Introduction
Fire is a natural process that has shaped landscapes globally for more than 420 million years (Bowman et al., Reference Bowman, Balch, Artaxo, Bond, Carlson and Cochrane2009) and is crucial in maintaining the health of many ecosystems. Yet wildfires can lead to long-lasting changes in ecosystem structure, species distributions and species persistence (Haslem et al., Reference Haslem, Kelly, Nimmo, Watson, Kenny and Taylor2011; Lovich et al., Reference Lovich, Quillman, Zitt, Schroeder, Green and Yackulic2017; Merrick et al., Reference Merrick, Morandini, Greer and Koprowski2021; Smith et al., Reference Smith, O'Connor, Granger and Beaumont2021). At a regional scale, the extent, frequency and severity of fires can shift because of land-use changes such as human settlement (Spyratos et al., Reference Spyratos, Bourgeron and Ghil2007) and logging (Lindenmayer et al., Reference Lindenmayer, Kooyman, Taylor, Ward and Watson2020). At a global scale, human-induced global warming is driving shifts in fire regimes, with forecasts and correlative studies predicting increases in the extent, frequency and severity of wildfires (Pitman et al., Reference Pitman, Narisma and McAneney2007; Canadell et al., Reference Canadell, Meyer, Cook, Dowdy, Briggs and Knauer2021; van Oldenborgh et al., Reference van Oldenborgh, Krikken, Lewis, Leach, Lehner and Saunders2021). This increased prevalence of megafires has resulted in some authors referring to the contemporary period as the Pyrocene (Nimmo et al., Reference Nimmo, Carthey, Jolly and Blumstein2021).
Determining how different fire regimes directly and indirectly affect species is a central research priority in contemporary conservation science (Driscoll et al., Reference Driscoll, Lindenmayer, Bennett, Bode, Bradstock and Cary2010). Although species in fire-prone landscapes have probably adapted to cope with the effects of fires, increased severity and frequency of fires could exceed the capacity of these adaptations to facilitate population persistence (Pausas & Parr, Reference Pausas and Parr2018; Nimmo et al., Reference Nimmo, Carthey, Jolly and Blumstein2021). For example, in a fire-adapted landscape, a survey conducted 20 years after an extreme fire found only 7.9% of 1,630 shrubs showed any resprouting (Nicholson et al., Reference Nicholson, Prior, Perry and Bowman2017). Beyond direct mortality, fire can have species-specific effects on habitat structure (Costa et al., Reference Costa, Pantoja, Sousa, de Queiroz and Colli2020), fecundity (Smith et al., Reference Smith, Bull and Driscoll2012) and predation pressure (Leahy et al., Reference Leahy, Legge, Tuft, McGregor, Barmuta and Jones2016) that can cause shifts in ecosystem dynamics. Because these effects are species-specific, comprehensive information across a range of taxa is required to obtain a clear understanding of how ecosystems respond to fire, which is particularly important considering the increased risk of megafires (Canadell et al., Reference Canadell, Meyer, Cook, Dowdy, Briggs and Knauer2021).
Mounting fears regarding the increased extent of wildfires were realized in the 2019–2020 austral bushfire season (hereafter ‘Black Summer’), during which prolonged drought led to catastrophic bushfires in the forested landscapes of south-eastern Australia (van Oldenborgh et al., Reference van Oldenborgh, Krikken, Lewis, Leach, Lehner and Saunders2021). During the Black Summer c. 97,000 km2 of vegetation on the Australian mainland burnt (Ward et al., Reference Ward, Tulloch, Radford, Williams, Reside and Macdonald2020). Bushfires of such an extent and severity were historically unprecedented in the Australian landscape (Gibson et al., Reference Gibson, Danaher, Hehir and Collins2020; Ward et al., Reference Ward, Tulloch, Radford, Williams, Reside and Macdonald2020). Understanding the conservation implications of these megafires has been a key concern for scientists and land managers, with estimates suggesting that 2.8 billion vertebrates were occupying areas affected by the fires (van Eeden & Dickman, Reference van Eeden, Dickman, Rumpff, Legge, van Leeuwen, Wintle and Woinarski2023) and that the range of more than 13,000 species of invertebrates was affected (Marsh et al., Reference Marsh, Bal, Fraser, Umbers, Latty and Greenville2022). Some field surveys have confirmed the dire predictions of the impacts of the Black Summer fires, with reports of adverse impacts on some species of lizards (Letnic et al., Reference Letnic, Roberts, Hodgson, Ross, Cuartas and Lapwong2023), snails (Decker et al., Reference Decker, Foon, Köhler, Moussalli, Murphy and Green2023), birds (Lee et al., Reference Lee, Callaghan and Cornwell2023), frogs (Beranek et al., Reference Beranek, Hamer, Mahony, Stauber, Ryan and Gould2023) and bats (Law et al., Reference Law, Madani, Lloyd, Gonsalves, Hall and Sujaraj2022a). However, the effects of the Black Summer appear to be species- and context-specific, with some frogs, invertebrates, lizards and snakes showing varying responses to the severity of the fires and some being relatively unaffected (Webb et al., Reference Webb, Jolly, Hinds, Adams, Cuartas-Villa, Lapwong and Letnic2021; Foon et al., Reference Foon, Moussalli, McIntosh, Laffan and Köhler2022; Reid et al., Reference Reid, Runagall-McNaull, Cassis and Laffan2022; Hartley et al., Reference Hartley, Blanchard, Clemann, Schroder, Schulz, Lindenmayer and Scheele2023; Letnic et al., Reference Letnic, Roberts, Hodgson, Ross, Cuartas and Lapwong2023).
The variability in species’ responses to fire highlights that effects of such extreme disturbance events are complex and poorly understood (Ratnayake et al., Reference Ratnayake, Kearney, Govekar, Karoly and Welbergen2019; Jolly et al., Reference Jolly, Dickman, Doherty, van Eeden, Geary and Legge2022). Further complications arise regarding species that are rare or elusive because of their low detectability (Bellemain et al., Reference Bellemain, Swenson, Tallmon, Brunberg and Taberlet2005). Similarly, for species or ecosystems for which baseline ecological knowledge is lacking, ecological shifts may not be detected because of shifting baselines (Mills et al., Reference Mills, Waudby, Finlayson, Parker, Cameron and Letnic2020). Snakes are a taxonomic group that is subject to these combined effects because they are generally under-represented in ecological research (Pyšek et al., Reference Pyšek, Richardson, Pergl, Jarošík, Sixtová and Weber2008; Trimble & van Aarde, Reference Trimble and van Aarde2012) and often show cryptic behaviour, which makes them difficult to detect (Mazerolle et al., Reference Mazerolle, Bailey, Kendall, Royle, Converse and Nichols2007). The lack of baseline information combined with generally low detectability hampers the conservation of snakes in the wake of events such as the Black Summer fires.
The mustard-bellied snake Drysdalia rhodogaster, also referred to as rose-bellied snake or Blue Mountains crowned snake, is a small (maximum total length c. 400 mm), diurnal, elapid snake (Plate 1) endemic to New South Wales in south-eastern Australia. Prior to the Black Summer bushfires, D. rhodogaster was considered abundant and was categorized as Least Concern on the IUCN Red List when it was last assessed in 2017 (Shea et al., Reference Shea, Cogger and Greenlees2018). However, because of its small size and cryptic habits, little is known about the species’ ecology (Shine, Reference Shine1981; Scanlon, Reference Scanlon2000). Previous studies indicate that D. rhodogaster is live-bearing, occurs in areas with mild to cool climates and primarily consumes small scincid lizards (Shine, Reference Shine1981). Because of a lack of knowledge about its ecology and preliminary assessments estimating that more than 30% of its distribution had burnt, D. rhodogaster was provisionally listed as needing conservation assessment following the Black Summer bushfires (Legge et al., Reference Legge, Woinarski, Garnett, Nimmo, Scheele and Lintermans2020).
One of the mustard-bellied snakes Drysdalia rhodogaster that we detected during the field surveys. Photo: M. J. Hodgson.
Here we present a multifaceted investigation of the impacts of the Black Summer bushfires on D. rhodogaster. Firstly, we use habitat suitability models to predict the extent of bushfires across the range of D. rhodogaster during the Black Summer. Secondly, we report on the findings from field surveys and occupancy modelling evaluating the effects of fire severity and fire extent on the site occupancy and detectability of D. rhodogaster. Thirdly, to further evaluate whether D. rhodogaster continued to occur in areas burnt during the Black Summer bushfires, we extracted records of D. rhodogaster made after the fires from a publicly accessible database (Atlas of Living Australia, 2021) and intersected these with publicly available information on the severity of the Black Summer bushfires.
Study area
Drysdalia rhodogaster is found in dry sclerophyll forest, heath and woodlands across coastal regions of south-eastern Australia (Plate 1; Shine, Reference Shine1981). Accordingly, we conducted surveys in burnt and unburnt forests in the Greater Sydney Basin of New South Wales. We established survey sites within forest dominated by Eucalyptus spp. in Morton National Park (n = 27), Blue Mountains National Park (n = 15) and Wollemi/Yengo National Park (n = 19; Fig. 1) in close proximity (within 5 km) to localities where D. rhodogaster had been recorded previously according to the Atlas of Living Australia (2021). Individual sites were at least 500 m apart. Movement ecology studies of the Australian elapid Hoplocephalus bungaroides indicate that the maximum distance snakes travel in a month rarely exceeds c. 500 m (Webb & Shine, Reference Webb and Shine1997), thus we assumed a distance of 500 m would ensure independence amongst sites for D. rhodogaster, which is a similarly sized elapid. Surveys were conducted in successive years following the Black Summer, during October 2020–April 2021, September 2021–February 2022 and September 2022–December 2022, in the warmer months when D. rhodogaster is most active (Shine, Reference Shine1981).
The area surveyed for mustard-bellied snake Drysdalia rhodogaster occupancy in south-eastern Australia during October 2020–April 2021, September 2021–February 2022 and September 2022–December 2022, indicating survey sites with and without detections of the species, and the area burnt in the Black Summer bushfires (7 January 2019–24 March 2020).
Methods
Habitat suitability models
We constructed a habitat suitability model for D. rhodogaster using the maximum entropy (MaxEnt) algorithm in the package dismo (Hijmans et al., Reference Hijmans, Phillips, Leathwick and Elith2017) in R 4.1.2 (R Core Team, 2021). We downloaded occurrence records from the Atlas of Living Australia (2021) and New South Wales Bionet Atlas (New South Wales Government, 2022). The Atlas of Living Australia is a biodiversity database that aggregates records from Australian institutions (museums, public sightings, citizen science projects and surveys). It also contains weekly integrations of Australian-based records from the global citizen science project iNaturalist (2024). Similarly, Bionet aggregates records from state-issued scientific datasets and surveys conducted by the environmental regulator. We cleaned the records from both sources and removed duplicates. This resulted in 271 occurrence records, further filtered to 198 unique records within 1 km2 cells (reducing sample redundancy) that we input into the MaxEnt model algorithm along with 38,704 background records (Supplementary Material 1). We ran the models with 43 environmental variables that estimate climate, vegetation and soil conditions at those points (Supplementary Table 1). We chose these variables a priori to represent key determinants of habitat suitability for reptiles (Cabrelli et al., Reference Cabrelli, Stow and Hughes2014; Cabrelli & Hughes, Reference Cabrelli and Hughes2015).
Our model provided gridded estimates of habitat suitability at a spatial resolution of 280 m (the resolution of the gridded environmental variables). We then categorized these raw habitat suitability values into a binary representation of suitability (0 = no to low suitability, 1 = suitable) using a species-specific threshold: the 10th percentile training presence logistic threshold. This is based on the weighting of different model errors (commission errors, whereby a grid cell is classified as suitable when it is not suitable, vs omission errors, whereby a grid cell is classified as unsuitable when it is suitable).
To explore the effect of elevation on the distribution of D. rhodogaster, we plotted the latitude of database records and the centroids of grid cells generated by the habitat suitability model with a probability of occurrence > 0.5 against their elevation (in m). We extracted elevation data from the Australian Government's 3 second Shuttle Radar Topographic Mission Derived Digital Elevation Model Version 1.0 (Geoscience Australia, 2010).
To determine the approximate area of predicted habitat for D. rhodogaster that was burnt during the Black Summer, we intersected points from the suitability model with an aggregated layer of the fire extent from the Black Summer fires. A more detailed description of the methods used for environmental variable selection, data cleaning, model construction, calibration and evaluation is provided in Supplementary Material 1 and Supplementary Table 2, and the environmental layers used are in Supplementary Table 1. We treated our model predictions as hypotheses for the potential distribution of D. rhodogaster (Lee-Yaw et al., Reference Lee-Yaw, McCune, Pironon and Sheth2022), and therefore these hypotheses required independent testing through further field sampling.
Field surveys for D. rhodogaster
At each site on each survey occasion we searched a 100 × 50 m area for 1 person-hour (active search) to standardize for the variable number of surveyors between sites. We deployed artificial refugia consisting of four roof tiles (42 × 33 cm) and two sheets of tin (c. 1 m2) at each site and checked these at the conclusion of the searches. Little is known about the activity periods of D. rhodogaster, with anecdotal evidence suggesting that the species is primarily diurnal; however, the species has also been found to be active at night (MJH, pers. obs., 2022). Because of the dearth of knowledge about D. rhodogaster activity patterns, we conducted the surveys under conditions assumed to maximize snake detection, avoiding excessively warm conditions or heavy rain. Searching included looking for surface active animals, turning over suitable refuges and raking litter. Where possible we minimized disturbance to the microhabitat and in all instances we replaced turned objects back to their original positions.
There was a minimum of 7 days between surveys at each site, with most sites having at least 14 days between surveys. The number of repeat surveys conducted within and across years varied between sites (1–8 surveys per site). We sampled only a subset of sites (Blue Mountains National Park, n = 14; Morton National Park, n = 21) 3 years post-fire. We sampled one Blue Mountains National Park site only 2 years post-fire.
Post-fire occurrences of D. rhodogaster
To complement our field surveys and further assess responses of D. rhodogaster across the species range, we analysed post-fire presence records from the Atlas of Living Australia (downloaded on 25 March 2022). We extracted records of D. rhodogaster that were observed in the 24 months following the end of the bushfires in the greater Sydney region (10 February 2020–1 March 2022). We excluded records of D. rhodogaster in the Atlas of Living Australia that were generated from the field surveys conducted for this study. We determined whether each record had been detected in a burnt or unburnt region by extracting burn severity values from a remote sensing dataset that was created after the Black Summer fires to quantify the extent and severity of fires in New South Wales (Fire Extent and Severity Mapping dataset; Department of Planning, Industry and Environment, 2020).
Covariates and statistical analysis
To determine site occupancy, we ran a single-species, single-season occupancy model in R. We built detection histories for each site by assigning a score of ‘0’ if we did not detect a snake, ‘1’ if we did detect a snake and ‘NA’ for no survey. We ran the occupancy models in the unmarked package in R, using the occu function (Fiske & Chandler, Reference Fiske and Chandler2011). As surveys occurred across multiple years, we included ‘year’ as a site-level covariate. We also included ‘park’ as an occupancy covariate because the initial occupancies probably varied between the parks.
To assess the impacts of bushfire on D. rhodogaster, we calculated the extent of habitat burnt within a 1,000-m radius around each survey site using data extracted from the Fire Extent and Severity Mapping dataset in ArcGIS (Esri, USA) and estimated the fire severity at each site. We assessed fire severity based on evidence of scorching and canopy condition in the first round of surveys conducted in spring 2020 (Letnic et al., Reference Letnic, Roberts, Hodgson, Ross, Cuartas and Lapwong2023). We classified sites as being burnt at low severity if their understorey showed evidence of recent burning (scorch marks on trees, burnt stumps and shrubs) but the canopy of Eucalyptus trees remained intact. We classified sites as being burnt at high severity if evidence of recent burning was observed in the understorey and in the canopy (i.e. leaves in the canopy were either absent or evident as epicormic buds).
To account for seasonality in the detectability of reptiles, we included an observation-level covariate of ‘day of year’, which we defined as the difference between the survey date and the start of the austral spring (1 September). Finally, because air temperature affects the detection of cryptic reptiles (Scroggie et al., Reference Scroggie, Peterson, Rohr, Nicholson and Heard2019), we included ‘daily maximum temperature’ as an observational covariate. We extracted daily maximum temperature data from the SILO climate database (Jeffrey et al., Reference Jeffrey, Carter, Moodie and Beswick2001; Queensland Government, 2023), which provides interpolated daily temperature data at a 5 × 5 km grid resolution. Given that this method treats sites across years as different, if a site had no surveys in a given year, then we omitted that year from the final analysis (n = 27). We standardized fire extent, day of year and daily maximum temperature (mean = 0 ± SD 1) using the scale function in R. If daily temperature data could not be accurately assigned to a site visit, we removed it from the final occupancy analysis.
We constructed a global model, a null model and several candidate models (Table 1) to investigate our a priori hypotheses regarding D. rhodogaster occupancy. For all models excluding the null model we included daily maximum temperature and day of year as detection covariates. Our model hypotheses were: (1) site occupancy is constant, (2) occupancy will decrease with greater burn extent and decrease in the years following fire, (3) occupancy will decrease with greater burn extent, (4) occupancy will decrease with greater burn extent and greater burn severity, (5) occupancy will decrease with greater burn severity and in the years following fire and (6) occupancy will decrease with greater burn extent and differ across parks. We constructed the final global model with burn extent and severity as well as year and park predicting occupancy probability and daily maximum temperature and day of year predicting detection probability.
Candidate model structures and model selection to investigate hypotheses regarding the detection p(.) and occupancy ψ(.) of the mustard-bellied snake Drysdalia rhodogaster in south-eastern Australia (Fig. 1). Models are ranked by quasi-likelihood Akaike information criterion corrected for small sample size (QAICc), which was used because of the small sample sizes and model variances being inflated by a global modal ĉ value of 2.14. Models used in model averaging are marked with an asterisk (*).
1 day, day of year (difference between the survey date and the start of the austral spring on 1 September); ext, fire extent; park, location (national park in which the survey took place); sev, fire severity; temp, air temperature; year, survey year
We confirmed model fit by performing a Mackenzie–Bailey goodness-of-fit test using the function mb.gof.test in the R package AICcmodavg (Mazerolle, Reference Mazerolle2023) on the global model. We ran goodness-of-fit tests with 10,000 iterations. χ 2 test results returned non-significant values (χ 2 = 1,423.24, P = 0.07), indicating that our global model fit the data; however, we found the models to have mild overdispersion (ĉ = 2.14). To account for this overdispersion, we inflated the variances of the candidate model covariates by the value of ĉ prior to model selection. We assessed model suitability using the quasi-likelihood Akaike information criterion corrected for small sample size (QAICc) and considered model structures suitable if they had the lowest QAICc value or ΔQAICc < 2 compared to the leading model. To provide inferences on covariate impacts on detection and occupancy, we used model averaging with shrinkage on the supported models to generate beta estimates and confidence intervals using the modavgShrink function of AICcmodavg with an adjusted ĉ (Mazerolle, Reference Mazerolle2023). We generated site occupancy and detection estimates for the most parsimonious model using the predict function in R.
Results
Habitat suitability models
Our model predicted the range of D. rhodogaster to be considerably larger than current presence records indicate. We predict that the range of D. rhodogaster extends north of Hunter Valley along the Great Dividing Range towards the border with Queensland (Fig. 2b). North of 34 °S the predicted distribution of D. rhodogaster is restricted to areas > 250 m elevation (Fig. 3c). When intersected with fire extent mapping, c. 46% of the predicted range of D. rhodogaster was burnt during the Black Summer.
(a) Occurrence records used for the habitat suitability model of D. rhodogaster across the eastern seaboard of Australia. Approximate locations of reliable D. rhodogaster records not recorded in the Atlas of Living Australia (2021) near the towns of Gloucester and Tenterfield are indicated. (b) Probability of suitable habitat generated by the habitat suitability model: values closer to 0 represent a low suitability and values closer to 1 represent high suitability. The location of Hunter Valley is indicated. (c) Plot of latitude vs elevation (m) of the locations for which the habitat suitability model for D. rhodogaster predicted a probability of suitable habitat > 0.5.
Records of D. rhodogaster used to assess occurrence across the species distribution in south-eastern Australia in the years following the Black Summer bushfires. We obtained locality records from the Atlas of Living Australia (extracted 25 March 2022) and field surveys reported in this study. The shaded area represents areas burnt during the Black Summer bushfires.
Field detections and occupancy models
Across the three parks we surveyed 61 sites, representing a total effort of 542 person-hours. Of the 61 sites, we classified 24 as burnt at high severity, 19 as burnt at low severity and 18 as unburnt. We recorded 41 detections of D. rhodogaster, with 20 individuals recorded under tin sheets, seven under roof tiles and 14 found during active searches.
We detected D. rhodogaster at 16 sites: 11 observations in five sites classified as unburnt, 10 snakes in three sites burnt at low severity and 20 snakes in eight sites burnt at high severity. In the first 12 months after the Black Summer bushfires, 80% of the snakes detected were in burnt areas.
The most parsimonious occupancy model was the null model. However, two additional models were supported. One included detection covariates of daily maximum temperature and day of year and the other contained these detection covariates and also fire extent (Table 1). Park and year were poorly supported (Table 1). Model averaging of the top models found no support for temperature (β = −0.29, 95% CI: −1.14, 0.55) or day of year (β = −0.1, 95% CI: −0.39, 0.59) influencing detection or of fire extent influencing occupancy (β = −0.11, 95% CI: −0.66, 0.45). When only considering the detection covariate model, detection decreased with higher daily maximum temperature (β = −0.64, 95% CI: −1.21, −0.07), and there was no effect of day of year (β = 0.19, 95% CI: −0.26, 0.65). Similarly, the model including fire extent found decreased detection with higher maximum temperature (β = −0.66, 95% CI: −1.22, −0.10) as well as no effects of day of year (β = 0.23, 95% CI: −0.21, 0.69) or fire extent (β = −0.53, 95% CI: −1.07, 0.01). However, we refrain from making inferences regarding these findings based on the low weight of this model relative to the null. Hereafter, we only present data from the most parsimonious (null) model. D. rhodogaster showed a low mean site occupancy (0.3 ± SD 0) as well as low detectability (0.2 ± SD 0).
Post-fire presence records
Excluding snakes observed by our survey team, a total of 38 records of D. rhodogaster were recorded in the Atlas of Living Australia in the 2 years after the Black Summer bushfires. Of these, 28 records were from unburnt areas and 10 records were from burnt areas (Fig. 3). Notably, in the first year after the fires all five snakes recorded in the Atlas of Living Australia were detected in areas that had been burnt.
Discussion
Our results show that c. 46% of the predicted distribution of D. rhodogaster was burnt during the Black Summer bushfires. However, our field surveys suggest that fire severity and burn extent at the site level probably had negligible impacts on the occupancy of D. rhodogaster. Similarly, occurrence records reported in public databases show that D. rhodogaster was recorded in areas that had been burnt during the Black Summer bushfires, with many of these snakes being observed during the first 12 months following the fires. Collectively, the results of our surveys and citizen science records indicate that D. rhodogaster has continued to occur in areas burnt by the Black Summer fires, suggesting that the fires had limited effects on the distribution and occupancy of D. rhodogaster within forest habitats.
During and immediately following the Black Summer there was significant concern regarding the effects of the fires on wildlife populations. Consequently, understanding the environmental conditions and life histories predisposing taxa to declines or persistence after severe fires is currently a strong focus of research in Australia (Ensbey et al., Reference Ensbey, Legge, Jolly, Garnett, Gallagher and Lintermans2023). Several studies have found that rainforest species appear to have been adversely affected by the Black Summer fires (Law et al., Reference Law, Madani, Lloyd, Gonsalves, Hall and Sujaraj2022a; Beranek et al., Reference Beranek, Hamer, Mahony, Stauber, Ryan and Gould2023), yet for grassland and dry forest species evidence of adverse effects is mixed (Webb et al., Reference Webb, Jolly, Hinds, Adams, Cuartas-Villa, Lapwong and Letnic2021; Hartley et al., Reference Hartley, Blanchard, Clemann, Schroder, Schulz, Lindenmayer and Scheele2023). However, for some forest species fire severity appears to be an important factor, with the effects of fire being greatest in areas burnt at high severity (Law et al., Reference Law, Madani, Lloyd, Gonsalves, Hall and Sujaraj2022a,Reference Law, Gonsalves, Burgar, Brassil, Kerr and O'Loughlinb; Letnic et al., Reference Letnic, Roberts, Hodgson, Ross, Cuartas and Lapwong2023). Our occupancy models support emerging insights suggesting that the effects of the Black Summer fires have varied markedly between species, as fire severity and burn extent had little effect on the occupancy of forest habitat by D. rhodogaster.
The occurrence of D. rhodogaster across the landscape post-fire is not unexpected given that the long and narrow bodies of these snakes are well suited to seeking thermally buffered refugia to avoid mortality during wildfires (Pausas, Reference Pausas2019). Soil is an effective buffer against lethal temperatures, with depths as shallow as 6 cm being sufficient to reduce temperatures to c. 30 °C during surface fires (Bradstock & Auld, Reference Bradstock and Auld1995), which is significantly below lethal temperatures for several closely related elapid snakes (Heatwole & Taylor, Reference Heatwole and Taylor1987). Given that D. rhodogaster is a terrestrial snake that often shelters under debris, it is plausible that the animals may have persisted in situ in buffered microsites rather than recolonizing from adjacent unburnt areas (Pausas, Reference Pausas2019). Post-fire composition of reptile communities has previously been found to be better explained by in situ persistence than by recolonization for adjacent unburnt regions (Santos et al., Reference Santos, Belliure, Gonçalves and Pausas2022).
Similarly, the post-fire occurrence of D. rhodogaster is likely to have been assisted by the low metabolic rates and energy demands that D. rhodogaster shares with other reptiles (Else & Hulbert, Reference Else and Hulbert1981). These low energetic demands are complemented by the availability of the small skinks on which the snakes prey (Shine, Reference Shine1981), which are often abundant in post-fire environments (Lunney et al., Reference Lunney, Eby and O'Connell1991). Although we did not quantify prey availability, we frequently observed small skinks (e.g. Lampropholis spp. and Saproscincus mustelinus) at burnt sites, suggesting that food was amply available for snakes after the fires. Understanding the role of prey species in driving the occurrence of snakes post-fire could be an important focus of future research.
Our analysis of post-fire records showed c. threefold more D. rhodogaster observations in areas that were unburnt (Fig. 3). However, the presence of records in burnt areas during the 12 months immediately following the fires, including in areas burnt at high severity, suggests that these snakes survived the fires. The number of detections must be interpreted cautiously, as records from the databases used were collected in a non-systematic manner and are thus open to sampling bias. For example, after the Black Summer bushfires, many reserves were closed to the public for varying periods of time because of safety concerns and therefore would not have been accessible to citizen scientists. Covid-19 probably also decreased the input of records because of restrictions on the movements of citizen scientists (Stenhouse et al., Reference Stenhouse, Perry, Grützner, Rismiller, Koh and Lewis2022). Therefore, although most records of D. rhodogaster in the 2 years after the fires were from unburnt areas, it is important to note that records from burnt areas may have been under-reported because citizen scientists had less access to these areas.
Our habitat suitability model (Fig. 2b) predicted a much broader potential distribution for D. rhodogaster than is evident from previous occurrence records (Fig. 2a). Our habitat suitability model predicted that the range of D. rhodogaster extends north of Hunter Valley at elevations > 250 m along the Great Dividing Range towards the border with Queensland (Fig. 2b,c). It is possible that our model overestimates the distribution of D. rhodogaster because it extends the range of the species into areas where there are no records in the Atlas of Living Australia. However, the reliability of our model is strengthened by published records of D. rhodogaster near Gloucester and Tenterfield in northern New South Wales (Fig. 2a), as well as by a specimen collected from the Tenterfield region in November 2020 (Australian Museum, R.188326; Goldingay et al., Reference Goldingay, Daly and Lemckert1996; Daly & Lemckert, Reference Daly and Lemckert2011), none of which are reported in the Atlas of Living Australia or included in our model. These populations north of Hunter Valley are probably genetically distinct as the valley is a dispersal barrier for many woodland reptiles (Chapple et al., Reference Chapple, Hoskin, Chapple and Thompson2011). Moreover, they may be heavily fragmented or patchily distributed. Further studies are warranted to confirm the relationship between populations separated by the valley and to confirm whether northern populations warrant additional conservation measures. This is particularly important as much of the predicted range of D. rhodogaster north of Hunter Valley was burnt in the Black Summer bushfires and because both published records from the region are from land used for recent or ongoing native forestry (Goldingay et al., Reference Goldingay, Daly and Lemckert1996; Daly & Lemckert, Reference Daly and Lemckert2011).
Given that the reptile fauna of north-eastern New South Wales has been extensively surveyed (e.g. Milledge Reference Milledge, Lunney and Ayers1993; Daly et al., Reference Daly and Lemckert2011) and there are few reliable records of D. rhodogaster from the region, it is possible that biotic factors such as habitat type or competition may be restricting the species. However, because the literature records suggest that D. rhodogaster does occur in north-eastern New South Wales (Goldingay et al., Reference Goldingay, Daly and Lemckert1996; Daly et al., Reference Daly and Lemckert2011) and our suitability model suggests that the species will be geographically restricted to elevated areas with a cool climate, we believe it is more likely that survey efforts using methods appropriate to detect this cryptic species have been insufficient in climatically suitable habitats. Consequently, we recommend that targeted searches for D. rhodogaster are undertaken in areas of suitable habitat (e.g. high-altitude forest) in this region to better determine its status and how it is affected by fire.
Overall, the data on D. rhodogaster occupancy in post-fire environments that we have collated from our own surveys and the Atlas of Living Australia suggest that even though a considerable portion of the known and predicted range of D. rhodogaster was burnt in the Black Summer bushfires, the species persists in areas affected by fires. Like many other species, D. rhodogaster was of low conservation concern prior to the fires, and little was known about its ecology or population status. Although our research provides some insight into the occurrence of D. rhodogaster after an extensive bushfire, the strength of our conclusions is limited by the absence of pre-fire information for this cryptic species. Moving forward, there is a need for a concerted effort to build baseline data on cryptic species categorized as Least Concern, so that more comprehensive comparisons can be made in the wake of future catastrophic events.
Author contributions
Study conception: ML, with contributions from MJH; data collection: MJH, AKR, YL, SC, BR, OP, JW, NS, JL, SWL, ML; data analysis: MJH, SWL, HMB, FM, ML; writing: MJH, with contributions from HMB; revision: all authors; funding acquisition: JW, ML.
Acknowledgements
We thank the many volunteers who assisted with the fieldwork; Tony Thorne for donating the roof tiles used in the field surveys; Marc Mazerolle for providing statistical advice; and two anonymous reviewers for their comments. This project was funded by the Bushfire Recovery Fund (BWHR-T2_GA-2000634). The University of Sydney has a publishing agreement with Cambridge University Press that facilitated the open access publication of this article.
Ethical standards
This research abided by the Oryx guidelines on ethical standards. All fieldwork methods and animal handling were approved by the University of New South Wales Animal Ethics Committee (20/160B) and the University of Wollongong Animal Ethics Committee (AE1912). The field research was conducted with approval from New South Wales National Parks and Wildlife Service (SL102394).
Conflicts of interest
None.
Data availability
Occupancy data are available at doi.org/10.6084/m9.figshare.24804003. Pipelines for habitat suitability models are available at github.com/HMB3/habitatIntersect.
Open access
