Determining the geographic distribution and ecology of the Critically Endangered Kaputar rock skink (Egernia roomi)

Study species

The Kaputar rock skink (Egernia roomi) is a moderately sized (average adult snout-vent length is 105 mm), diurnal, heliothermic and saxicoline skink species, and a member of the Egernia striolata species complex (Sadlier et al. 2019). Like most of its congeners, the Kaputar rock skink is crevice obligate, and prior to this study, known only to use cracks and fissures on exposed basaltic outcrops and plateaux above 1360 m on the Nandewar Range within MKNP (Sadlier et al. 2019; Swan et al. 2022). The species is reliant on rock-on-rock microhabitat and does not construct burrows like members of the related Liopholis genus (which were once considered to be part of Egernia; Gardner et al. 2008). Complex sociality is a common trait within the Egernia genus (Chapple 2003; Riley et al. 2021), and Kaputar rock skinks have been commonly sighted occurring in pairs or small groups (NG and AC, pers. obs.). Prior to our study, the extent of occurrence (EOO) and area of occupancy (AOO) of the species (based on the three known sites) were both 8 km² (NSW Threatened Species Scientific Committee 2021).

Study site

Located 52 km east of Narrabri in north-central NSW (Fig. 1), the Nandewar Range stands as the eroded remains of the Nandewar shield volcano, which is estimated to have formed from volcanic eruptions occurring between 21 and 17 million years ago (Abbott 1969). A westerly isolate of the Great Dividing Range, the highest elevations of the Nandewar Range (MKNP) serve as an outlying mesic upland refuge surrounded by drier slopes and plains, with the isolated, high elevation areas of the range sharing many floral and faunal communities with the disjunct New England Tableland on the Great Dividing Range to the east (Sadlier et al. 2019; Copeland and Backhouse 2022). The Kaputar Summit reaches 1509 m AHD (Australian Height Datum), while the surrounding lowland plains sit at only 230 m (Sadlier et al. 2019). MKNP (~51,400 ha) constitutes most of the remnant, native habitat within the Nandewar Bioregion, which has otherwise largely been converted into agricultural land (DPIE 2021). The park includes the declared wilderness areas of Grattai, Nandewar, and Rusden.

Fig. 1.

Mount Kaputar National Park study region, and the study species (Kaputar rock skink, Egernia roomi, top right inset). Inset map displays location of study area within the state of NSW. G, Grattai; N, Nandewar; R, Rusden designated wilderness areas. Image credit: Nicholas Gale. Aerial imagery sourced from Google Earth (https://earth.google.com, accessed February 2023).

Areas over 1350 m in elevation are considered subalpine and are subjected to snowfall and prolonged frost events during winter months (DPIE 2021). Rugged rocky peaks and plateaux at this elevation are interspersed with open Snow Gum (Eucalyptus pauciflora), Ribbon Gum (Eucalyptus viminalis) and White Gum (Eucalyptus dalrylmpleana) woodland, with patches of dry heathland and Tea Tree (Leptospermum polygalifolium) thickets also present (Murphy and Shea 2015). The central section of the national park, commonly dubbed the Kaputar Plateau, encompasses the Mount Kaputar summit itself (Fig. 1), hosting the highest peaks of the range, and is the most accessible region within MKNP, possessing a partially sealed road to the summit and the majority of the maintained walking tracks within the park. While vaguely defined in literature, we define the Kaputar Plateau as the areas above the 1200 m contour east of Coryah Gap, which encompasses the summit region and associated smaller outcrops and peaks. This includes the Governor, West Kaputar Rocks, Mount Dowe, Mount Lindsay, Bundabulla Circuit/Lookout, Lindesay Rocks, and rock outcroppings east of the Horton River headwaters (Fig. 1). We consider the Bullawa Creek valley a northerly cut-off for the plateau, despite a rock-depauperate lower narrow ridge at 1000 m elevation connecting the area to the Pound Mountain area to the north (seen between sites 23 and 24 in Fig. 2).

Fig. 2.

Individual occurrence records of the Kaputar rock skink detected in the central Nandewar Ranges during targeted surveys. Grattai Mountain (Site 30) is not visualised. Inset map displays survey locations within the MKNP boundary. For site information associated with site numbers, see Supplementary material. Aerial imagery sourced from Google Earth (https://earth.google.com, accessed February 2023).

Site selection and definition

Prior to our study, the Kaputar rock skink was known only from three sites: (1) the Mount Kaputar summit (Site A); (2) Mount Dowe (Site B); and (3) The Governor (Site C) (Fig. 2). Survey sites were selected using aerial satellite imagery in order to discern areas of outcropping rock within the study area (Mount Kaputar NP region >1000 m), constituting the known suitable habitat for the target species. To be certain that all accessible rock habitat was covered, all formed walking tracks (including those under a canopy) over this elevation were hiked in order to confirm rock habitat coverage. This identified one site that was hidden from aerial imagery, a westerly facing rocky cliff along Coryah Gap Fire Trail (Site 4). We considered sites as distinct if there was a greater than 50 m discontinuity in rocky habitat between them. Twelve survey sites fell within two declared wilderness areas within the park. These included the Nandewar Wilderness north of the Kaputar Plateau, and Rusden Wilderness south of the Kaputar Plateau (Fig. 1). The third wilderness area north of Killarney Gap (Grattai) was surveyed in October 2023. Sites ranged in size from <0.01 km² (Sinclair Peak, Site 9) to 0.62 km² (East Bundabulla Cliffs, Site 18). In total, 30 individual sites were surveyed, inclusive of Mount Lindsay and Grattai Mountain (see Supplemetary Table S3).

Visual encounter surveys

Although range restricted, the Kaputar rock skink is not cryptic, and is readily detected when basking under ideal conditions. The species can also be seen in crevices during times of inactivity, including periods of light to moderate rain and fog (NG, pers. obs.). Repeated diurnal surveys were carried out from early September 2022 to mid-January 2023, during the skink’s active season. Searches were carried out by two surveyors, and only during conditions where Kaputar rock skink detection was favourable (i.e. temperatures over 10°C, and clear of rain). Suitable habitat was scouted from afar using a combination of binoculars (Vortex Diamondback HD 10 × 42) and unaided eyes to observe basking or active lizards, and when approached, crevices were inspected using a torch to detect sheltering skinks. After a skink was detected and presence at a site was confirmed, the survey continued until all suitable habitat was exhausted, in order to maximise individual occurrence records. All sites were surveyed at least once, or twice if the Kaputar rock skink was not detected on the first visit. The iPhone application Sightings (Ug Media https://www.ugmedia.com.au) was used to record each individual sighting, taking coordinates to an accuracy of ±5 m.

Updating EOO, AOO and fine-scale habitat mapping

The minimum convex polygon surrounding all records was calculated in QGIS (ver. 3.28.3) using the attribute table field calculator to determine the EOO of the species. The IUCN recommended 2 × 2 km² grid method was used to quantify the AOO. To achieve fine-scale habitat mapping of the Kaputar rock skink, once a site was confirmed to have the skink present, ends and beginnings of habitat were recorded in Avenza Maps (Avenza Systems; https://www.avenza.com/avenza-maps/) and modified into a polygon using QGIS to encompass discrete skink habitat; the area of these polygons (and thus habitat area) was calculated in km². This included the three Sadlier et al. (2019) sites, which we revisited and surveyed.

Species distributional modelling (SDM)

We predicted the distribution of suitable Kaputar rock skink habitat using the Maximum Entropy species distribution modelling algorithm, MaxEnt (Phillips et al. 2006). Species distribution modelling explores the contrast between environmental conditions of occupied sites (i.e. presence locations) and potentially unoccupied sites (i.e. locations of the environmental background), to estimate habitat suitability for a species across a landscape. As a presence only algorithm, MaxEnt performs well for small sample sizes and therefore is the best choice for modelling the distribution of small range and micro-endemic species (Hernandez et al. 2006). Individual Kaputar rock skink occurrence records (118), obtained from presence/absence surveys and the visitation of historically known sites, were used to model the species’ distribution. To minimise spatial autocorrelation for the species distribution modelling, we spatially thinned the Kaputar rock skink occurrence records to 1 km using the ‘thin’ function in the R package spThin (Aiello-Lammens et al. 2015), resulting in 10 occurrence points (i.e. 10 1 km² presence pixels) used for modelling.

Our primary aim was to model habitat suitability within a geographical area approximating the potential range of the Kaputar rock skink, to confirm all suitable habitat for the species has been surveyed. We therefore set the modelling background as the extent of all biogeographic regions (Interim Biogeographic Regionalisation for Australia (IBRA) ver. 7.0) occupied or likely to be occupied by the Kaputar rock skink. This only included the Nandewar Bioregion (NAN). Within this bioregion, the species’ currently known distribution is limited to the Kaputar subregion (NAN03). The only other subregion included in the background was the neighbouring Peel subregion (NAN04), abutting the Kaputar subregion to the east, relevant due to its possessing mountainous terrain >1000 m. The other Nandewar subregions, NAN01 (Nandewar Northern Complex) and NAN02 (Inverell Basalts) do not have any mountainous terrain above 1000 m and were therefore excluded from the modelling background.

We pre-selected a set of 27 candidate predictor variables likely affecting the species’ distribution. These variables relate to climate, topography, terrain and vegetation. We primarily included bioclimatic variables because as a high elevation ectotherm, the distribution of the Kaputar rock skink is likely influenced by temperature and precipitation regimes, as is the case for many ectothermic species (Araújo and Guisan 2006; Powney et al. 2010). As the species is known only to be associated with high elevation rock habitat (Sadlier et al. 2019; Swan et al. 2022), we included Terrain Ruggedness Index (TRI), topographic wetness, and aridity as candidate predictor variables. All environmental layers were downloaded at, or resampled to, a spatial resolution of ~1 km² (30 arc-seconds). Variables were resampled to fit the study background region using the ‘mask’ function of the ‘raster’ package (Hijmans et al. 2013). We then extracted environmental data from the locations of the 10 thinned Kaputar rock skink occurrence points.

To investigate and reduce multicollinearity between variables, we visually inspected PCA biplots and calculated the variance inflation factor (VIF) using the ‘vifstep’ function of the usdm package for R (Naimi et al. 2014) retaining those with VIF <5. We also calculated Pearson correlation coefficients between each pair of variables and retained the variable with greater ecological relevance from each pair of highly correlated variables (P ≥0.7; (Dormann et al. 2013). The final set of variables (which were uncorrelated and biologically relevant) used for model fitting were as follows: Temperature Seasonality (Bio4), Thornthwaite Aridity Index, Normalised Difference Vegetation Index (NDVI), and TRI. Having few predictor variables is beneficial to reduce the risk of overfitting, a common issue when modelling ranges of taxa with small distributions (Anderson and Gonzalez 2011). We used the ‘ecospat.mantel.correlogram’ function in the ecospat package (Di Cola et al. 2017) to confirm that spatial autocorrelation of occurrence records was not significantly different than 0.

Models were built with regularisation multiplier options set to 1, 2 and 3 and with five different feature classes combinations (1, Linear; 2, Quadratic; 3, Linear + Quadratic; 4, Linear + Quadratic + Hinge; 5, Hinge); this resulted in 15 parameter combinations (and hence 15 candidate models). Feature classes determine the flexibility of the shape of the relationship of covariates to response. The regularisation multiplier dictates the penalty for model complexity, with lower values resulting in a model more fitted to the presences (over-fitted), and higher values in more spread out (under-fitted) model predictions (Phillips and Dudík 2008; Elith et al. 2011; Merow et al. 2013). Notably, as one of our aims was to predict new areas of potential occurrence, we did not use values of the regularisation multiplier lower than 1. Together, these settings influence model complexity by controlling the number and complexity of features in the final output. All the models were evaluated using the metrics AICc, AUC_TEST, AUC_DIFF, ORmtp, OM₁₀ and Boyce Index (see Supplementary material for definitions and explanations of these metrics). We then summarised the evaluation metric scores across five models of each parameter combination. We visualised the predictions of each parameter combination’s full model, and we qualitatively evaluated their plausibility against our knowledge of the species’ ecology and of the region’s biogeography (following Petitpierre et al. 2017). We ultimately chose the parameter combination that yielded MaxEnt models with the lowest AIC score (corrected for small sample sizes; AICc; Warren and Seifert 2011), that performed well for multiple other metrics, was relatively parsimonious compared to other models (Merow et al. 2013; Coelho et al. 2019), and yielded ecologically plausible predictions.

Threats

While conducting fieldwork, we noted the presence of feral goats (Capra hircus) through (1) visual observation of animals, and (2) pellets (scats) at each site. In particular, the fouling of crevices with pellets was examined, and could be assessed during surveys whilst crevices were searched for skinks. Evidence of fire was recorded at each site, to ground truth the fire footprint. On top of this, human-induced damage was also noted, including obvious dislodgment and fracturing of large rocks and the creation of rock cairns.

Fire overlap

The minimum convex hull including all records of the species (i.e. the EOO) was overlain atop the Australian Google Earth Engine Burnt Area Map (AUS GEEBAM) (DAWE 2020) to ascertain the extent to which fire has overlapped the newly defined range of the species.

Presence/absence surveys

At total of 114 occurrence observations of the Kaputar rock skink were recorded at 15 additional sites on the Nandewar Range, extending the species’ range 3.5 km east and 2.5 km south of Mount Dowe (Site B), which was previously considered the most south-easterly site for the species (Fig. 2). This brings the total number of known occurrence sites to 18, from the three previously noted sites in the recognition paper (Sadlier et al. 2019). At sites where the species was present, skinks were always found on the first survey, and time until first detection ranged from 2 to 35 min (Table 1), with the mean time until detection being 13.2 min.

Along with the Kaputar rock skink, eighteen other reptile species were detected during diurnal skink surveys, including the congeneric Cunningham’s Skink (Egernia cunninghami) and Tree Skink (E. striolata). This species distribution and assemblage pattern data is to be dealt with elsewhere.

A new elevational low for the species was recorded at 1147 m (Site 18, East Bundabulla Cliffs, Fig. 2), with the previously known lowest elevation for the species at The Governor recorded at 1360 m (Sadlier et al. 2019). Skinks were also observed on rocks immediately at the Summit lookout (Site A, 1509 m), with the Kaputar rock skink now known to have an elevational range of 1147–1509 m. This is a narrow band of only 362 m, although a three-fold increase from the previously noted 120 m, between 1360 and 1480 m (Sadlier et al. 2019). The species is now known from elevations that are not considered strictly subalpine (characterised by the presence of E. pauciflora woodland), but montane areas <1350 m. We did not detect the Kaputar rock skink at 14 sites (Fig. 2), including Grattai Mountain.

The species’ updated distribution is still limited to the Kaputar Plateau, as well as rock outcroppings coming off the plateau to its south (Bundabulla Cliffs east and west, Sites 15 and 18). The Kaputar rock skink’s range is bound in the east by the end of suitable rock habitat as the Nandewar Range tapers off into lowland plains, and in the west by Coryah Gap, as habitat becomes dry open sclerophyll woodland dominated by Stringybark (Eucalyptus laevopinea) and Manna Gum (E. viminalis), lacking rock habitat. Bullawa Creek and its associated deeply incised valley provides a northerly dispersal barrier, whereas the end of extensive rock on the Kaputar Plateau and the beginning of open Eucalypt woodland south of Bundabulla Cliffs constitutes the southern cut-off for the species.

We did not observe Kaputar rock skinks using trees or fallen logs, neither during surveys nor when walking between survey sites. This is despite the congeneric tree skink (E. striolata) commonly using trees and logs for shelter, including in the nearby (albeit lowland and semi-arid) Pilliga Forest (Duckett et al. 2012; Murphy and Murphy 2015). As expected, the majority of our new occurrence sites came from areas lacking walking tracks, being generally difficult to access (requiring rock scrambling or considerable hikes off formed trails in wilderness areas). New sites that were easily accessible on walking tracks included Mount Lindsay (Site 20), Eckford Lookout (Site 12) and Bundabulla Lookout (Site 14). Skink observations did not come from beside the walking track itself, but more so on the craggy cliffs bordering these aforementioned sites.

The most isolated populations, east of the Horton River (Sites 27 and 28), are noteworthy given their disjunction from the two other closest populations by what appears to be substantial dispersal barriers. They are separated from the Lindesay Rocks population (1.1 km to the west) due to the deeply incised Horton River valley. They are even further disjunct from summit populations (3.5 km to the north-west), with complete lack of habitat connectivity between these sites and Mount Capel atop the Kaputar Plateau, as the area only supports open woodland depauperate of rock habitat, now thick with vegetation regrowth (Acacia spp.) post fire. Furthermore, no E. striolata-complex skinks were detected on or around Mount Capel (Fig. 2). It is unknown whether these populations were present before the carving of the Horton River valley or later colonised from the summit population.

Lindesay Rocks (Site 25) constitutes the second-most isolated population, straddled by the Horton River in the east and Middle Creek to the west, separating the population from those at Mount Lindsay (Site 20) by 800 m (Fig. 2). While previous Kaputar rock skink records were only known from the Kaputar Plateau rim, the skink was detected on Sinclair Peak (Site 9), a miniscule, sharp rocky peak at only 1200 m² in area, surrounded by open woodland toward the westerly extent of the plateau. No obvious habitat corridors connect the peak to The Governor, the next closest population, 600 m to the north-west (Fig. 2). The Governor remains the most westerly population of the species.

Updated EOO, AOO and fine-scale habitat mapping

The updated EOO of the species is 11.97 km². The updated AOO is 40 km², when calculated using the IUCN recommended 2 km × 2 km grid method. In cases where the AOO is larger than EOO, the IUCN suggests that EOO should be changed to equal the AOO for the purpose of conservation status assessment (IUCN 2022). Therefore, the EOO and AOO of the Kaputar rock skink are both 40 km². Fine scale habitat mapping revealed 1.51 km² of suitable rocky habitat, thus it is likely that only 1.51 km² of habitat within the species’ AOO is ecologically suitable.

Species distributional model (SDM)

Model performance

The optimal model (i.e. with the lowest ΔAICc value) performed well with an AUC of 0.997 and included a regularisation multiplier of 2 and LQ feature class. The geographical prediction of the model is in Fig. 3. Aridity and temperature seasonality (Bio4), contributed 90% and 10% to the model respectively, suggesting that aridity is the primary environmental variable that correlatively predicts the species’ distribution. NDVI and TRI, used as proxies for rock habitat, did not contribute substantially to the model.

Fig. 3.

Species distribution model (SDM) of the Kaputar rock skink, in Mount Kaputar National Park within the Kaputar Subregion. Inset map displays a close up of the Mount Coryah landform and the Kaputar Plateau, along with associated peaks and mountains important for context. Only habitat suitability grids >0.5 are visualised.

Threats

At MKNP, the presence of goats was noted at almost all sites except Sinclair Peak (Site 9), by visual observations of live animals, scat, or both. However, no scats were ever found to physically contaminate crevices during skink surveys, despite being commonly found in small piles atop outcrops. Substantial human damage to rock habitat was noted at four locations with Kaputar rock skinks, all of which are easily accessible to the public and on walking tracks; these included the Summit, Mt Dowe, The Governor and Mount Lindsay.

Fire overlap

When the newly defined EOO of the Kaputar rock skink is overlain atop the Australian Google Earth Engine Burnt Area Map (AUS GEEBAM) (DAWE 2020), the entire range of the species falls within the 2019 burn scar (Fig. 4). Previous projections of the species EOO (areas above 1000 m on the Nandewar Range) led to only a 67% burn overlap estimate (Sadlier et al. 2019; Legge et al. 2022b).

Fig. 4.

Kaputar rock skink records and extent of occurence (EOO) overlapping the 2019 fire scar on the central Nandewar Range region. Aerial imagery sourced from Google Earth (https://earth.google.com, accessed February 2023).

Distribution

Our comprehensive surveys throughout the higher elevation areas of MKNP, combined with species distribution modelling, have allowed us to determine the distributional range of the Kaputar rock skink. While we have detected the Kaputar rock skink at sites throughout a broader region of the National Park, the distribution of the species remains extremely localised and limited to small, discrete rocky habitat ‘islands’ atop the Kaputar Plateau. This mirrors the ranges of other short-range taxa endemic to the Nandewar Range, namely a community of endangered land snails (Murphy and Shea 2015; Murphy et al. 2019). Despite having the bulk of their distributions on the Kaputar Plateau and nearby high elevation peaks, these land snails have also been detected in lower elevation dry rainforest thickets (Murphy et al. 2019), with the Kaputar rock skink proving to be even more restricted than these species. Given that small distribution size is the greatest predictor of extinction risk for terrestrial vertebrates (Sekercioglu et al. 2008), the species remains a high conservation priority.

With extensive exploration and survey of the Kaputar plateau, supplemented by sampling of nearby high peaks, the distributional extent of the Kaputar rock skink is closer to being fully understood. Most accessible rock outcrops above 1000 m have been surveyed for skinks in the central region of MKNP, constituting the highest peaks of the Nandewar Range, and being the most likely habitat to harbour the Kaputar rock skink. All areas deemed potentially suitable for the species have been surveyed, making it unlikely that the species exists elsewhere. The Kaputar rock skinks’ range is now known to be smaller than the previously projected AOO (134 km²), but larger than the previously known AOO (8 km²). With our new distributional knowledge of the Kaputar rock skink, its IUCN listing can be assessed under criteria B (geographic range). With the newly calculated AOO of 40 km², the Kaputar rock skink exceeds the threshold for Critically Endangered (<10 km²), and now only qualifies for Endangered (AOO <500 km²), under IUCN guidelines and criteria (IUCN 2022).

Threatening process

Limitations and future direction

With our surveys being concentrated on rocky habitat, we cannot rule out the absence of the species from wooded areas, given other Egernia species readily utilise crevices in both standing trees and fallen timber (Chapple 2003; Swan et al. 2022). However, on Mount Kaputar, Eucalypt species (such as E. dalrylmpleana, E. pauciflora, and E. viminalis) from sub-alpine elevations (>1350 m) may not provide suitable shelter sites for the skink, as they possess a smooth trunk that do not crack and fissure, as opposed to Ironbark species such as Eucalyptus crebra, of which are utilised by the congeneric Tree Skink (E. striolata) in the nearby Pilliga Forest (Murphy and Murphy 2015).

Despite our surveys capturing much of the rock from the highest elevation areas of the Nandewar Range, we cannot definitively call this a full survey effort of the Nandewar Range. Rocky areas around the Pound Mountain/Doyles Peak in the Nandewar Wilderness and Gins Mountain in the Grattai Wilderness possess extensive rock that may harbour the species. Further focused surveys in wooded areas and in other high elevation rock habitat in the national park may further extend the species’ known range, as well as challenge the assumptions of the species having strictly saxicoline habits.