Predominant colonization of Malesian mountains by Australian tree lineages

Massive biota mixing due to plate‐tectonic movement has shaped the biogeography of Malesia and during the colonization process, Asian plant lineages have presumably been more successful than their Australian counterparts. We aim to gain a deeper understanding of this colonization asymmetry and its underlying mechanisms by analysing how species richness and abundance of Asian versus Australian tree lineages in three Malesian subregions change along environmental gradients. We hypothesize that differing environmental histories of Asia and Australia, and their relation to habitats in Malesia, have been important factors driving assembly patterns of the Malesian flora.

Malesian Floristic Interchange point to the importance of phylogenetic biome conservatism in biotic interchanges and resemble those resulting from the Great American Biotic Interchange in the Neotropics.

| INTRODUC TI ON
The distribution of terrestrial life on earth is spatially structured into biogeographical regions with more or less homogeneous biota, separated by biogeographical boundaries (Cox, 2001;Holt et al., 2012;Wallace, 1876). Long-term geographical isolation has shaped the deepest boundaries between these regions (Ficetola, Mazel, & Thuiller, 2017) by limiting dispersal to and from other land masses, thus leading to the independent evolution of biotas (e.g. Madagascar, Australia) for millions of years. Rare but frequent long-distance dispersal events followed by successful establishment can lead to the exchange of organisms even between distant landmasses (Crisp et al., 2009). Events where landmasses approach each other due to sea-level changes and/or collision of moving tectonic plates are much less frequent than long-distance dispersal, but cause massive biotic exchange with profound impacts on biotas (Antonelli et al., 2018;Vermeij, 1991). Examples include the collision of the Indian continent with Asia in the Eocene (Dutta et al., 2011) and the repeated presence of the Bering Land Bridge between East Asia and North America during the Tertiary (Donoghue & Smith, 2004;Graham, 2018). Much more recently, transport of organisms by humans has been accelerating the mixing of biotas, causing massive environmental and economic impacts (van Kleunen et al., 2015). Understanding the patterns of past biotic exchanges and their underlying processes can help predict future impacts due to anthropogenic biotic interchange (Heberling, Jo, Kozhevnikov, Lee, & Fridley, 2017;Vermeij, 1991).
One of the best-studied of such events was the successive establishment of the Central American land bridge (Graham, 2018;Montes et al., 2015), which culminated in the Great American Biotic Interchange (GABI; Marshall, Webb, Sepkoski, & Raup, 1982;Wallace, 1876), the extensive mixing of South American and North American faunas and floras in the late Miocene to Pliocene. Much of our current understanding of tropical biotic exchange resulting from geological processes stems from the intense study of the GABI.
The past isolation of landmasses, availability of different habitats in space and time and taxon-specific traits like dispersal capability have all had an influence on the interchange patterns, but their relative importance is debated (Faurby & Svenning, 2016;Marshall et al., 1982;Simpson, 1980;Webb, 2006;Woodburne, 2010). Furthermore, it remains unclear whether the mechanisms of biota mixing unravelled from the study of the GABI are of universal validity throughout the tropics or largely specific to the Neotropical realm.
Another striking example of geology-driven biota mixing in the tropics is the Indo-Australian or Malay Archipelago. This part of the Palaeotropics encompasses the insular region between Asia and Australia (plus the Malay Peninsula) and is usually referred to as Malesia in botanical research (Raes & van Welzen, 2009;Zollinger, 1857). Similar to the situation in Central America before the closure of the Central American land bridge, the extensions of the Asian continent on the Sunda shelf and of the Australian continent on the Sahul shelf have never had a continuous land bridge connecting them. Today, they are separated by narrow stretches of ocean and the islands referred to as Wallacea (Hall, 2017). Already in the Palaeogene, when the continents were still far apart from each other, sporadic long-distance dispersal events by plants occurred from Asia to Australia and vice versa. However, biotic interchange, termed the Malesian Floristic Interchange (MFI) or Sahul-Sunda floristic exchange, sharply intensified in the Early Neogene when the two land masses approached each other and the Wallacean islands emerged in between (25-20 Ma; Crayn, Costion, & Harrington, 2014;Lohman et al., 2011;Richardson, Costion, & Muellner, 2012;Sniderman & Jordan, 2011). Hence, we can generally assume that the species present today in Malesia belong to lineages that were present on either of the two continents before the Neogene but did not occur on both of them.
Today, the fauna of Malesia shows strong geographical structure and includes numerous taxa endemic to the region's diverse subregions.
This pattern, known since the initial observations by Wallace (1860) and consistent with relatively low dispersal capabilities in many animal taxa, indicates dispersal filters from Asia to Australia and vice versa.
The situation in plants, however, is different: Malesia is a well-defined floristic region, albeit with internal geographical structuring (Culmsee & Leuschner, 2013;Raes & van Welzen, 2009). Among the most diverse floras worldwide (Slik et al., 2015), the Malesian flora is furthermore said to be derived predominantly from Asian ancestors, indicating asymmetric colonization . This assumption is mainly based on the fast-growing number of phylogeographical studies of single plant taxa (e.g. Grudinski, Wanntorp, Pannell, & Muellner-Riehl, 2014;Thomas et al., 2012). These studies have proven highly valuable to understand the evolutionary history and biogeography of numerous Malesian taxa and have allowed insights into more general patterns of colonization (summarized in Crayn et al., 2014).
While the available information suggests that overall, Asian lineages dominate the Malesian flora today (e.g. Richardson et al., 2012), detailed phylogeographical studies are still lacking for many species-rich taxa and the mechanisms of the MFI are generally much less understood than those of its Neotropical counterpart GABI. In particular, the colonization patterns of plants since the onset of the MFI under different environmental conditions have not received sufficient attention (but see Yap et al., 2018). Earlier phytogeographical research has highlighted the abundance of Australian elements in certain Malesian forest habitats, such as in mountains and areas with ultramafic parent material (van Steenis, 1935b & Sabir, 2011). Furthermore, previous studies have predominantly used species richness as a measure for evolutionary success (e.g. Richardson et al., 2012) neglecting other quantitative measures like abundance or biomass due to lack of adequate data (but see Culmsee, Leuschner, Moser, & Pitopang, 2010). These other parameters are needed, however, to comprehensively describe patterns of colonization and dominance. Tropical mountain forests are well suited to study colonization processes in the distant geological past: They possess bioclimatic belts with a steep turnover in tree community composition (Körner, Paulsen, & Spehn, 2011), are typically less affected by anthropogenic disturbance and their tree floras are often more natural than those of remaining lowland forests (Cannon, Summers, Harting, & Kessler, 2007).
Malesia and the MFI can serve as an independent model to test hypotheses resulting from decades of research on the GABI. If long isolation leads to lower competitiveness (Faurby & Svenning, 2016;Simpson, 1980), then Australian lineages should be less successful in the colonization of Malesia than their Asian counterparts, regardless of habitat suitability (in the following we speak of Asia and Australia instead of Sunda shelf and Sahul shelf, as Asian species dispersed into Malesia not only from Sundaland, but also via the Philippines; see van Steenis, 1935a). Alternatively, if colonization depends strongly on the available habitat (Cody, Richardson, Rull, Ellis, & Pennington, 2010;Woodburne, 2010), colonization asymmetry will vary between habitats and relate to the environmental history of the source and sink regions. Trees with ancestors in continental Southeast Asia and Sundaland, where rainforest has been present since at least the mid-Eocene (Hall, 2013;Morley, 2012), may possess adaptations to hotter and moister climates than taxa from clades that originated on the Australian continent. The latter underwent strong aridification in the Neogene after its separation from Antarctica when mesic biomes remained confined to mountain areas along the eastern coast of Australia and in parts of New Guinea (Byrne et al., 2011;Kooyman et al., 2014;Quarles van Ufford & Cloos, 2005;Sniderman & Jordan, 2011). Furthermore, the large extent of nutrient-poor soils on the Australian continent could have favoured plant lineages adapted to these soils (Hill, 2004;van Steenis, 1979).
Dispersal filters such as stretches of ocean between suitable terrestrial habitats limit biotic exchange (Bacon et al., 2015;Graham, 2018;Woodburne, 2010). However, plants have comparatively high dispersal capabilities, so that the filter function of ocean barriers may be less-important to them than to many animal taxa, as exemplified by the relative commonness of long-distance dispersal in plants (Bacon et al., 2015;Donoghue & Smith, 2004;Sanmartín, Ronquist, & Cunningham, 2004). During the MFI, newly emerging islands in Wallacea likely facilitated stepping-stone dispersal, so that the occurrence of plant lineages in Malesia may be only weakly dependent on distance from their host region, that is, the land masses of Asia and Australia, whereas suitable habitat may play a larger role.
In this study, we attempt to close the knowledge gap about colonization asymmetry in the MFI. More specifically, we try to quantify the proportion of tree species and individuals with Asian versus Australian ancestry in Malesian forests by adopting a mixed approach. First, we compiled plot-level data of trees from the three major Malesian subregions and biodiversity hotspots Sundaland, the Philippines and Wallacea, to establish a dataset with >15,800 tree individuals of c. 1,640 species from a wide phylogenetic range (c. 35% of all seed plant families containing trees). The dataset further contains information on species abundances and environmental variables for each plot. Second, we inferred the geographical ancestry (Asian or Australian) for each species by building on the wealth of phylogenetic information that has become available in the last three decades supplemented by relevant fossil records. 3. Habitat, rather than distance to the source region (Asia or Australia), drives the differences in community composition.

| Malesian tree inventory data
We compiled a dataset of 55 tree inventory plots (all trees with diameter at breast height ≥10 cm) in old-growth forests (Figure 1), including 42 plots from published studies and 13 plots from our own work on Sulawesi, Indonesia (Table 1, Appendix 1). We classified the western and central parts of Malesia covered by our dataset into three phytogeographical areas, Sundaland, the Philippines and Wallacea, following the nomenclature of biodiversity hotspots of Myers, Mittermeier, Mittermeier, da Fonseca, and Kent (2000).
We selected only such studies that provided species identification to genus or species level with a high taxonomic standard, including the deposition of voucher specimens in herbaria and a full list of the recorded species per plot with their abundances, that is, the per-plot number of individuals per species.

Wallace's Line
New Guinea F I G U R E 1 Diversity and abundance of Australian tree lineages in Malesian forest plots are higher at higher elevations, on ultramafic soils, and east of Wallace's line. Shown are the locations of 55 old-growth forest plots in 12 Malesian areas in the phytoregions Sundaland, the Philippines and Wallacea (a) and the per-plot proportions of tree species (b-d) and individuals (e-g) with Australian ancestry in relation to elevation, parent material and phytoregion. Regression lines show significant results based on multiple logistic regression models (model 1 for b-d, model 3 for e-g; see Table 2 for details). Symbols represent the 12 studied Malesian areas (a), each with 1-16 forest plots (b-g  , 1948-2019The Plant List, 2019;WCSP, 2019). We pruned the original names to species level and manually corrected spelling errors.
We removed cf.-and aff.-qualifiers but retained the following epithets unless the purportedly similar species did not occur naturally in the respective biogeographical region (in that case we used 'sp.'). We treated morphospecies as different species when they were clearly distinguished as such in the original sources (e.g. as sp.1, sp.2, etc., see Appendix 1). In addition, we assumed that morphospecies in the same genus but from different original studies represented different species. The total number of species in our dataset (see Appendix S2) is thus probably inflated and should be seen as an estimate.
However, since our main results are based on calculated percentages on a plot basis, this possible bias does not affect our analyses.
Assuming that each species or its ancestor was present in only one continent-Asia or Australia-before the intensification of the MFI in the early Neogene (see Crayn et al., 2014), we separated the species into two groups: those with Asian ancestry (hereafter: Asian species) versus those with Australian origin (Australian species). To compile this information, we carried out an exhaustive literature search for phylogeographical studies (e.g. phylogenetic studies including direct reconstruction of ancestral areas), other dated and undated phylogenetic studies, relevant fossil data and taxonomic literature, using in total c. 300 published sources (see references in Appendix S1 in the Supplementary Information). As most phylogenetic studies use incomplete taxon sampling, many of our species were not directly represented in the references. We therefore assumed that all species of a genus shared the same biogeographical ancestry (Asian or Australian) unless the results of studies indicated otherwise. In the latter cases (e.g. Macaranga, Ficus), we attempted to match our species to infrageneric clades, mostly based on systematic studies and morphological characters. Likewise, we attempted to assign species of polyphyletic genera to the correct clades in the phylogenetic studies. When phylogeographical studies were not available, we resorted to other phylogenetic sources. These often did not explicitly contain information on biogeographical history, but with the distribution of sampled species taken from floristic accounts and databases, we could usually infer the geographical ancestry nonetheless, especially when the phylogenies were dated. Tree individuals which could not be assigned to one of these ancestry groups were excluded from further analyses (n = 245). The majority of those ambiguous individuals belonged to morphospecies which did not have sufficient taxonomic resolution to infer their geographical origin, but we also included alien species here (11 individuals in four species), as they do not convey any meaningful biogeographical information in the context or our study. The percentages of individuals with ambiguous ancestry per plot ranged from 0% to 11.8% (median 0%, mean 1.3%). Plots on ultramafic parent material contained a significantly higher proportion of individuals with unknown geographical origin than plots on other substrates (logistic regression model with quasibinomial error structure: D 2 = 0.16, p < .0001), possibly highlighting the high number of insufficiently known endemics expected to occur there; the other variables did not have any significant effects (Appendix S1: Figure S1.1). Most of the trees that lack information on their geographical origin do so because of insufficient taxonomic resolution (i.e. completely unidentified or identified to family level only). We consider it unlikely that species with Australian ancestry are generally more difficult to identify than those from Asia or vice versa. Therefore, despite the unequal representation of trees with ambiguous geographical origin per plot, deleting these records prior to the main analyses is unlikely to add a significant bias to our results.
The final dataset contained 15,886 individuals assigned to 1,636 species and morphospecies. We were able to classify the majority of individuals (73%) using phylogeographical studies or a combination of these with information on fossils. For most of the rest, we found dated (17%) or undated phylogenies (9%) without direct inference of geographical origin and combined these with information on the fossil record and/or distribution data to infer the respective ancestral regions from other sources. Only for the remaining 2% of individuals, we used fossil record and/or distribution data alone.
Details regarding the methods, references and specifications of species' assignation to their geographical ancestry are summarized in Appendix S1. A list of all species with their respective inferred origin is given in Appendix S2. Finally, we calculated the proportion of Australian species per plot and the proportion of individuals belonging to these species (Australian individuals) per plot. The Australian and Asian proportions per plot amount to 100%.

| Data analysis
To  (Table 1). We followed two strategies to account for the geographical position: first, we calculated the distance of each plot to the nearest border of the Sahul Shelf (displayed in Figure 1a) using the ruler tool in QGIS (QGIS Development Team, 2018). Second, we used the three phytogeographical regions Sundaland, the Philippines and Wallacea to define the plots' regional affiliations.
We then tested for correlation between the environmental parameters elevation, parent material (geology), distance from the Sahul Shelf (Sahul distance) and phytogeographical region (phytoregion).

| Environmental conditions driving biogeographical patterns
Multiple logistic regression models (LRM) uncovered significant differences in the proportion of both Australian species and Australian individuals between the 55 plots from 12 Malesian areas in relation to all investigated environmental variables (Table 2). For the proportion of Australian species per plot, the model with elevation, parent material and phytogeographical region as independent variables (model 1, D 2 adj = 0.71***) performed better than the one using elevation, parent material and distance from the Sahul shelf (model 2, D 2 adj = 0.66***). In both models, plots at higher elevation had higher percentages of Australian species and this factor accounted for about half of the deviance. Australian species were also better represented on plots with ultramafic parent material compared to non-ultramafic localities but the explanatory power of the parent material was much lower (D 2 = 0.12***) compared to elevation. In the phytoregion model (model 1), phytoregion was the second most explanatory variable (D 2 adj = 0.16). Here, Wallacean plots had significantly more Australian species than those from the Philippines and Sundaland. In the distance model (model 2), the explanatory power of distance to the Sahul shelf (D 2 adj = 0.08**) was lower than that of parent material (Table 2, Figure 1b-d).
For the proportion of Australian tree individuals in a plot, differences between phytoregion model (model 3, D 2 adj = 0.69***) and the distance model (model 4, D 2 adj = 0.67***) were negligible. Again, elevation explained more than half of the deviance (D 2 adj = 0.54***) with an increasing proportion of trees from Australian lineages towards higher elevations. The patterns of higher proportions of Australian trees in ultramafic than in non-ultramafic parent materials remained constant (models 3 and 4, D 2 adj = 0.12***). Phytoregion (model 3) and the distance to the Sahul shelf (model 4) both explained a similar proportion of the respective model deviance, and in model 3, there was again a significant difference between the Wallacean plots with more and those from Sundaland and the Philippines with fewer Australian tree individuals (Table 2, Figure 1e-g).
The patterns retrieved for the elevational distribution of Australian species and tree individuals were remarkably similar but the increase with elevation was stronger for individuals than species. Likewise, the increase of Australian individuals from non-ultramafic to ultramafic plots was larger than that of Australian species (Table 2, Figure 1b-g).
On non-ultramafic parent materials, our models predicted c.
35%-55% Australian species at an elevation of 2,500 m in the three F I G U R E 2 Elevational tipping points with a 50%-share of Australian tree species showing the transition from Asian-dominated forests below the coloured lines to Australian-dominated forests above them. Data show the 50%-tipping points for species (a) and individuals (b) in forests of the Malesian phytoregions Sundaland (Sunda.), the Philippines (Phil.) and Wallacea (Wallac.) on non-ultramafic (solid lines) and ultramafic (dashed lines) parent materials. Values from multiple logistic regression models with elevation, parent material and phytoregion as independent variables (model 1 for a, model 3 for b; see Table 2

| Elevational tipping points
On non-ultramafic substrates, the elevational tipping point with a 50%-share of Australian species was reached at 2,270 m in Wallacea
Due to lack of available data, we did not include plots from New Guinea in our study. However, previous studies have suggested that even in New Guinea and tropical northern Australia, Asian lineages contribute more to regional floras than their Australian counterparts Yap et al., 2018).

| Plate tectonics, climate and the Malesian Floristic Interchange
We used a dataset of tree inventory plots in three Malesian biodi- Sunda Islands) including some of the largest areas of ultramafic bedrock worldwide (Galey, van der Ent, Iqbal, & Rajakaruna, 2017;Hall, 2013Hall, , 2017. At the same time, temperatures rose again and moist tropical habitats expanded in Australia until the mid-Miocene climatic optimum. Global cooling since the mid-Miocene (c. 14 Ma) was offset in Australia by the continued rafting towards the equator, at least in the north but intensive aridification occurred, leading to an overall contraction of mesic biomes, which remained mostly in montane refugial areas along the eastern coast of the continent and in eastern New Guinea, where mountains had been present since the early Oligocene (Bryant & Krosch, 2016;Byrne et al., 2011;Macphail, 2007;Martin, 2006;Quarles van Ufford & Cloos, 2005).
The emergence of Wallacea greatly reduced the distances of open sea between Asia and Australia and coincided with the dramatic intensification of the MFI (Crayn et al., 2014) likely due to the facilitation of stepping-stone dispersal between the continents van Welzen, Parnell, & Slik, 2011 (Sniderman & Jordan, 2011;Yap et al., 2018). Because the Australian flora had suffered widespread extinction of typical tropical rain forest taxa before the intensification of the MFI, source populations for the colonization of Malesian lowlands were probably small, when rain forest habitats expanded in the early Miocene (Byrne et al., 2011;Richardson et al., 2012;Sniderman & Jordan, 2011).
Nevertheless, our results show that some Australian lineages were highly successful in colonizing Malesia from the east, especially in azonal habitats with poor soils including ultramafics and on mountains. Taxa adapted to poor soils and/or colder upland habitats were widely present in Australia at the onset of the MFI (Bryant & Krosch, 2016;Hill, 2004;van Steenis, 1979) and were presumably able to colonize the newly emerging extensive ultramafic areas in New Guinea and Wallacea in due time (Galey et al., 2017). Malesian mountain building has mainly occurred in the last 10 My but the highest ranges like the New Guinea Highlands, Central Sulawesi Mountains, Barisan Mountains in Sumatra, or Mt. Kinabalu in Borneo were only formed in the Plio-Pleistocene (Baldwin, Fitzgerald, & Webb, 2012;Hall, 2013;Merckx et al., 2015;Nugraha & Hall, 2017). As montane habitats in Malesia were formed and aridification intensified in Australia, species spread from refugia along the eastern Australian coast and in already existing mountain ranges of New Guinea to the large emerging mountains in New Guinea, Wallacea and Sundaland (Hill, 2004;Kooyman et al., 2014;Morley, 1998;Sniderman & Jordan, 2011 (Biffin et al., 2010).

| Dominant Australian taxa differ in their history and ecology
Today it is the most species-rich genus of Malesian trees, occurring in a variety of moist forest ecosystems, especially dominant on nutrient-poor soils and in montane areas .
Elaeocarpus shows a similar spatial and temporal pattern of diver-  Morley, 1998).
Notably, conifers with Asian affinities (Pinaceae) were common in mountains of Borneo until the Miocene, but Podocarpaceae apparently replaced them afterwards (Muller, 1966 Figure 3).

| Habitat-suitability is more important than distance
We found evidence that Asian species predominantly colonized lowland habitats and medium to rich soils during the MFI, whereas Australian species were more successful in the colonization of montane areas and poorer (ultramafic) soils. This pattern across Malesia, one of the major tropical regions of the world, is remarkable given the region's large extension and archipelagic nature.
Geographical distance from the source regions (Asia and Australia) only had a minor influence on the colonization success compared to habitat. This influence was best explained using Malesian subregions as categorical variables, indicating a nonlinear relationship (  (Bryant & Krosch, 2016;Byrne et al., 2011;Martin, 2006), providing the nearest source population of plants adapted to Malesian mountain habitats.
We therefore interpret the colonization pattern found here in the light of similar ecological conditions between source and sink areas, Sundaland as a source area for lowland Malesia on the one hand and Australia as a source area for montane Malesia on the other. This likely points to the importance of phylogenetic biome conservatism (Crisp et al., 2009;Crisp & Cook, 2012), that is the tendency of lineages to retain their ancestral ecology over long time spans and continental scales, in the assembly of the Malesian vegetation (Grudinski et al., 2014;Kooyman et al., 2014). Notably, Asian immigrants have also been relatively successful in the colonization of tropical lowland habitats in northern Australia but not in temperate habitats further south due to environmental filtering (Yap et al., 2018). However, speciation events including biome shifts between montane and lowland forest and vice versa must have occurred in numerous clades, as indicated by speciesrich genera spanning wide environmental gradients today like Syzygium, Lithocarpus (Fagaceae), Litsea (Lauraceae), Elaeocarpus and Symplocos (Symplocaceae). Biome shifts between tropical lowland and montane forests have received relatively little attention, possibly due to the blurred boundaries and close spatial interconnectedness between the two (Antonelli et al., 2018;Donoghue & Edwards, 2014). Nevertheless, tropical mountain areas with their close proximity of widely differing habitats are known to be cradles of diversity with strong species turnover along the elevational gradient, facilitating speciation and associated niche evolution (Merckx et al., 2015;Sanín et al., 2016). The dynamic history of Malesia since the onset of the MFI together with two ecologically and geographically different source regions may thus have contributed to the exceptionally high plant diversity in Malesia today (Slik et al., 2015). Malesian montane forests, while less speciesrich than their lowland counterparts, today harbour many survivors of dramatic extinction events in Australia during the Tertiary (Kooyman et al., 2014) as well as elements of originally tropical Asian families. Their unique evolutionary history and associated higher phylogenetic diversity (Culmsee & Leuschner, 2013;Slik et al., 2009) attest to their exceptional conservation value.

| Patterns resembling the Great American Biotic Interchange
The patterns uncovered in this study allow the assessment of several hypotheses regarding biota mixing that have been postulated based on studies of the GABI, the mixing of North American and South American biotas during the Neogene and Quaternary. Malesia presents an independent model system, which has similarities to the Neotropics but also shows some differences. In both regions, tectonic movement and climatic changes have led to the mixing of biotas between Gondwanan fragments that had long been isolated before (Australia and South America) and Laurasian regions, which had repeatedly been connected to each other during the Tertiary (Eurasia and North America; Donoghue & Smith, 2004;Lawver et al., 2013).
No land bridge connects Asia and Australia, which is similar to the situation before the closure of the Central American land bridge in the Americas, but the absence of a land connection is less important to plants, which have relatively high dispersal capabilities compared to many animal groups (Bacon et al., 2015;Sanmartín et al., 2004). More importantly, before the GABI began, the largest source population of lowland rainforest plants was located in the Amazon basin, that is on the formerly isolated continent of South America. In contrast, Australia, had not only undergone a 10 My-long isolation before the onset of the MFI, but also large-scale extinction of rainforest plants due to continent-wide cooling and aridification (Byrne et al., 2011), whereas large tropical rainforests persisted in Sundaland.
The dominance of Australian plants in Malesian mountain habitats today adds to the growing evidence against the isolation hypothesis (Antonelli et al., 2018;Bacon et al., 2015;Cody et al., 2010), which states that biotas are less successful in events of biotic interchange after long isolation due to lower competitiveness or higher susceptibility to predators (Faurby & Svenning, 2016;Simpson, 1980 (Antonelli et al., 2018;Bacon et al., 2015;Cody et al., 2010;Graham, 2018;Woodburne, 2010). While our results have to be viewed with caution due to the persisting lack of data from a key area, New Guinea, the congruence of scenarios from the MFI and the GABI shown here provide support to the idea that patterns and mechanisms that have been found through decades of studying the GABI are not specific to the Neotropics but have more universal validity.

ACK N OWLED G EM ENTS
We thank the authors of the original tree-inventory studies (Appendix 1) for making their data publicly available. Furthermore, we thank Darren Crayn and Gunnar Keppel, who provided valuable input that greatly helped to define the focus of this study, Ingrid M.
Angel-Benavides, who helped with the statistical analyses, and two anonymous reviewers for their very helpful and constructive re-

DATA AVA I L A B I L I T Y S TAT E M E N T
The tree inventory data are available from the original studies (see Table 1, Appendix 1).

R E FE R E N C E S B I OS K E TCH
The authors, based at Goettingen University, Germany and Tadulako University, Indonesia, are interested in patterns of plant diversity and composition in relation to ecological and evolutionary processes, especially in tropical montane habitats.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section.