Endemism and diversity of small mammals along two neighboring Bornean mountains

Study sites

Mt. Kinabalu and Mt. Tambuyukon are two neighboring peaks inside Kinabalu National Park in the Malaysian state of Sabah, Borneo (Fig. 1). This park covers 764 square kilometers. Mt. Kinabalu is the tallest peak in Borneo at 4,095 m and is home to thousands of endemic plant and animal species (Payne et al., 2007; Van Der Ent, 2013). Mt. Tambuyukon (the third highest peak in Borneo, 2,579 m; Fig. 1), despite being only 18 km away, is far less scientifically explored. The vegetation zones as described by Kitayama (1992) for Mt. Kinabalu have been used for simplicity as well as for consistency with previous elevational surveys (Nor, 2001): lowland (>1,200 m), lower montane (1,200–2,000 m), upper montane (2,000–2,800 m) and subalpine (2,800–3,400 m).

The lowland dipterocarp forest dominates both mountains from the lowest elevations up to 1,200 m. Above this elevation begins the lower montane oak forest of 10-25 m trees up to around 1,800-1,900 m on both Mt. Kinabalu and Mt. Tambuyukon. On Mt. Tambuyukon at 1,440 m there is a sharp break to an ultramafic outcrop and the vegetation changes to a low productivity forest with shorter trees. The mossy or cloud forest begins at around 2,000 m on both mountains. This zone is usually immersed in clouds, moss covers most surfaces, and pitcher plants (genus Nepenthes), epiphytes, orchids, and climbing bamboos are abundant. At 2,350 m on Mt. Tambuyukon and 2,600 m on Mt. Kinabalu there is a fast transition to an open stunted forest dominated by Dacrydium and Leptospermum species. At these elevations, the vegetation develops a sclerophyllous and microphyllous syndrome. At 2,800 m the subalpine vegetation appears on Mt. Kinabalu, which is absent on Mt. Tambuyukon.

Field survey

Surveys were conducted in two consecutive field seasons along elevational gradients following climbing trails along Mt. Tambuyukon and Mt. Kinabalu. We targeted small non-volant mammals and further included opportunistic observations and data from trail cameras. Species identification was performed according to Payne et al. (2007). During the first field season, we surveyed Mt. Tambuyukon in June–August 2012. Surveys for the second field season were conducted on select locations on Mt. Tambuyukon (to fill gaps in the first season sampling) and along the full elevational gradient of Mt. Kinabalu in February–April 2013.

We set traps from ~331 to 2,509 m on Mt. Tambuyukon, and from 503 to 3,466 m on Mt. Kinabalu (Dataset S1). The taxa we expected in the small mammal trap surveys included members of the families Soricidae (shrews), Erinaceidae (gymnures), Tupaiidae (treeshrews), and rodents in the families Muridae (mice and rats) and Sciuridae (squirrels). Trapping was conducted following ethical standards according to the guidelines of the American Society of Mammalogists (Sikes, Gannon & Animal Care and Use Committee of the American Society of Mammalogists, 2011). Animal care and use committees approved the protocols (Smithsonian Institution, National Museum of Natural History, Proposal Number 2012-04 and Estación Biológica de Doñana Proposal Number CGL2010-21524). Field research was approved by Sabah Parks (TS/PTD/5/4 Jld. 45 (33) and TS/PTD/5/4 Jld. 47 (25)), the Economic Planning Unit (100-24/1/299) and the Sabah Biodiversity Council (JKM/MBS.1000-2/2(104)).

Line transects were set at approximately every 400–600 m in elevation. On Mt. Tambuyukon, transects were placed along the mountaineering trail markers (placed every one km along the trail) as follows: from Monggis substation to km 1 at 500 m, 900 m (km 7.5), 1,300 m (km 10.3), 1,600 m (km 11), 2,000 m (km 12.6) and 2,400 m (km 13.5). On Mt. Kinabalu the 500 and 900 m transects were located at Poring Hot Springs, from the entrance and along the trail to the Langanan Waterfall. The next elevation transect for Mt. Kinabalu was set at ~1,500 m at the Park Headquarters, ~2,200 m along the Timpohon mountaineering trail (km 2, Kamborangoh), 2,700 m (km 4, Layang-Layang), and 3,200 m (around Waras, Pendant hut and Panar Laban). For reporting results and data analysis, we hereafter refer to these locations as “trapping locations.” Each trapping location gathered the trap data from transects which could span a distance of over 100 m up- or downhill (raw data in Dataset S1).

We set traps at approximately 5–10 m intervals for a total of around 40 traps per transect. Trapping locations are shown in Fig. 1. Collapsible Tomahawk live traps (40 cm long), collapsible Sherman traps (two sizes 30 and 37 cm long), and local mesh-wire box traps were used. We considered traps as “close to the ground” if set below approximately three meters off the ground. Most of these traps were directly set on the ground, while a small proportion was set on branches or vines at the reach of the hand. We considered the three-meter threshold as reasonable to describe the terrestrial small mammal community given the arboreal habits of many small terrestrial Bornean mammals (Wells et al., 2004) and the high-complexity of the vegetation in most transects. Any trap above that Threshold was considered “arboreal.” A bait mixture (of varying composition) consisting of bananas, coconuts, sweet potatoes, palm fruit and oil, vanilla extract and dried fish was placed in each trap. A small number of pitfall traps were distributed from 500 to 2,000 m on Mt. Tambuyukon (number and distribution were limited by the substrate) (Table 1).

Each trapping location had a total of two to four transects. The highest elevation had a lower number of trap nights due to the smaller area available for placement of traps. Coordinates for trapping locations were recorded using Garmin eTrex^® series and Garmin GPSmap 60CSx. The minimum number of trap nights was based on the saturation rates obtained from Nor (2001) at approximately 300 trap nights. For every trapping location the cumulative trap-nights and species richness were calculated by adding the cumulative number of species caught and trap-nights for each trap for each successive day from the date the trap was set.

We set up four camera traps (Reconyx RapidFire RC55™ cameras, and ScoutGuard HCO™ cameras) along the mountaineering trail on Mt. Tambuyukon. Camera 1 was placed at 500 m, at the first-kilometer marker for the hiking trail. Cameras 2 and 3 were placed along the Kepuakan River near km 8 and at approximately 900 m. Camera 4 was placed at approximately 1,300 m near km 10.5. No cameras were deployed along the Mt. Kinabalu trail due to a large number of day hikers and mountain climbers.

Additionally, while on Mt. Tambuyukon we opportunistically recorded mammal observations while walking to, from and along our trap lines, while setting cameras, or while in our campsite.

Alpha diversity

Beta diversity

We calculated beta diversity for each mountain using a Sorensen-based dissimilarity index (β_SOR) with its turnover (β_SIM) and nestedness (β_NES) components (Baselga, 2010). For calculations we used function beta.multi, from package “betapart” 1.3 (Baselga & Orme, 2012) in R. These calculations were also done pairwise between all trapping locations within the same mountain using the function betapart::beta.pair. These distances were used for a cluster analysis using neighbor-joining (Saitou & Nei, 1987) with ape::nj function in R to evaluate the community similarity between mountains and elevations.

We estimated the contribution of endemics to β_SIM and β_NES components of beta diversity. For each mountain, we removed endemic species from the dataset and recalculated β_SIM and β_NES (hereafter referred to as β_SIM-end and β_NES-end). We created subsets of the data to which these measures could be directly compared by randomly sub-sampling 5,000 times without replacement, several species from the complete dataset equal to the non-endemics present in that mountain. P-values were determined from the distributions of the permutated values.

Field survey

The overall trapping success increased toward high elevation on both mountains. On Mt. Tambuyukon, we trapped a total of 295 different individuals (not including recaptured animals) from 21 different species (Dataset S1; Table S1) over 5,957 trap nights, for a total of 5.0% trap success (not including arboreal or pitfall trapping; Table 1). Trap success at each elevation ranged from 2.1% at 1,600 m to 9.6% at 2,400 m. One non-target capture of a carnivore, a Kinabalu ferret-badger (Melogale everetti), was recorded. The trap success calculations were done excluding pitfall traps and arboreal traps (due to inconsistent placement of traps). The use of pitfall and arboreal traps was limited by the time necessary to set and check arboreal traps, difficulty finding appropriate places to set pitfall traps and the high number of tourists on Mt. Kinabalu, which limited the sites we could set traps to those out of sight of the trails. The accumulation of species across trap nights varied across elevations, and appeared near saturation in all trapping locations (Fig. S1). A species of shrew, Suncus sp., was collected in a pitfall trap, and a gray tree rat (Lenothrix canus) in an arboreal trap, bringing the total number of species to 23.

On Mt. Kinabalu, we trapped a total of 20 species from 209 different individuals (Dataset S1; Table S1) over 2,022 trap nights, for an average trap success of 10.3% (Table 1). The trap success across elevations was much higher on Mt. Kinabalu, ranging from 5.6% (at 900 m), to 15.4% (at 2,700 m) (Table 1). This overall higher capture rate resulted in species saturation with a lower number of trap nights on Mt. Kinabalu than on Mt. Tambuyukon (Fig. S1).

Species distribution

The mountain treeshrew, Tupaia montana, was the most frequently caught species (35.7% of all catches) and it had a wide elevational distribution from 836 to 3,382 m. The Bornean mountain ground squirrel (Sundasciurus everetti, formerly Dremomys everetti; Hawkins et al., 2016), the long-tailed giant rat (Leopoldamys sabanus), and Whitehead’s spiny rat (Maxomys whiteheadi) also had large elevational distributions on both mountains (Fig. 2). The lowland (<1,000 m) terrestrial small mammal community was the most diverse, with 19 species trapped. We captured 16 species in the community associated with montane forest between 1,000 and 2,400 m and only seven species at 2,400 m and above (Fig. 2).

We captured a single shrew (Suncus sp.) after 176 pitfall trap nights. A less intensive arboreal trapping effort of 76 trap nights yielded seven individuals from six small mammal species: T. montana (n = 2), Lenothrix canus (n = 1), Callosciurus prevostii (n = 1), Sundasciurus jentinki (n = 1), Sundamys muelleri (n = 1) and T. minor (n = 1). Despite the smaller effort on arboreal trap nights, we still captured two species that were not trapped elsewhere (Lenothrix canus and Sundasciurus jentinki). However, we also caught arboreal species in ground traps or traps close to the ground (<3 m), including Chiropodomys pusillus, Callosciurus prevostii and T. minor.

We set four trail cameras on Mt. Tambuyukon to document larger mammals not targeted by our traps. They documented an additional eight species of mammals (Table 2; Fig. S2). The number of species captured by the cameras varied from one to five, with the camera at 500 m exhibiting the most diversity, both in number of species and number of independent visits (Table 2). They documented four species that were not documented in any other way.

On Mt. Tambuyukon, several species were detected only through direct observation (Table 3). Of these sightings many were documented only a single time, including the orangutan (Pongo pygmaeus), the Bornean giant tufted ground squirrel (Reithrosciurus macrotis), Whitehead’s squirrel (Exilisciurus whiteheadii), and the bearded pig (Sus barbatus). The Bornean gibbon (Hylobates muelleri) was heard singing on an almost daily basis, but only directly observed a single time. The sambar deer (Rusa unicolor) was heard vocalizing once at 1,400 m. Only one observation was made of a carnivore, the Malay civet (Viverra tangalunga), which was observed during a late night-walk. The visual observations increased the diversity of species documented, especially for primates and tree squirrels.

Alpha diversity

Both mountains showed a similar pattern for alpha diversity indices across elevations (Fig. 3; Table S2). Species richness and Shannon diversity were maximum in low elevations and decreased gradually toward high elevations. However, evenness was lowest at middle elevations (U-shaped) (Fig. 3). The high dominance of some species at middle elevations (e.g., mountain treeshrew) leads the Shannon diversity to sink at around 1,500 m in both mountains. However, Shannon diversity increases again toward the highest elevations due to the more even occurrence of the species in the small mammal communities at those elevations, despite lower species richness (Fig. 3). The species richness we report in low elevations for both mountains are above the upper 97.5% quantile of the expected richness expected under the hypothesis of the MDE (Fig. 3A).

We detected a significant negative relationship between species richness and elevation (χ₁² = 8.81, P = 0.003). However, species richness was not correlated with expected species richness under the MDE (χ₁² = 0.54, P = 0.46) (Fig. 3).

Elevation was positively correlated with higher proportion of endemic species (χ² = 7.96, d_f = 1, P = 0.005) and a greater proportion of captures belonging to endemic species (χ² = 10.7, d_f = 1, P = 0.001) (Fig. 4).

Beta diversity

Variation in the species composition assemblages, or beta diversity, was very similar for both mountains (β_SOR = 0.77 on Mt. Kinabalu and β_SOR = 0.74 on Mt. Tambuyukon). Most of this beta diversity derived from the turnover component (β_SIM = 0.73 for Mt. Kinabalu and 0.65 for Mt. Tambuyukon). The nestedness component was very low on both mountains (β_NES = 0.04 for Mt. Kinabalu and 0.09 for Mt. Tambuyukon) (Fig. 5). This indicates that the assemblages at different elevations are not the product of species loss from the richest assemblages. Instead, they are singular assemblages with different species compositions. We found large dissimilarities in the turnover component between pairwise locations within each mountain (β_sim) associated with < ~1,900 vs > ~ 1,900 m locations, compared to lower values within lowland or highland locations (Table S3). When removing endemic species from the dataset, β_NES-end increased and β_SIM-end decreased on both mountains (Fig. 5). The permutations indicated that this decrease in the turnover component (β_SIM-end) was significant in Tambuyukon (β_SIM-end = 0.36, P = 0.002), but not in Kinabalu (β_SIM-end = 0.68, P = 0.36), and that the increase in nestedness (β_NES-end) was significant for Mt. Tambuyukon (β_NES-end = 0.38, P = 0.001), but only marginally significant for Mt. Kinabalu (β_NES-end = 0.14, P = 0.05) (Fig. 5).

The clustering analysis grouped trapping locations in two main groups, above and below ~1,400 m (Fig. S3). Within the high elevation group, all locations above ~1,900 m clustered together.

Alpha diversity

Species richness peaked at low elevations on both mountains coinciding with the lowland dipterocarp forest. Then, it decreased gradually toward the highest elevations where it was lowest (Fig. 3). This pattern deviates from the expectations of the MDE (Fig. 3), and the MDE reported by Nor (2001) in a previous small mammal survey on Mt. Kinabalu with a very similar survey scheme to ours. Colwell & Lees (2000) suggested that a MDE should constitute the null hypothesis over which deviations should be interpreted. However, this point of view is not universal (Rahbek, 1995; McCain, 2007, 2009). Incomplete sampling of the low elevations happens regularly and can artificially create a MDE (Rahbek, 1995; Lomolino, 2001), which may explain the difference between Nor (2001) and our results for the same mountain. Our surveys detected 12 more species as compared to Nor (2001), including a climbing mouse, Chiropodomys pusillus, a tree rat Lenothrix canus, Prevost’s squirrel, Callosciurus prevostii, Jentink’s squirrel, Sundasciurus jentinki, the rats Maxomys rajah, Maxomys alticola and Sundamys muelleri, the non-native species Rattus exulans and Rattus tanezumi, two species of treeshrews Tupaia. longipes and T. minor, and two species of shrews, one trapped in a small Sherman trap, Crocidura sp., and one in a pitfall trap, Suncus sp.. The effects of an incomplete sampling should be more acute in lowland elevations where there is more habitat heterogeneity (Rosenzweig, 1992, 1995) and species might tend to occupy smaller ranges (Rosenzweig, 1995; Brown, 2014). However, we report the highest alpha diversity in the lowest elevations, which supports that our interpretations are not affected by this low-elevation sampling bias. Another observation supporting the comprehensiveness of our sampling is that we documented species in all sites between the lowest and highest occurrences, except for Maxomys whiteheadi on Mt. Kinabalu, and Maxomys ochraceiventer and N. rapit on Mt. Tambuyukon.

Conversely, the gradual decrease of species richness we recorded was explained by elevation alone. Even in the presence of a MDE, a gradual decrease of species richness with elevation seems to be a general pattern in mountain gradients (Rahbek, 1995). There are multiple factors that are correlated with elevation which have been proposed to explain diversity across mountains gradients, but disentangling their effects is difficult given the multicollinearity (Heaney, 2001; Fu et al., 2006; Kluge, Kessler & Dunn, 2006). The strong correlation of alpha diversity with elevation enables further discussion. Diversity has been proposed to peak with precipitation (Heaney, 2001), but on Mt. Kinabalu the peak of precipitation, at around 2,000 m (Kitayama, 1992), did not match the diversity peak. Productivity (Aiba, Takyu & Kitayama, 2005) and temperature (Kitayama, 1992) are negatively correlated with elevation in Kinabalu, and could potentially explain the diversity pattern. Nevertheless, it has been suggested that examining resource availability for this taxonomic group is more appropriate than simply looking at productivity (Brown, 2001; Heaney, 2001). This is a difficult variable to measure that was not incorporated in our original survey. A more plausible explanation for the change in diversity is area and habitat complexity. Available area (Camacho-Sanchez et al., 2018) and complexity of the forest (Kitayama, 1992) decrease with elevation on Mt. Kinabalu. The peak in diversity we find in low elevations is consistent with the spatial hypothesis which states that (1) at the regional level, larger areas (such as the lower elevations in mountains) have lower rates of extinction over speciation (Rosenzweig, 1992) and (2) that larger areas have more types of different habitats, so greater species diversity should be observed in larger areas (Rosenzweig, 1995). Perhaps, the relationship between area and diversity on elevational gradients along large mountains falls somewhere between these processes (McCain, 2007).

Beta diversity

The composition of the small mammal assemblage changed across elevation in a similar way for both mountains. Indeed, all trapping locations in high elevations (above ~1,900 m) were very similar in composition (Fig. S3). Mid-mountain locations showed intermediate compositions whereas the 500 and 900 m locations on both mountains also clustered together. This montane fauna transition was already identified by Nor (2001) to be at around 1,800 m, and it matches approximately the vegetation limit between the lower and upper montane forests on Mt. Kinabalu (Kitayama, 1992). The influence of this shift is reflected in the overall high turnover component of beta diversity (β_SIM) for both mountains (Fig. 5). The pairwise turnover components (β_sim) were highest between lowland—highland locations (Table S3). This indicates the lowland and highland communities are composed of different species, rather than the community with the lowest richness (highland) being a subset of the species present on the richest one (lowland). This pattern has already been described for other small mammals across several mountain systems (Mena & Vázquez-Domínguez, 2005). Unfortunately, there is no consistent data collection from other tall mountains in Sundaland to discuss a common mountain biogeography pattern in this region. Historical expeditions to Sumatra also point to a similar shift in vegetation structure and different mammal assemblages at high elevations above 2,000 m (Robinson & Kloss, 1918, 1919; Miller, 1942).

Pattern of endemism

Beyond the net alpha and beta diversity we describe, we put special attention on the nature of the species that could be driving these diversity patterns. Our models predicted that the proportion of Bornean endemism as well as the proportion captures belonging to endemic species increased with elevation (Fig. 4). At the same time, we found that the high-elevation endemics were responsible for the high turnover component in beta diversity on Mt. Tambuyukon (Fig. 5). The species that had the greatest contribution to this high-elevation endemism were Rattus baluensis, T. montana, Sundasciurus everetti and Maxomys alticola (Table S1). These species are restricted to mountain areas in northern or central-northern Borneo, and were present in high abundance and evenness from 2,000 m (Table S1). Nor (2001) reported high trapping success on Mt. Kinabalu, associated with higher abundance at high elevations, but he did not record Maxomys alticola. Heaney (2001) also recorded the highest abundances in the top elevations in the Philippines, and a peak of endemics at higher elevations. Three additional trapped species are mountain endemics (Melogale everetti, Sundasciurus jentinki and N. rapit) but contributed less to our analysis because of their lower densities or detectability. This pattern of high-elevation endemics could be even more pronounced as the distribution and taxonomy of the highland mammals also found on other islands in Sundaland are further updated and revised (i.e., Hylomys suillus and Sundamys infraluteus; Camacho Sánchez, 2017). A majority of the lowland species are widespread, and also distributed across other Sundaland landmasses such as Sumatra and the Malay Peninsula (Corbet & Hill, 1992).

The pattern of mountain endemics on Mt. Kinabalu has previously been described for other taxa (Merckx et al., 2015). For mammals, this has also been observed by Heaney (2001), Sánchez-Cordero (2001), Swenson et al. (2012) and Chen et al. (2017) on other mountains. Mt Kinabalu is unique in Sunland due to its high elevation (4,095 m) and the scarcity of nearby peaks above 2,000 m. Higher isolation on mountain peaks boosts mountain endemism worldwide (Steinbauer et al., 2016), which could explain the greater endemism in the higher elevations in the Kinabalu range. For instance, a pattern of mountain endemism linked to divergence in allopatry induced by isolation from the combination of topography with past climate changes has been described in Bornean birds (Sheldon, Lim & Moyle, 2015; Moyle et al., 2017; Manthey et al., 2017). A similar pattern of high mountain endemism driven by intra-island speciation has been reported for shrews in Sumatra and Java (Esselstyn et al., 2013; Demos et al., 2016). Presumably, the high degree of isolation of the habitats on the higher slopes of Mt. Kinabalu help to maintain a highly endemic community. This could be due to reduced dispersal and colonization to/from nearby similar areas as proposed by Steinbauer et al. (2016).