Local and regional diversity of frog communities along an extensive rainforest elevation gradient in Papua New Guinea

Rainforests on high tropical mountains are globally important species diversity hotspots. We studied amphibians along an extensive rainforest elevation gradient on Mt. Wilhelm (4509 m) in Papua New Guinea. We established eight sites at 500 m elevation increments between 200 and 3700 m a.s.l. and relate their community composition to the known species pool of New Guinea island. We recorded 3390 frogs from 55 species, which is three times more species than at any local community along the elevation gradient. Species diversity peaked at 1700 m a.s.l. for Mt. Wilhelm communities, and at 500–1100 m a.s.l. in the broader New Guinea fauna, probably reflecting increasing speciation and decreasing dispersal rates with increasing elevation. The beta diversity between frog communities was high and increased with increasing elevation. The change in frog community composition across 500 m elevation corresponded to the change over 200 km distance within lowland forests. A majority of frog species were distributed over narrow <500 m elevational ranges, at Mt Wilhelm and the New Guinea fauna more broadly. We did not detect Rapoport's pattern of wider elevation range for species at higher elevations than for lowland species, for Mt. Wilhelm communities or the New Guinea fauna. The high beta diversity patterns along elevation gradients generated by rapid species turnover with narrow elevation ranges make frog communities vulnerable to change in environment, including climate change.

Species diversity trends range from a monotonous decrease with increasing elevation to a mid-elevation peak in diversity (Colwell et al., 2016;McCain, 2010).However, the key drivers of these trends are not well understood, since numerous potentially important environmental factors tend to covary with elevation (Beck et al., 2017).
These factors include temperature and land area, as well as biotic factors such as predation pressure and primary productivity that decline with increasing elevation (Malhi et al., 2017;Roslin et al., 2017).
They may drive a parallel monotonous decline in species diversity in many taxa.A mid-elevation diversity maximum may be generated by geometric constraints, that is, enrichment of mid-elevations by dispersal from both lower and higher elevations, by a combined effect of several opposing drivers, some increasing and others decreasing the diversity with elevation, or by a larger regional species pool generated by faster speciation in mountains (Beck et al., 2017;Colwell et al., 2016;McCain, 2010;Souto et al., 2019;Stevens, 1992;Toussaint et al., 2014).The importance of both local and regional ecological factors argues for the study of both communities and their regional species pools (McCain, 2009).
Community-level elevation trends in the diversity of rainforest amphibians may be driven by the abiotic environment, particularly temperature (Matavelli et al., 2022), or its combination with biotic factors, including food supply (Supriya et al., 2019) and predation pressure (frog-eating snakes are more diverse at lower altitudes in New Guinea: Tallowin et al., 2017).The dominance of abiotic drivers would lead to a monotonous decline in amphibian diversity with elevation, while biotic factors would generate a mid-elevation peak in diversity, either mirroring the trend in food supply, represented by insect abundance (Supriya et al., 2019), or resulting from a combination of high predation pressure at low elevations (by e.g.snakes) and harsh environment at high elevations.Reproductive mode may also play a key role in determining patterns of amphibian diversity with elevation.The number of species with water-dependent reproduction declines with increasing elevation while those with water-independent reproduction exhibit an increase in richness at intermediate elevations (Siqueira et al., 2021).
The regional-level trends in elevation diversity are determined by speciation dynamics, dispersal, and both biotic and abiotic ecological factors (García-Rodríguez et al., 2021;Hubbell, 2001;Tenoriol et al., 2023).A highly efficient dispersal across the entire study area would result in regional diversity patterns matching the local ones, both determined solely by ecological conditions.A universally low dispersal would produce a monotonous decline in regional species diversity with elevation, following the elevational trend in the available land area.The trend of decreasing dispersal with increasing elevation would lead to a mid-elevation peak in diversity, with lowland diversity constrained by low beta diversity due to high dispersal, and high-altitude diversity constrained by limited land area available, together with harsh environment.In this scenario, the regional and community mid-elevation peaks do not necessarily occur at the same elevation.These scenarios may be further complicated by elevation trends in diversification rate, and species density per unit area (García-Rodríguez et al., 2021;Tenoriol et al., 2023).
The change in species composition with elevation is driven by species turnover where new species are introduced to communities, and by nestedness, where species poor communities lose species so that they are subsets of more species rich communities (Baselga & Leprieur, 2015).The beta diversity among communities, defined here as dissimilarity in species composition between two communities, can be constant along the elevation gradient if determined primarily by the mean temperature.Alternately, the Rapoport's pattern of montane species having wider elevation ranges than lowland species (Stevens, 1992) predicts decreasing beta diversity with elevation.A mid-elevation peak in beta diversity can indicate the contact between lowland and montane faunas (Mena & Vázquez-Domínguez, 2005).New Guinea's topographic diversity supports a diverse amphibian fauna, with more than 400 described species reported to date (Frost, 2022) and many additional species yet to be described (Oliver et al., 2022).Many New Guinea frog species are restricted to montane environments, and these often have limited geographic ranges (Köhler & Günther, 2008;Zweifel et al., 2005).The importance of high elevation for regional species pools of frogs has been documented by Tallowin et al. (2017).Here we contrast the regional geographic patterns of anuran amphibians across New Guinea with elevational trends in communities on Papua New Guinea's highest mountain in order to elucidate the key drivers of amphibian diversity on an extensive altitudinal transect.Specifically, we test the following mutually exclusive hypotheses: (1) Amphibian species richness will peak at mid-elevations due to high speciation dynamics generating species with narrow elevation ranges, (2) Amphibian species richness will peak at mid-elevations due to an overlap between lowland and montane species with broad elevation ranges, and (3) If present, a regional species richness pattern determined by the mid-elevation peak will diverge from the species richness peak in local communities, which will be situated in the lowlands with the most favorable ecological conditions including a rich resource base combined with a large area of lowland rainforests.
In addition to exploring patterns of alpha (local) and beta diversity, the study aims to establish baseline information on the elevational distribution of frogs for the future monitoring of threats to frog diversity from climate change (Sheridan & Bickford, 2011) and chytrid fungus (Dahl et al., 2012).

| Study area
The island of New Guinea comprises some of the best developed and preserved rainforest elevation gradients in the tropics.Its mountains have some of the globally highest plant diversities, at >5000 plant species per 100 × 100 km area (Barthlott et al., 2007), and harbor high species diversity of amphibians, birds and mammals (Tallowin et al., 2017).Our study area was situated at Mt. Guinea (Paijmans, 1976).The sites have annual rainfall from 3200 to 4400 mm (local meteorological stations) with a mild dry season (>100 mm rainfall per month) from June to August, stronger only during El Nino events.The condensation zone is around 2600 m a.s.l and the mean annual temperature decreases at the rate of 0.54°C per 100 elevation m, from 27.4°C at the lowland site to 8.37°C at the timberline (Figure S1) while humidity remains at 95%-100% throughout the elevational gradient, decreasing to 80% towards the timberline (one-year data from data loggers; Sam et al., 2019).

| Sampling methods
Each site was surveyed once during the wet season between May 2009 and January 2010 for 14-15 nights (a total of 114 survey nights).A team of 3-4 surveyors searched for frogs in accessible terrestrial and aquatic habitats along a network of existing forest trails at each elevation (Figure S2).We detected frogs visually using a headlamp and by tracking their calls (Dahl et al., 2009(Dahl et al., , 2013)).The calls of vocalizing frog species were recorded, their habitat photographed and snout to vent length measured.Further, we took tissue samples for mitochondrial cytochrome c oxidase I (COI) barcoding to assist species identification (data at www. bolds ystems.org; Project-DBNGF DNA Barcoding New Guinea Frogs).Voucher specimens were taken for taxonomically difficult species and deposited at the Papua New Guinea National Museum (Port Moresby) and the South Australian Museum (Adelaide).Each species was also identified as a terrestrial or aquatic breeder (Table S2).
We extracted information on elevational ranges of the entire known frog fauna of mainland New Guinea (397 species) from a dataset generated by an expert panel assembled for the IUCN 2019 Melanesia-region Amphibian Red List Assessments (IUCN, 2020).
We used these data to generate the New Guinea regional species pools comprising all species present at the eight elevations used for the sampling of frog communities at Mt. Wilhelm (Table S2).We also analyzed the full set of 397 mainland New Guinean species binned to elevation intervals 100 m wide, from 0 to 4000 m a.s.l., or by 500 m intervals where needed for comparison with the community data.
The regional species pool data are based on the total available evidence for New Guinea which could not be standardized to equal sampling effort across elevations.

| Statistical analysis
Species diversity per elevation was standardized by the equal sampling effort at each site.The completeness of the sampling and the species richness per a constant number of individuals were explored by randomized species rarefaction and extrapolation, using 'iNEXT' (Hsieh et al., 2016).The EstimateS (Colwell, 2013) was used for beta diversity (1-Sorensen index) estimates.We also partitioned the overall Sorensen dissimilarity between pairs of communities into the species turnover (Simpson dissimilarity) and nestedness (Sorensen-Simpson dissimilarity, Baselga & Leprieur, 2015).Each species was characterized by mean elevation (average of elevations for all specimens), elevation mid-point (between minimum and maximum elevation) and elevation range (maximum-minimum elevation).Our analyses were implemented in the R language (R Core Team, 2014).We have applied identical analyses to both the community data and the regional species pool data at each elevation.The community data are compared with the regional pools of species whose elevation range includes the elevation of the sampled community.The regional species data were also binned into 100 m elevation intervals, or 500 m intervals where needed for the comparison with community data.

| RE SULTS
A total of 3390 individuals from 55 species, 14 genera and five families was recorded.The family Microhylidae represented most of the diversity, 36 species, followed by Pelodryadidae with 12 species, Ranidae with five species, and Ceratobatrachidae and Limnodynastidae each with one species (Figure S3, Table S1).The Microhylidae and Pelodryadidae dominate also the entire New Guinea fauna, comprising respectively 65% and 29% of species (Table S2).The proportion of Microhylidae increased from lowlands to higher elevations; the frog communities above 2000 m a.s.l comprised only Microhylidae and Pelodryadidae (Figure S3).The proportion of species with terrestrial development (Microhylidae and Ceratobatrachidae), 67% of species in total, increased with elevation from 46% at 200 m a.s,l to 100% at 3700 m a.s.l (Spearman r = .814,p < .05).
The species accumulation curves reached the asymptote at all sites except 2200 m a.s.l and 2700 m where we may have underestimated total species richness (Figure 1).The local diversity varied from two to 18 species among communities, reaching maximum at 1700 m a.s.l.In contrast, the New Guinea species pool at the eight elevations used for the fieldwork yielded the species richness maximum at 700 m a.s.l, as part of a broad diversity peak between 500 and 1100 m a.s.l (Figure 2).
The overlap in species composition decreased rapidly with increasing elevation difference, both in local frog communities and across the New Guinea fauna so that <10% of species were shared between locations 1500 or more elevation m apart (Figure 3a,b).
The Sorensen dissimilarity between communities 500 elevation m apart (i.e., pairs of adjacent sample sites) increased with log(elevation), from 60% of species shared between lowland sites to only 40% of species shared at high elevations (Figure 3c).Most of the beta diversity between communities was due to species turnover, while the nestedness component was low, on average responsible for 13% of the total beta diversity value and exhibited no elevation trend (Figure 3c).The similarity of regional species pools separated by 500 elevation m did not show any statistically significant trend with elevation (Figure 3c).There was also no elevational trend in beta diversity between pairs of adjacent 100 m elevation zones for the New Guinea fauna (Figure 3d).
The high beta diversity is consistent with narrow elevation ranges documented for most frog species in the Mt.Wilhelm communities, where half of all species were found at a single elevation, i. e. across <500 m elevational range.This pattern is also mirrored by the regional species pools (Figure 4a).In the New Guinea fauna, a large number of species is found only within a 100 m elevation zone, while 50% of all species have elevation ranges of <500 m (Figure 4b).
The median elevation range of species in frog communities does not change with elevation at Mt. Wilhelm (Spearman r, p > .05).Similarly, there is no trend in elevation range of frog species at different elevations in the New Guinea fauna (Figure 5).

| DISCUSS ION
Our study shows the importance of long elevation gradients for the maintenance of high local diversity in the tropics.It also explores the importance of geographic scale for ecological patterns, with important similarities as well as differences in species distribution and diversity patterns between communities and the fauna of New Guinea.
The mid-peak maxima in frog communities have also been documented in the Andes and Himalayas (Chettri & Acharya, 2020;Fu et al., 2006;Hu et al., 2011;Navas, 2006).In New Guinea, both local and regional species richness are driven by microhylid frogs, which exhibit evolutionary plasticity in many life history traits facilitating adaptation to various habitats (Köhler & Günther, 2008), including having a terrestrial vs. aquatic mode of reproduction (Günther, 2006;Oliver et al., 2017).The proportion of species with terrestrial development was higher at Mt. Wilhelm than in Neotropical rainforest communities (Haddad & Prado, 2005).Amphibian species with reproductive modes independent of water are particularly common in humid montane habitats (Goin & Goin, 1962)  lowlands.They could include food resources and predation risk.
Insect abundance peaks at approximately 2000 m a.s.l along the Mt.
Wilhelm elevation gradient (Supriya et al., 2019) while the species diversity of snakes and birds as potential predators of amphibians peaks in the lowlands in New Guinea (Sam et al., 2019;Tallowin et al., 2017).The mid elevation peak in species diversity was not caused by the enrichment of mid elevations by lowland and/or montane fauna; that would lead to an increase in mean elevation range per species in mid elevation communities that was not observed.
The regional mid-elevation peak in frog diversity occurs at a lower elevation than the community diversity peak, suggesting greater isolation and therefore speciation of lower-montane populations, compared to the lowlands (Smith et al., 2007).Frog communities in the lowland rainforests of New Guinea are characterized by low beta diversity with a majority of species shared between sites even >200 km apart (Dahl et al., 2009(Dahl et al., , 2013)).This results in a relatively depauperate lowland fauna, with only six of 55 frog species in our Mt Wilhelm data limited to elevations ≤200 m a.s.l, despite the fact that areas below 100 m a.s.l are larger than the combined area of all other elevations in New Guinea.
In contrast to the lowlands, frog communities exhibit high beta diversity along the elevation gradient.The change in species composition, quantified by the Sorensen dissimilarity, between two communities 500 elevation m apart at Mt Wilhelm is comparable to the Sorensen dissimilarity between two lowland communities that are 200 km apart (Dahl et al., 2009).The beta diversity is driven by rapid turnover of species with narrow elevational ranges so that the total frog diversity of 55 species represents 306% of the maximum local diversity of 18 species in a single community.This is the second highest diversity enrichment due to elevation gradient among the seven plant and animal taxa studied at Mt. Wilhelm, ranging from 140% to 330% (Novotny & Toko, 2015).Although beta diversity increases with elevation for frogs F I G U R E 3 Beta diversity of frogs in the Mt.Wilhelm communities and the New Guinea fauna.Sorensen dissimilarity for all pairs of the sampled communities (Local) and their species pools (Regional) (a), and for all pairs of 100 m elevational zones for the New Guinea fauna (b).Elevation trends in beta diversity (Sorensen dissimilarity and Nestedness) across 500 m elevational difference, calculated for all pairs of adjacent communities (Local) and their species pools (Regional) (c), and Sorensen dissimilarity for all pairs of adjacent 100 m elevation zones for the New Guinea fauna (d).The Local Sorensen dissimilarity is correlated with log(elevation) (Pearson r = .758,p < .05),while there is no significant elevational trend for other beta diversity measure.The boxplots in A and B show minimum, maximum, 1st and 3rd quartiles, and median.at Mt. Wilhelm, and more broadly in the New Guinea fauna, it does not exhibit the Rapoport pattern of increasing elevation ranges with increasing elevation (Stevens, 1992).The rapid species turnover and lack of the Rapoport pattern in elevational ranges were also found in Neotropical frog communities (Goyannes-Araújo et al., 2015).These patterns point again to the importance of mountains for the generation of frog diversity in New Guinea and elsewhere.
A mid elevation peak in local diversity, rapid species turnover with elevation, and narrow elevation ranges with no Rapoport pattern are all structural features making the composition of frog communities sensitive to climate change, as any increase in temperature and environmental conditions it drives may lead to large changes in community composition (Laurance et al., 2010).In addition, the environmental changes may include new or expanding pathogens and predators of frogs, in particular the chytrid fungus that affected tropical frog communities in Australia but is so far absent on the Mt.Wilhelm transect (Dahl et al., 2012) and more broadly in New Guinea (Bower et al., 2019).
The high species richness per area, tight species packing, high species turnover and high phylogenetic diversity make large tropical montane ranges globally important for the maintenance of amphibian diversity (García-Rodríguez et al., 2021;Tenoriol et al., 2023).

| CON CLUS IONS
Our data demonstrate significant changes in alpha and beta diversity patterns among rainforest frog communities along a steep tropical mountain gradient, allowing us to test our hypotheses on the determinants of amphibian diversity.In particular, amphibian species richness peaks at mid-elevations both for communities and the regional species pool, therefore likely reflecting high speciation dynamics of species with narrow elevation ranges, rather than an overlap between lowland and montane species with broad elevation ranges.
Wilhelm (4509 m a.s.l.), the highest peak in Papua New Guinea (PNG)We established eight survey sites, at 500 m elevational intervals along an extensive rainforest gradient from 200 m a.s.l (05 o 44' S, 145 o 20' E) in the flood plains of the Ramu River Basin to the forest timberline at 3700 m a.s.l (05 o 47' S, 145 o 03' E; Figure S2).The survey sites represented typical vegetation at each elevation and were selected with the help of indigenous landowners.Our study transect was 30 km long and traversed all elevation forest zones in New and this pattern has been observed elsewhere including Africa (Liedtke et al., 2017) and South America where Siqueira et al. (2021) reported a mid-elevation peak in their richness in the Brazilian Atlantic Forest.Climatic variables such as humidity, precipitation and temperature have generally been considered good predictors of the occurrence of terrestrial reproduction in amphibians (Lion et al., 2019), but Liedtke et al. (2017) demonstrated that steep terrain and absence of aquatic waterbodies, both features of the Mt Wilhelm montane environment, may be more important factors.The mid-elevation peak in community diversity may also reflect other biotic drivers potentially reducing amphibian diversity in the F I G U R E 1 Frog species accumulation curves (95% confidence intervals shaded) for the Mt.Wilhelm communities at each elevation, extrapolated to 700 individuals.F I G U R E 2 Species diversity of frogs observed at eight sites along the Mt.Wilhelm elevation gradient (Local), in the New Guinea fauna recorded from the sampled elevations (Regional) (a) and the entire New Guinea fauna, binned to 100 m elevation bands (b).
Beta diversity was high and increased with elevation, with the entire gradient supporting three times more species than any local community.This contrasts with studies of lowland frog communities across northern Papua New Guinea that reported low beta diversity such that a change in frog community composition across 500 m of elevation on Mt Wilhelm corresponded to equivalent changes across 200 km distance within nearby lowland forests.These patterns highlight the importance of New Guinea's complex geological history resulting in extensive mountain formation for generating and maintaining the island's extraordinary frog diversity.The high proportion of New Guinean frogs restricted to narrow (<500 m) elevational ranges suggests that many frog communities may be vulnerable to climate change.This study has established baseline information on the elevational distribution of frogs on New Guinea's highest mountain that can permit future monitoring of threats to frog diversity from climate change and other threats such as pathogens, including chytrid fungus.AUTH O R CO NTR I B UTI O N S C. D. and N. V conceived the study.C. D. and I. B. conducted fieldwork.C. D., I. B., and S. J. R. identify the specimens.C. D. and S. J. R. conducted taxonomic analysis, and all authors contributed to the analyses, and the manuscript writing was led by C. D. ACK N OWLED G M ENTS The study was funded by the Association of Zoos and Aquariums; Conservation Endowment Fund No. 08-857, USA; The Christensen Fund, USA; Grant Agency of the Czech Republic (19-28126X); Darwin Initiative for the Survival of Species (22-002) and PNG Mama Graun

F
Distribution of frog species elevation ranges derived from the Mt.Wilhelm community samples (Local) compared with their regional species pool (elevation range is binned in 0-499, 500-999, etc. intervals) (a), and elevation ranges for the New Guinea fauna (binned in 0-99, 100-199 etc. intervals) (b).F I G U R E 5 Median (1st-3rd quartile) elevation range of frog species in 100 m elevation bands in New Guinea.There is no significant elevation trend of elevational range (Spearman r, p > .05).