Orchid diversity: Spatial and climatic patterns from herbarium records

Abstract Aim We test for spatial and climatic patterns of diversification in the Orchidaceae, an angiosperm family characterized by high levels of species diversity and rarity. Globally, does orchid diversity correlate with land area? In Australia, does diversity correlate with herbarium collecting effort, range size, or climate niche breadth? Where are Australia's orchids distributed spatially, in protected areas, and in climate space? Location Global, then Australia. Methods We compared orchid diversity with land area for continents and recognized orchid diversity hotspots. Then, we used cleaned herbarium records to compare collecting effort (for Australian Orchidaceae vs. all other plant families, and also among orchid genera). Spatial and climate distributions were mapped to determine orchids’ coverage in the protected area network, range sizes, and niche breadths. Results Globally, orchid diversity does not correlate with land area (depauperate regions are the subantarctic: 10 species, and northern North America: 394 species). Australian herbarium records and collecting effort generally reflect orchid species diversity (1,583 spp.), range sizes, and niche breadths. Orchids are restricted to 13% of Australia's landmass with 211 species absent from any protected areas. Species richness is the greatest in three biomes with high general biodiversity: Temperate (especially southwest and southeast Australia), Tropical, and Subtropical (coastal northern Queensland). Absence from the Desert is consistent with our realized climate niche—orchids avoid high temperature/low rainfall environments. Orchids have narrower range sizes than nonorchid species. Highly diverse orchid genera have narrower rainfall breadths than less diverse genera. Main conclusions Herbarium data are adequate for testing hypotheses about Australian orchids. Distribution is likely driven by environmental factors. In contrast, diversification did not correlate with increases in range size, rainfall, or temperature breadths, suggesting speciation does not occur via invasion and local adaptation to new habitats. Instead, diversification may rely on access to extensive obligate symbioses with mycorrhizae and/or pollinators.


| INTRODUC TI ON
Orchidaceae is the most diverse flowering plant family with approximately 25,000 species worldwide, comprising ~10% of all angiosperms (Cribb, Kell, Dixon, & Barrett, 2003). Orchids have colonized all vegetated continents and several subantarctic islands, have one of the broadest latitudinal ranges of all plant families, and occupy a wide range of habitats in epiphytic, terrestrial, and even subterranean forms (Cribb et al., 2003;Givnish et al., 2016;Skotnicki et al., 2009). Despite this diversity, many orchids are naturally rare, exacerbated by human activities especially habitat destruction and degradation (including via climate change), and illegal collecting (Cribb et al., 2003;Reiter et al., 2016;Wraith & Pickering, 2017).
Some of these threatening processes could potentially be mitigated if orchid species are present in protected areas or can be translocated (Swarts & Dixon, 2009b;Reiter et al., 2016, but see Wraith & Pickering, 2017).
Orchid distributions and rarity are also likely to be affected by their many intimate, and often obligate, multispecies relationships (Pemberton, 2010). For example, orchids rely on mycorrhizal fungi for nutrients, especially as seedlings (Rasmussen & Rasmussen, 2009). The degree of specialization in these orchid-mycorrhizae relationships varies, with implications for orchid distributions and rarity (Davis, Phillips, Wright, Linde, & Dixon, 2015;Jacquemyn, Brys, Waud, Busschaert, & Lievens, 2015). In turn, the distributions of mycorrhizal fungi depend largely upon edaphic conditions such as soil moisture, pH, nutrients, and organic content (McCormick & Jacquemyn, 2014;Nurfadilah, Swarts, Dixon, Lambers, & Merritt, 2013). Orchid-pollinator relationships are often unusual and sometimes highly specialized (Gaskett, 2011;Xu, Schlüter, & Schiestl, 2012), and the distribution and availability of these specialized pollinators are critical for successful orchid conservation efforts (Reiter et al., 2017(Reiter et al., , 2016. The insects that pollinate orchids have additional symbioses with other taxa, for example, food plants and larval hosts (Brown & Phillips, 2014;Kelly, Toft, & Gaskett, 2013;Reiter, Lawrie, & Linde, 2018). These extensive, interconnected relationships could limit the distribution of orchids to sites with high general biodiversity (Pemberton, 2010). If orchids are potential bioindicators of general diversity (Newman, Ladd, Batty, & Dixon, 2007), mapping their presence/absence in protected areas gives insight into the distribution and conservation of not just orchids, but a broader network of symbiotic insect, plant, and fungal partners.
Despite strong research interest in orchid biology and conservation, there are few comparative investigations of the distribution, diversity, and niche characteristics of orchids at continental scales.
A single, landscape scale study is available, addressing orchid diversity, habitat, and climate in China (Zhang et al., 2015). Most orchid species (90%) were distributed across only 2.7% of China's landmass, coinciding with regions of high general plant diversity. Nearly, all species were covered by nature reserves (~97% of 1,449 species), and patterns of orchid species richness were best explained by abiotic factors including net primary productivity (24.5%) and moisture index (16.2%).
For example, for orchids in the southwest Australian Floristic Region (SWAFR), presence-absence data from herbarium records were tested against biogeographic provinces, ecological factors including pollination strategy, and climate variables, to explore patterns in species rarity (Phillips, Brown, Dixon, & Hopper, 2007. Taxon-specific distribution analyses have been performed for some Australian orchid species by coupling field or historical orchid abundances with climate and environmental data, for example, for rare species such as Cryptostylis hunteriana (Clark, deLacey, & Chamberlain, 2004) and underground Rhizanthella gardneri (Bougoure, Brundrett, Brown, & Grierson, 2008), and the diverse genus Pterostylis (Janes, Steane, & Vaillancourt, 2010). However, despite the wide accessibility of data on plant distribution, phenology, and climate (Hijmans & van Etten, 2012;Lavoie, 2013;Willis et al., 2017), studies of orchids at continental and global scales are surprisingly lacking. Baseline ecological information about where orchids occur and why is needed for understanding evolution and diversity, and more pragmatic goals such as conservation and climate change adaptation planning.
In this study, we analyze the patterns of occurrence of 1,538 native Australian orchid taxa using a dataset of 174,592 digitized herbarium records and long-term climate averages. Herbarium records are effective for exploring changes in orchid distributions (Kull & Hutchings, 2006) and pollination rates (Pauw & Hawkins, 2011;Robbirt, Davy, Hutchings, & Roberts, 2011), and evidence of phenological cues that track climate and the consequences for this under future climate change (Gallagher, Hughes, & Leishman, 2009).
Orchids originated in Australia 112 Mya before dispersing globally (Givnish et al., 2015), making it the ideal place to investigate longstanding patterns of diversity. The current Australian orchid flora is almost entirely within the tribe Diurideae, which arose in Africa from Neotropical origins ~50 Mya, then reinvaded Australia (Givnish et al., 2015(Givnish et al., , 2016Kores et al., 2001;Weston, Perkins, Indsto, & Clements, 2014). This shared origin means that comparative analyses are unlikely to be compromized by phylogenetic elements.
Although Australia is occasionally considered depauperate in orchid species given its land area (Dafni & Bernhardt, 1990; van der Cingel, | 11237 GASKETT And GALLAGHER 2009b; WCSP, 2016), and may rank alongside well-recognized orchid hotspots such as South and Central America and Southeast Asia (Cribb et al., 2003). Australian environments are highly varied (montane meadows, arid grasslands, tropical rainforests), and digitized vouchered herbarium collections are plentiful and freely available online (CHAH, 2009).
Here, we use herbarium data to explore orchid diversity relative to land area, collecting effort, spatial and climatic distributions, and the protected area network. We test the following hypotheses: (a) Globally, Australia is relatively depauperate in orchid diversity per unit land area, (b) herbarium collecting effort for orchids is similar to other highly diverse Australian plant families, (c) orchid species are adequately represented in protected areas, (d) orchid richness differs between biomes within Australia, (e) orchids from highly diverse genera have larger range sizes and occur across a wider breadth of temperature and rainfall conditions than those from less diverse genera.

| Land area and orchid diversity
Numbers of orchid species per continent and for known orchid diversity hotspots were obtained from the World Checklist of Selected Plant Families (WCSP, 2016), the World Geographical Scheme for Recording Plant Distributions (Brummitt, 2001;Hopper & Gioia, 2004;Zhang et al., 2015). Land areas were from the 2015 United Nations Demographic Yearbook (UN Statistics Division, 2015) and the Island Directory (Dahl, 1991). We performed linear regressions for the number of orchid species (log 10 transformed) and area in km 2 for (a) continents and (b) known orchid diversity hotspots. To test whether Australia is depauperate, we compared the number of orchid species per km 2 between Australia and the other continents via ANOVA. Continental comparisons were performed both with and without Antarctica since Antarctic orchids are restricted to small, ice-free subantarctic islands comprising just 0.35% of Antarctica's area (Dahl, 1991).

| Species occurrence records
We used records from Australia's Virtual Herbarium (AVH; via the Atlas of Living Australia [ALA] Data Download Portal; https://collections.ala.org.au/) to characterize the distributional range of native Australian orchid species. The AVH is the largest source of Australian floral distribution data and is based on digitized records of vouchered specimens from Australia's nine major herbaria (CHAH, 2009). This preliminary dataset was cleaned by removing records which were (a) not identified to species level (i.e., consisting only of a genus name and the epithet "sp."); (b) collected in a country other than Australia; (c) lacking in georeferencing information (latitude and longitude coordinates); (d) duplicates (i.e., nonunique combinations of latitude, longitude, and species name); (e) cultivated; (f) hybrid combinations between species; and (g) not native (i.e., introduced species identified in both (Randall, 2007) and using the tag "naturalized" provided in the Australian Plant Census [APC] https://biodiversity.org.au/nsl/ services/apc). The dataset thus consisted of 174, 592 occurrence records for 1,538 orchid species (Supporting Information Table S1.1 in Appendix S1). Records in the ALA provide the name submitted at the time of collection, plus a corrected name where relevant. Here, we use the corrected name, reflecting the accepted names in current literature.
Use of herbarium records for orchid studies is hindered by taxonomic uncertainty, often due to orchids' capacity to readily hybridize and speciate. Many Australian orchid groups have undergone revisions, sometimes involving controversial splitting of genera (Hopper, 2009;Jones, 2006)

| Do herbarium records and collecting effort reflect species diversity?
We tested whether herbarium records represent Australian plant diversity with linear regressions comparing the number of species versus the number of collecting records for (a) all plant families or (b) all orchid genera (data were log 10 transformed to approximate normality). Collecting effort (log 10 [number of records/number of species]) was compared between the Orchidaceae and the other top 10 most diverse plant families using ANOVA.

| Distribution: Spatial and climatic
We mapped patterns of orchid richness by projecting all occurrence records into a 0.5° × 0.5° equal area grid of the Australian continent and counting the number of orchid taxa (either species or genera) in each grid cell. All mapping was performed in R version 3.0.2 (R Foundation for Statistical Computing, Vienna, Austria) using the "raster" package (Hijmans & van Etten, 2012) and in ArcGIS v. 10.2 (Environmental Systems Research Institute, Redlands, CA, USA). We determined range size (the area of occupancy in km 2 ) by counting the number of 10 km × 10 km (100 km 2 ) equal area grid cells occupied by the species across Australia. We compared the log 10 transformed range sizes of orchid species and other higher plants (angiosperm and gymnosperm species) using ANOVA with post hoc Tukey tests.
We overlaid occurrence records with the 2014 Collaborative Australian Protected Area Database (https://www.environment.gov. au/land/nrs/science/capad) and extracted the number of species in each protected area. We used this to then count how many species were (a) absent from all protected areas, (b) present only in protected areas, or (c) found in both protected and unprotected areas.
We calculated the richness of orchid genera and species in differ- We visualized the climate conditions that orchids occupy across Australia by extracting values of mean annual temperature (°C; MAT) and annual rainfall (mm; AP) for each orchid occurrence from gridded climate datasets and overlaying them onto a two-dimensional bi-plot of the total available climate space for Australia. Total Australian climate space was determined by extracting MAT and AP for all grid cells across the continent at a 5-arc minute resolution.

| Orchid diversification: Spatial and climatic patterns
To explore climatic factors associated with orchid diversity, we characterized the range of temperature and rainfall conditions occupied by each orchid species across its range (i.e., niche breadth). Niche F I G U R E 1 Orchid species richness across Australia. (a) The biomes of Australia (Koppen Climate Classification; Bureau of Meteorology, 2006). Values indicate the number of orchid species found in each biome. (b) The number of orchid species recorded in each 0.5° grid cell across Australia based on records from Australia's Virtual Herbarium (http://avh.chah.org.au/) for 1,538 orchid species. (c) The realized climate niche of orchid species in Australia. Black circles represent location records in a climate space defined by mean annual temperature and annual precipitation. Large gray circles represent all the available climate space for these two variables across Australia. Climate data were extracted at a 5 arc minute resolution using baseline data for the period 1950-2000 available from the Worldclim dataset (http://www. worldclim.org/; Hijmans et al., 2005). (d) The number of orchid genera relative to the number of orchid species in each 0.5° grid cell breadths were determined by extracting data on MAT and AP for each occurrence record using the Worldclim dataset and calculating the range of conditions occupied by each species. Then, we compared the log 10 transformed niche breadths between the five most highly diverse orchid genera, and all the remaining orchid genera, using ANOVA with post-hoc Tukey tests.

| Land area and orchid diversity
There was no significant relationship between orchid diversity and land area for continents (Figure 2; with Antarctica: R 2 = 0.199,

| Do herbarium records and collecting effort reflect species diversity?
Herbarium records indicate that orchid species comprise 6.77% of the Australian flora as captured in cleaned digitized specimens from the Australian Virtual Herbarium (1,540 of 22,731 spp.). The Orchidaceae are the third most speciose family, and the third most collected (Supporting Information Table S2.1 in Appendix S2; Figure 3a). For all families, there was a significant linear relationship between species diversity and the number of herbarium records (R 2 = 0.751, F 1,300 = 900.875, p < 0.0001; Figure 3a). However, collecting effort (number of herbarium records per species) was marginally, but significantly, lower for the Orchidaceae than for the other top 10 most diverse plant families (ANOVA: F 1,10 = 5.243, p = 0.048; Figure 3a).
Collections have been made for 120 Australian orchid genera, and the most highly collected genera are also the most speciose (terrestrial genera Caladenia and Pterostylis; Figure 3b). There was a significant positive linear relationship between the diversity of an orchid genus and number of herbarium records (R 2 = 0.589, F 1,119 = 168.978, p < 0.001; Supporting Information Table S2.2 in Appendix S2).

| Distribution: Spatial and climatic
Australia's orchid species are restricted to 13% of the landmass. Of the 10,812 protected areas in Australia, 2,809 have at least one orchid species present. Of the 1,538 orchid species in our study, 211 (13.7%) are found only outside of the protected area network and 124 (8.1%) are found only within the protected area network (see Supporting Information Table S1.1 in Appendix S1).
Australia's orchid diversity hotspots correspond with three biomes (Figure 1a

| Orchid diversification: Spatial and climatic patterns
Orchid diversification (as measured by the number of species per genus in each grid cell) is the greatest in the regions of highest general orchid richness, which occur in the southeast and southwest corners The top five most speciose orchid genera had different range sizes to the other orchid genera (F 5,1536 = 5.493, p < 0.001), but Tukey post hoc tests reveal this was driven largely by the only genus with an exclusively nectar rewarding pollination strategy, Prasophyllum, which had a smaller range than the other orchids (p values >0.05; Table 1). The top five most speciose genera all had significantly narrower rainfall breadths than the rest of the orchid genera (Table 1; rainfall F 5,1536 = 28.689, p < 0.001, post hoc Tukey tests all p values <0.05). Two of the genera also had significantly narrower temperature breadths than other orchid genera (F 5,1536

| D ISCUSS I ON
Australia is sometimes described as having low orchid diversity given the land area, perhaps due to the large expanses of desert (Dafni & Bernhardt, 1990;van der Cingel, 2001). Our global comparison found no evidence for this, despite orchids only occupying 13% of the Australian landmass (cf. 2.7% of China; Zhang et al., 2015). Instead, we found North America surprisingly depauperate given its large land size, especially when Mexican orchid species are considered separately (Mexico = 1,560 spp., USA, Canada and Alaska = 394 spp.; Krupnick, McCormick, Mirenda, & Whigham, 2013;WCSP, 2016).
Despite their diversity, widespread appeal, and conservation status (Brundrett, 2007;Swarts & Dixon, 2009a), orchids are less collected than the other similarly diverse plant families in Australian herbarium records. Species rarity can hinder collection, but in general, museum collections tend to overrepresent rare species (Garcillán & Ezcurra, 2011;Guralnick & Van Cleve, 2005). As yet, there are no other studies of whether herbarium collections and collecting effort represent the natural diversity and abundance of orchids. It may be that botanists are reluctant to collect orchids because they are aware of orchids' low rates of pollination, fruit set, and recruitment (Brundrett, 2007). Researchers and enthusiasts may also not wish to publicize orchid species and sites as this can lead to overexploitation by enthusiasts or illegal harvesting (Wraith & Pickering, 2017). Possibly, CITES regulations intended to prevent international trade in orchids may also deter some international scholarship and collecting efforts (Roberts & Solow, 2008).
Herbarium records often accurately reflected relative species diversity within orchid genera (e.g., Caladenia and Pterostylis), although some genera were surprisingly well-collected given their lesser diversity, for example, Acianthus (15 spp., 4,584 records) and Microtis (27 spp., 720 records). These genera have some of the broadest areas of occupancy (mean values: Acianthus 12,420 km 2 , Microtis = 13,222 km 2 , all other Australian orchids = 5,256 km 2 ). In contrast, Sarcochilus is neither widespread nor speciose (21 spp., 2,066 records, 5,038 km 2 mean area of occupancy). It may be collected more often because it has a very attractive floral display and is used horticulturally, leading to stronger interest in collecting the wild species. Undercollection appears to be associated with taxonomic issues and/or restricted ranges with high levels of endemism.
Our mapping confirms that Australian orchids are limited to 13% of the landmass, with obvious hotspots of higher diversity. In theory, there should be few barriers to orchid dispersal and colonization as their tiny, dust-like seeds are readily transportable via wind and water, and vegetative reproduction allows new populations to arise from very few individual colonists (Arditti & Ghani, 2000;McCormick & Jacquemyn, 2014). However, experiments and molecular data reveal that orchids rarely achieve long-range dispersal, and most dispersal events are over meters rather than kilometers (Brundrett, 2007;Givnish et al., 2016;Trapnell & Hamrick, 2005). Globally, orchid distribution is patchy with some smaller regions achieving much higher diversity than neighboring areas (e.g., Madagascar ~1,000 species vs. rest of Africa ~2,350 species; Mexico = 1,560 spp. vs. rest of North America = 394 spp.; Krupnick et al., 2013;WCSP, 2016). We identified three key Australian orchid diversity hotspots: the renowned SWAFR of Western Australia, plus the less well-recognized east coast Victoria and New South Wales region, and northwestern and central Tasmania region. The SWAFR is a globally recognized biodiversity hotspot, ranked one of the world's top 25 priorities for conservation (Hopper & Gioia, 2004;Myers, Mittermeier, Mittermeier, Fonseca, & Kent, 2000). The orchids of the SWAFR have received considerable research attention addressing species interactions, rarity, biogeography, and conservation (Phillips et al., 2010;Phillips, Backhouse, Brown, & Hopper, 2009;Phillips, Brown, Dixon, & Hopper, 2007;Swarts & Dixon, 2009b). The orchids of Victoria, New South Wales, and Tasmania are yet to receive such wholistic, region-based research.
We found that 211 orchid species (and presumably, their associated plant, fungal, and insect partners) are absent from Australian protected areas. While orchids in protected areas can be vulnerable to illegal collecting (especially in East Asia; Wraith & Pickering, 2017), protected areas may still offer better opportunities for policing this, and other major threats such as land clearing (Reiter et al., 2016) Australian orchid diversity is the greatest within three of the six available biomes, but orchids are almost entirely absent from desert regions, unlike the other most diverse Australian angiosperm families (Fabaceae and Myrtaceae; Crisp, Cook, & Steane, 2004).
Orchids appear to be particularly vulnerable to water availability; the absence of orchids from drier zones that we report here is Orchid genera n species Range size (km 2 )

Rainfall breadth (mm)
Top five most speciose genera TA B L E 1 Temperature and precipitation breadths for the top five most speciose orchid genera in Australia and all other orchid genera consistent with the few available studies of orchid distribution and climate. For example, the distributions of Pterostylis orchid species and orchid-associated mycorrhizal fungi in Australia are generally associated with water availability, including rainfall, drainage, and moisture index (Janes et al., 2010;McCormick & Jacquemyn, 2014;Nurfadilah et al., 2013). Chinese orchid distributions were also largely determined by net primary productivity and moisture index (Zhang et al., 2015). A study on the distribution of Macedonian orchids did not explicitly test for an effect of water availability, but the restricted distribution of wet meadow and bog specialist species was noted (Tsiftsis et al., 2008).
It is unclear whether it is the orchids or their mycorrhizae that are most dependent on water availability. Physiological con-  (Noy-Meir, 1973). Orchid flowering, seed set, and even pollination are sometimes triggered by rainfall events (Bodley, Beggs, Toft, & Gaskett, 2016;Brown & York, 2017;Fan et al., 2012). Some orchid species exhibit leaf trait adaptations compatible with drought conditions (e.g., sunken stoma, thick surface cuticles in Slipper Orchids, Paphiopedilum; Guan, Zhang, Guan, Li, & Hu, 2011) and so their paucity in Australian deserts may not be due to a lack of adaptive potential for arid environments. There may well be more orchid species yet to be formally collected from drier regions, for example, entirely subterranean and therefore rarely encountered species of Rhizanthella, although this genus is associated with Mediterranean rather than desert climates (Bougoure et al., 2008). Desert biomes are also unlikely to be suitable for maintaining mycorrhizal fungal due to low soil moisture content (McCormick & Jacquemyn, 2014;Nurfadilah et al., 2013).
Our data suggest that orchid speciation is facilitated by geographically linked or spatial factors, rather than any innate traits associated with particular orchid genera. We found that spatial patterns of orchid diversity were consistent at the level of genus and species, that is, regions with more genera also had more species.
Furthermore, for the most diverse genera, species richness was the greatest in geographic regions of high general orchid diversity.
Intriguingly, diversity was not associated with having a wider range size, or a broader tolerance for rainfall and temperature conditions, suggesting genera are not diversifying in response to invading new sites and adapting to their abiotic conditions. Therefore, the most important drivers of orchid speciation are likely to be both spatially linked and biotic. A critical factor is likely to be insect diversity; switching to a new pollinator species is instrumental in orchid speciation (Breitkopf, Onstein, Cafasso, Schlüter, & Cozzolino, 2015;Peakall & Whitehead, 2014;Peter & Johnson, 2014;Sun, Schlüter, Gross, & Schiestl, 2015). Mycorrhizal fungi diversity is less important than pollinators in orchid speciation , but mycorrhizae do nonetheless determine the subset of sites that an orchid can colonize within its possible climatic range (McCormick & Jacquemyn, 2014;Nurfadilah et al., 2013). Dependence on mycorrhizae could underly the narrower range sizes we report here for orchids versus nonorchid species. Regions with diverse and abundant mycorrhizae may well support a richer orchid biota, although this is untested. Worldwide, orchid diversity hotspots do tend to occur in areas of high plant diversity (Cribb et al., 2003;Myers et al., 2000;Zhang et al., 2015), although no formal analyses have been performed.
Although several studies propose that orchid diversification is facilitated by their unique deceptive pollination systems (Givnish et al., 2015), here, we note no relationship between any particular pollination system and current species diversity. The most speciose genera in our study differ in their reproductive strategies: Caladenia has frequent transitions between food and sexual deception and rewarding strategies, Pterostylis are all sexually deceptive,
How abiotic and biotic factors influence the distribution of these pollination strategies on a landscape scale is an intriguing but untested avenue for future consideration (Herberstein, Baldwin, & Gaskett, 2014).

ACK N OWLED G M ENTS
We thank Stuart Allen for assistance in R and Marie Herberstein for useful comments when designing this study. RVG is supported by an Australian Research Council Discovery Early Career Award (DE170100208). We thank Greg Holwell, Stuart Allen, Val Gaskett and Ian and Terry Holwell for providing childcare essential for the completion of this research.

AUTH O R CO NTR I B UTI O N S
ACG and RVG conceived the ideas, collated, and analyzed the data and wrote the manuscript.

DATA ACCE SS I B I LIT Y
All data were extracted from publicly available online repositories.
Details and URLs are provided in the Section .