: study design, results interpretation, manuscript writing.