Severe wildfires promoted by climate change negatively impact forest amphibian metacommunities

Changes to the extent and severity of wildfires driven by anthropogenic climate change are predicted to have compounding negative consequences for ecological communities. While there is evidence that severe weather events like drought impact amphibian communities, the effects of wildfire on such communities are not well understood. The impact of wildfire on amphibian communities and species is likely to vary, owing to the diversity of their life‐history traits. However, no previous research has identified commonalities among the amphibians at most risk from wildfire, limiting conservation initiatives in the aftermath of severe wildfire. We aimed to investigate the impacts of the unprecedented 2019–2020 black summer bushfires on Australian forest amphibian communities.


| INTRODUC TI ON
Anthropogenic climate change alters the extent, severity and affected habitat of wildfires, potentially influencing the evolutionary trajectory of species and their persistence (Flannigan et al., 2009;Williams et al., 2019). Such changes are expected to have increasingly negative consequences for entire ecological communities, including those both with and without a long history of fire exposure (Nolan et al., 2020). Recent wildfire events in Africa, Greece, California and Australia (Nolan et al., 2020), as well as across the Amazon (Chakraborty et al., 2019), already demonstrate such consequences. Understanding the impacts of climate-driven wildfires on species across ecological communities is critical to predicting their viability and identifying opportunities for adaptive management to improve their resilience (Folke et al., 2004).
The characteristics of the wildfire, including its extent and severity, influence the response exhibited by animal communities . The effects of wildfires can be both direct and indirect, operating over short-term, moderate-term and evolutionary timescales (Engstrom, 2010). Direct or 'first order' effects include death from burns, smoke inhalation and oxygen deprivation (Tomas et al., 2021), with mortality increasing with fire severity (Jolly et al., 2022). Indirect or 'second order' effects include starvation, increased predation and habitat modifications that may affect breeding, movement and population viability (Davis et al., 2019;Hradsky et al., 2017). Exposure to periodic, low-severity fires over long evolutionary periods allows organisms to develop adaptations to fire that are behavioural (Pausas & Parr, 2018), biological and morphological (Keeley et al., 2011;Mahony, Gould, et al., 2022;Mahony, Hines, et al., 2022). Yet, even fire-tolerant species may be overwhelmed by fires that have undergone rapid shifts in schedules, intensity and extent because of climate change. Thus, the persistence of ecological communities in the presence of climate-changed induced wildfires will be dictated by the characteristics of future fires, as well as the adaptative capacity of its species members Lindenmayer et al., 2013).
While there has been research into the responses of amphibians to relatively small fires, including ecological burns and small-scale wildfires, there has been limited research on the effect of largescale severe wildfires due to their rarity (Hossack & Pilliod, 2011;Mahony, Gould, et al., 2022). Previous investigations have been concentrated around fires over small geographic scales and are mostly species-specific, demonstrating that such disturbances can cause population reductions (Hossack & Pilliod, 2011), local extinctions (Lemckert, 2000) and reduced genetic diversity (Potvin et al., 2017).
Amphibian communities impacted by wildfires primarily exhibit a reduction in species richness due to the local extinction of firevulnerable species (Muñoz et al., 2019). There is an urgent need to quantify how amphibian communities respond to high-severity large-scale wildfires and to identify species that lack resilience so that management resources can be more effectively allocated.
The impact of fire varies between amphibian species and is dependent on habitat associations and the presence/absence of specific traits. Rain forest obligates are less adapted to fire compared with species occupying dry forest, as fire is not a common feature of rain forests and thus unlikely to have been a driving factor in their evolutionary histories (Mahony, Hines, et al., 2022). This is apparent by the lower upper thermal limits of Australian frogs from cool temperate rain forests (between 28 and 29°C) compared with those from dry forests (up to 36°C;Brattstrom, 1970). Despite these physiological differences in species between communities, some amphibians have behavioural and morphological traits to survive fire (Pilliod et al., 2003;van Mantgem et al., 2015), even if not directly selected for this purpose. For example, most amphibians actively seek out moist refuges to prevent desiccation, which is a behaviour that may act as a preadaptation that improves their probability of surviving fire . Amphibians that naturally burrow to avoid hot dry conditions are also buffered from the rising temperature during the passing of a fire (Penman et al., 2006). Species with adaptations to prevent water loss, such as protective skin secretions (Withers et al., 1984) may fare better when exposed to dry conditions that can lead to fire. Species from the same community can differ enormously in these key traits despite sharing the same evolutionary history with fire, highlighting the need for speciesspecific assessments.
Changes to the availability of shelter and breeding habitat that can occur before or after a fire can also influence amphibian populations (Hossack et al., 2013). Wildfire can reduce the number of refuges, which may cause lagged declines (Hossack et al., 2013).
Similarly, droughts usually precede wildfire events and can reduce the availability of suitable breeding habitat and refuges which can lead to reductions in amphibian populations and communities before the passing of a wildfire Cayuela et al., 2016;Wassens et al., 2013).
We evaluated the impact of the unprecedented Black Summer Fires of 2019-2020 on forest amphibian metacommunities across the east coast of New South Wales, Australia. This fire event produced the largest extent of high-severity wildfire ever recorded raises serious concern for the persistence of amphibians under an increasingly fireprone climate.

K E Y W O R D S
2019-2020 Australian Black Summer fires, community ecology, disturbance, fire severity, multispecies occupancy model, severe weather, threatened species in Australia  and was preceded by a recordbreaking drought causing very low fuel moisture content (Nolan et al., 2020). The current cumulative evidence strongly supports the idea that these events were driven by climate change (Abram et al., 2021;Canadell et al., 2021;Collins et al., 2022). This wildfire burnt rain forest ecosystems with unprecedented scale and severity (Godfree et al., 2021), decimating the landscape, causing some species to have large portions of their distributions affected (Ward et al., 2020). Our knowledge of the impacts of these fires on amphibians is currently limited to a short-term investigation of species' persistence using presence-only citizen science data (Rowley et al., 2020) and expert elicitation analyses Mahony, Gould, et al., 2022). A formal analysis of the impacts of the fire using standardized and replicated field survey data has not yet been reported. This form of analysis is critical to precisely determine which communities and species were impacted by the wildfires.
We investigated the impacts of severe fire on two scales: the amphibian metacommunity and the individual species level, using multispecies occupancy models (MSOM). We analysed occurrence data of temperate forest amphibians from a large spatial area while accounting for imperfect detection. As the Black Summer fire event was unprecedented in severity and scale (Collins et al., 2021), we predicted there would be negative responses of metacommunity occupancy and species richness relative to the proportion of habitat severely burnt within the study area. We aimed to (1) highlight species that were fire-affected to address their conservation status and (2) determine which shared traits among species made them more susceptible to fire. We predicted that rain forest specialists would be vulnerable to wildfires as they do not have an evolutionary history with this type of disturbance and that burrowing amphibians would be buffered from the effects of wildfire as we expected that a fossorial refuge would enable such species to avoid first-order fire impacts. We expected arboreal species to be negatively impacted since they seek refuge in arboreal situations which would be exposed to adverse conditions during severe wildfire where full canopy consumption occurs.

| Study area & site selection
This study occurred along the Great Dividing Range in New South Wales, which is one of the largest continuous native forests within Australia (about ~815 km from the most northern to the most southern study sites) and was heavily impacted by severe fires in 2019-2020 ( Figure 1). Different forest community types were surveyed for threatened amphibians: cold-temperate and subtropical rain forest communities occurring in the northern areas of the study region.
These forests typically have an entirely closed canopy with a dense midstory and understory (Specht, 1983). Wet and dry sclerophyll forests were surveyed across the entire region. The canopy of these communities is dominated by Eucalyptus spp. and is variable in canopy coverage (30%-70%, Keith Classification).
Survey sites were chosen based on the occurrence of 10 threatened species that had distributional overlap with the wildfires and for which there was long-term monitoring information available (Heleioporus australiacus, Litoria daviesae, Lit. littlejohni, Lit. subglandulosa, Lit. watsoni, Mixophyes balbus, M. iteratus, Philoria pughi, P. sphagnicola and Pseudophryne australis). We assumed that our targeted threatened species surveys would also allow us to generate valid data for common, co-occurring species-enabling the evaluation of community responses to wildfire. Where possible, we chose equal numbers of sites in burnt and unburnt areas for each target threatened species. Two criteria were used for site selection (1) where target species were known from prefires and (2) sites which overlapped with the Fire Extent and Severity Mapping (FESM; Gibson et al., 2020). Due to differences in fire severity and amphibian community distribution, we divided the large geographic extent in two regions either side of the Hunter Valley (~32° S). This resulted in a total of 411 sites, with 293 sites in the northern region (n burnt = 188 and n unburnt = 105) and 118 sites in the southern region (n burnt = 67 and n unburnt = 51).

| Amphibian surveys
Surveys were conducted during the breeding season for each species after the fires (in some cases within weeks and others about 6 months after the fires) and continued for about 1 year (from February 2020 to July 2021). There were some instances where acoustic devices were deployed before the fires and were retrieved after the fires, allowing some direct postfire data to be included in the analyses (from January 2020). Two approaches were used to detect amphibians: visual encounter surveys (VES) and passive audio recorders. Recorders were deployed in the field for 3-9 months and needed 2-3 battery and data storage changes. Multispecies data were extracted from the audio files by researchers experienced in frog call identification. Due to the large quantity of data to sort through, we focussed efforts on the single 5-min records from 22:00 as most species appear to be active at this time. We increased the listening period to include 08:00 and 17:00 at sites targeting Philoria spp., as this group is known to peak in calling activity during these periods (Willacy et al., 2015). We were not able to reliably separate species of the Pseudophryne genus based on calls. Hence, we lumped three species together (P. australis, P. bibronii and P. coriacea) into 'Pseudophryne spp.' in the southern region where they are sympatric.
This was also true for U. laevigata and U. fusca, which were grouped together into 'Uperoleia spp.'.

| Environmental covariates of detection probability
Several environmental variables were recorded and collated to inform analysis of detection probability. Air temperature was recorded at the beginning of each VES survey with a Kestrel weather meter (model: 3000 RH). Hourly temperature data were associated with each acoustic recording using the R package ChillR (Luedeling, 2019), and this was also used in situations where there was missing temperature data (see Figure S1 for a comparison between the two temperature measures). Daily rainfall was also collected using ChillR.
Julian date was paired with each survey and audio recording as an integer to account for seasonal variation of detection, where the first day of September = 1 and the last day of August = 365.

| Spatial and habitat data collection for occupancy
For each survey site, fire severity was extracted from the Fire Extent and Severity Map (FESM; pixel size: 10 × 10 m; State Government of NSW and Department of Planning and Environment, 2020). A 500-m buffer was extended either side of the transects (ArcMap 10.7.1) to quantify the proportion of total area impacted by severe fire within the buffer zone. The 500-m buffer also corresponds to the upper limit of the dispersal distance of most amphibians (Lemckert, 2004) and provides a useful landscape measure of fire impacts to local frog communities. Only FESM categories 4 (high severity) and 5 (extreme F I G U R E 1 Map of study sites in reference to the 2019-2020 fire extent and severity map. Fire severity categories; 0 = unburnt, 2 = burnt surface with unburnt canopy, 3 = partial canopy scorch, 4 = full canopy scorch with partial canopy consumption and 5 = full canopy consumption. The dashed line indicates where the geographic split between the two regions occurs. severity) were used to calculate the proportions of area impacted by fire.
An additional seven covariates were included to help explain variation in the probability of occupancy. Six were extracted using ArcMap and included latitude, elevation above-sea-level (pixel size: 10 × 10 m), slope (pixel size: 28 × 28 m), aspect (pixel size: 28 × 28 m), root-zone soil moisture (soil moisture in the top 100 cm of the surface, pixel size: 5 × 5 km) and Normalised Difference Vegetation Index (NDVI, pixel size: 5 × 5 km). A mean value was extracted for a 10-m buffer area extending either side of the transect line using ArcMap. We selected this buffer distance as this was the extent of the transect coverage during surveys.
Elevation, aspect and slope data were extracted from a digital elevation model grid of Australia (Gallant et al., 2011) and monthly average of NDVI and root-zone soil moisture from the Bureau of Meteorology (2020). We included these covariates as they are important drivers of amphibian occupancy Seaborn et al., 2021). As hydroperiod can be a strong determinant of amphibian community structure in waterbodies (Snodgrass et al., 2000), we included ephemerality as a categorical covariate (0 = permanent; 1 = ephemeral), with any water body that dried during field surveys being classed as ephemeral. Ephemerality is also likely to be affected by climate change-induced drought, and thus, it was essential to incorporate the influence of the previous severe drought prior to the survey period (Hossack & Corn, 2007;Hossack et al., 2013).
Root-zone soil moisture was used as a proxy variable to measure drought impacts (Bureau of Meteorology, 2020). We chose this metric as (1)

| Statistics
A multispecies occupancy model (MSOM) was used to assess the relationships between fire severity and forest amphibian metacommunities using Bayesian inference. Under a hierarchical framework, a MSOM accounts for the imperfect detection of species during a survey while linking occupancy models for individual species. In doing so, MSOMs provide estimates of both metacommunity-level and species-level responses to environmental covariates (Devarajan et al., 2020). This unified approach increases the precision of parameter estimates for infrequently observed species, thereby enabling estimation of occupancy rates for both rare and common species (Dorazio et al., 2006;Kéry & Royle, 2008). Data from frequently observed species are borrowed from data-poor species, which results in the mean estimates of rare species being drawn towards the group average while increasing the precision of estimates (Link, 1999).
The structure of the MSOM was based on the methods of Dorazio et al. (2006). The occupancy submodel assumed a true but only partially observed presence-absence matrix z i,j for species i = 1, 2,…, N at site j = 1, 2,…, J, where z i,j = 1 if species i was present at site j, and z i,j = 0 if the species was absent, and where z i,j ~ Bernoulli(psi i,j ) and psi i,j is the probability that species i occurs at site j. Given that the state variable z i,j is typically uncertain, observed data x i,j,k are used for species i at site j during survey k, which are also assumed to be Bernoulli random variables if species i is present (Dorazio et al., 2006). A detection submodel specified that where z i,j is the true occurrence matrix and p i,j,k is the probability of detection of species i at site j during survey k. Detection was fixed to zero when species i was not present because z i,j = 0.
Relationships between the probability of occupancy and the fire severity within a 500-m radius, along with seven additional environmental covariates, were assessed in two models. The first model assessed the metacommunities of frog species inhabiting a broad range starting from south of the Hunter River to Mongo National Park (southern model), while the second model assessed metacommunities north of the Hunter River to the Queensland border (northern model). The Hunter River was chosen to delineate the two models given that it forms a biogeographical barrier along the Great Dividing Range and represents the turn-over point of the distributions of many frog species along the east coast of Australia (Mahony et al., 2001). The following covariates for the probability of occupancy at a site were modelled: (1) fire severity within a 500-m radius; (2) latitude; (3) elevation; (4) slope; (5) aspect; (6) root soil moisture; (7) NDVI; and (8) waterbody ephemerality.
Intercorrelations among the covariates for both models were assessed; however, no strong correlations were evident (r < .5; see Table S2). A random effects term (cluster) was included in both models as a categorical variable to account for potential spatial autocorrelation in site occupancy, as some waterbodies were within dispersal distance of neighbouring sites. We chose to incorporate random effects since spatial autocorrelation can result in overestimation of parameter estimates and bias occupancy rates (Wintle & Bardos, 2006). Sites within the southern and northern models were assigned into 110 and 95 spatial clusters, respectively. Clustering was applied to sites that were within 500 m of each other to prevent spatial pseudo-replication. A site-level random effects term was also included in both models to account for unexplained variation in site occupancy, given the wide geographical distribution of the study sites. For instance, we included only one local-scale habitat covariate (ephemerality), whereas other unmeasured local variables may also influence frog occupancy-including waterbody vegetation cover and habitat structure.
The detection submodel included four survey-level covariates: (1) Julian date to account for seasonality in detection among species; (2) a quadratic effect of Julian date because detection is likely to peak for many species during the breeding season; (3) survey type modelled as a categorical covariate (VES = 0; acoustic = 1); (4) air temperature; (5) rainfall; and (6) a quadratic effect of rainfall, since detection may be lower with low rainfall and with heavy rain, reducing the ability to detect frog calls with acoustic surveys due to background noise.
Missing survey-specific covariates for detection were replaced by mean values in instances when surveys at a site were not conducted.
In the model, species-specific parameters were linked to the wider metacommunity by estimating hyper-parameters (μ) that treated species-level parameters as random effects drawn from a normal distribution; that is, α i ~ N(μ α , sigma α ) where α i is the probability of occupancy for species i, μ α is the mean community response and sigma α is the standard deviation in α among species (Zipkin et al., 2009) covariates was determined by 95% CIs that did not overlap with zero, although a slight overlap of the CI with zero was still assumed to confer statistical importance (Cumming & Finch, 2005). Bayesian p-values for each model were computed to assess model fit using the Freeman-Tukey fit statistic; p-values close to .5 indicated good model fit (Gelman et al., 1996).
Detection-error corrected estimates of the number of species at each site obtained from the southern and northern multispecies occupancy models were used to evaluate relationships between species richness and fire severity within a 500-m radius. We used Poisson regression modelling with uninformative priors for the in- Both models were run using three MCMC iterations to generate 50,000 samples after discarding a burn-in of 10,000 samples with chains thinned by five. The estimated mean, standard deviation and 2.5th and 97.5th percentiles of the posterior distribution of the coefficients were obtained from each model.

| RE SULTS
A total of 40 species were detected during surveys across the study region (mean 2.27 ± 0.11 per site). Each of the species was from one of three anuran families, either: Limnodynastidae, Myobatrachidae or Pelodryadidae. Thirty-one species were detected in the south-  (Table S3). Mean estimated probabilities of detection were also highly variable among the frog species from each region, ranging from 0.008 (0.002-0.028) for Lit. peronii in the southern model to 0.574 (0.498-0.649) for C. signifera in the southern region (Tables S3 and S4). The models for both regions had good fit to the data with Bayesian p-values of .48 and .52, respectively.

| Summary of detection and occupancy probability covariate effects
There were negative relationships between the mean probability of detection and Julian date, Julian date 2 , survey type and rain 2 in both southern and northern models (Tables S5 and S6). There was a strong relationship between the probability of detection and temperature in the southern model, but this relationship was weaker in the northern model. In the southern model, there were negative relationships between the mean probability of community occupancy and the extent of severe fire, site latitude and elevation, along with a positive relationship with NDVI. In the northern model, there was a negative relationship between mean community occupancy and severe fire extent and a positive relationship with site latitude (Tables S5 and S6). There was a negative relationship between the probability of community occupancy and site ephemerality in the southern model.

| Metacommunity and species richness responses to fire
There was a strong negative relationship between the extent of severe fire and community occupancy in the southern region (mean: −0.52; −1.08 to −0.034, Figure 2), but this relationship was unclear in the northern region (mean: −0.24; −0.84 to 0.34, Figure 2). The extent of high-severity fire was the fourth strongest predictor variable for probability of occupancy in the southern region (Table S4) Table S6), whereas there was less variability in species responses to fire cover in the southern region (sigma α = 0.724; Table S5).

| DISCUSS ION
We demonstrate that the unprecedented climate change-driven fires of 2019-2020 resulted in an overall loss in site-level amphibian biodiversity. In particular, there was a reduction in both species' richness and metacommunity occupancy in the southern region of the study, but this effect was less clear in the northern region. This is the first investigation using unbiased estimates derived from a multispecies occupancy model to assess the response of an amphibian metacommunity to wildfire. Some species were impacted by the fires more severely than others, most noticeably rain forest specialists (A. darlingtoni, M. iteratus and P. pughi). Negative fire-related impacts were not just limited to rain forest-dwelling amphibians but also to species that have evolved and adapted to fire-prone vegetation communities Mahony, Hines, et al., 2022), highlighting the scale and severity of the fires. We discuss the impacts of wildfires on two ecological scales: the metacommunity scale and the species scale. We focus on explaining potential reasons behind the differential impact of the fires between species and offer recommendations for future research and conservation management to address the global threat of climate change-driven wildfires on amphibians.

| Metacommunity scale impacts of severe fires
In the northern region of our study, extensive severe fires did not occur often, resulting in limited statistical power to determine the impact of severe fire extent on amphibian communities in this region. Although the northern model did not reveal a correlation between metacommunity occupancy and severe fire extent, several species demonstrated a strong negative relationship to severe fire extent, indicating the complexity of species responses to fire and the need for species-level investigations to ensure vulnerable species are not overlooked based on average metacommunity responses.
Our findings demonstrate that large-scale severe wildfires can negatively affect the persistence of numerous species and amphibian communities. The extent and severity of wildfire has been shown F I G U R E 2 Regional metacommunity response to the proportion of habitat burnt by severe wildfire within 500 m for the (a) southern region model and (b) northern region model. Solid lines are the mean occupancy estimates, while shaded areas represent 95% Bayesian credible intervals. to be an important predictor of persistence in other fauna communities and species (Blakey et al., 2021;Lindenmayer et al., 2013). Numerous studies have reported amphibian persistence after wildfire, but these have been restricted to either relatively low-severity or small-scale fires (<10,000 ha, e.g. Brown et al., 2011;Greenberg & Waldrop, 2008;Hossack & Corn, 2007). As we enter a period where wildfires are predicted to burn with greater extents and severity (Silveira et al., 2022), proactive conservation actions need to be directed to amphibians.

| Impacts across amphibian species
Many amphibian species were affected by the wildfires-including a suite of those expected to have some ability to withstand fire.

This includes the burrowing frogs (H. australiacus and Lim. dumerilii),
which are considered to be afforded a degree of protection against wildfire as they can burrow deeply enough to avoid heat exposure Penman et al., 2006). Despite these expectations, the burrowing frogs present in the southern region had negative relationships with severe fire. A potential explanation for the susceptibility of species predicted to have some fire tolerance is the sheer scale and severity of the 2019-2020 wildfires, which may have exceeded the physiological limits of species with fire-resistant adaptations.
We found consistent impacts of fire among closely related taxa, including within the Litoria phyllochroa species group and Philoria spp. The L. phyllochroa species group includes seven species that are all similar in terms of morphology, behaviour and habitat preference; most are green, small bodied, and occupy and breed in F I G U R E 3 Effect sizes of each species occupancy in response to cover of high severity fire within 500 m in both the southern and northern models. Triangle = southern model. Circle = northern model. Red = negative relationship with BCI not overlapping 0, black = BCI overlapping 0, green = positive relationship with BCI not overlapping 0.  Cover of high severity fire within 500m Philoria pughi occupancy (d) riparian habitats (Donnellan et al., 1999). Three of four species in this group were less likely to occur where severe fire extent was larger (L. nudidigitus, L. pearsoniana and L. phyllochroa). Furthermore, we uncovered clear negative impacts of wildfire on P. pughi and P. sphagnicola. Philoria comprise a unique Gondwanan lineage of rain forest-dwelling species that create burrows in rain forest bogs and seepages where nonfeeding tadpoles develop (Mahony, Hines, et al., 2022). They have evolved in fire-sensitive vegetation communities where wildfire has not occurred to any great extent in the past (Cochrane, 2003), which may explain their responses to wildfire.
There are three plausible hypotheses that could explain why small, riparian-dwelling frog species are sensitive to fire: (1) direct mortality due to radiant heat from burning of the vegetation that constitutes the main habitat refuge; (2) indirect effects caused by reduced postfire survival due to enhanced predation; and (3) direct mortality of eggs and larvae leading to reduced recruitment. The former hypothesis is possible for species in this group, as they are commonly found in association with riparian vegetation (Lemckert et al., 2005;Parris, 2001). There may be a high degree of fire-induced mortality if riparian vegetation is burnt during severe fires. The second hypothesis is also possible as these frogs are green in colour and may be more vulnerable to predation postfire as they cannot camouflage against burnt vegetation. While this group can darken their skin pigment to blend in with their surroundings (Anstis, 2017), it is unknown how effective this behaviour is against burnt vegetation or how long colour changes can be sustained. European frogs with similar morphology, colour and habitat use are negatively affected by fire, including Hyla molleri and H. meridionalis (Muñoz et al., 2019).
These findings emphasize that frogs which seek cover in riparian vegetation may be universally sensitive to fire.
It is also plausible that some species may have suffered reduced recruitment due to direct mortality of eggs and tadpoles during the drought and fire season, which led to local extinctions. Some amphibians show a 'boom-bust' ecology, where recruitment is needed each year to sustain the population (Berven, 1990). Therefore, a catastrophic loss of recruitment over both large temporal (an entire breeding season) and spatial (an extensive proportion of their distribution) scales may increase local extinction risk. This is especially the case for species where populations are already fragmented by the effect of chytrid (Scheele et al., 2015). Furthermore, tadpoles are susceptible to water contamination, which can cause direct mortality or hampered growth and survival. This was exhibited by tadpoles of Rana catesbiana, which showed impaired tadpole feeding ability due to damaged mouth parts because of fly ash contamination (Christopher et al., 1996). However, the effects of water contamination on tadpole communities due to wildfire is an understudied topic and is in critical need of attention (but see McDonald et al., 2018). It is likely that the consequence of fire on community and species persistence is a culmination of several of these hypothesized impacts, with the severity or presence of each differing widely between species based on life-history traits.
The impacts of wildfire on Philoria spp. and A. darlingtoni raise concerns for amphibians that are rain forest obligates with embryonic and larval development that rely on micro-refugia away from standing water bodies Mahony, Hines, et al., 2022). In the northern region, the occupancy of three out of nine rain forest obligate species was negatively related to severe fire extent (P. pughi, M. iteratus and A. darlingtoni). Species restricted to rain forests that historically do not burn have not been exposed to this threat during their evolutionary history and are predicted to have limited adaptations for avoiding or tolerating wildfire Mahony, Hines, et al., 2022;Nimmo et al., 2021).
Susceptibility of rain forest fauna to severe wildfire has been demonstrated in other animal groups (Law, Madani, et al., 2022). In the case of amphibians, rain forest species are dependent on moist habitats and lack adaptations to reduce or avoid desiccation and are more likely to have adaptations such as tadpoles in terrestrial nests (Philoria spp.) or parental care of tadpoles (Assa spp.) that result in greater innate vulnerability to fire exposure .
For example, A. darlingtoni relies on moisture from the humid microclimate in rain forest leaf litter and has already been demonstrated to have suffered local extinctions due to wildfire (Lemckert, 2000).
This reliance on moisture raises the possibility that populations of rain forest species may already be impacted by the effect of drought before wildfires occur. If so, the resilience of amphibians to fire regimes must be considered in terms of the cumulative impact of other weather events that are also being altered by climate change.
The consistency of fire impacts on rain forest obligate species with specialized life histories calls for attention for similar amphibians occurring in other countries, as they are likely to be impacted by progressive climate change events. Amphibians which possess direct development and non-free-living tadpoles show high diversity in habitats that can retain high humidity (e.g. rain forest: Gould, Beranek, et al., 2022).
Globally, there is a high rate of endemism of amphibians with terrestrial reproductive modes in the tropics (Lion et al., 2019). Some of these regions have recently experienced unprecedented severe wildfires (e.g. Amazon: Chakraborty et al., 2019) and are likely to experience even more severe events in the future (Silveira et al., 2022). Rain forest obligate amphibians with terrestrial reproductive modes should be prioritized in conservation assessments of tropical regions where drought and wildfire events are predicted to increase in severity.
Our results do not support the prediction that arboreal frogs would be highly impacted by fires Mahony, Hines, et al., 2022). Most arboreal species showed weak relationships with severe fire extent (Lit. caerulea, Lit. chloris, Lit. dentata, Lit. peronii, Lit. quiritatus and Lit. tyleri). These results are in line with some of the citizen science data of Rowley et al. (2020), as Lit. dentata and Lit. chloris were found to be calling immediately after the fire in impacted areas and occurring in equal or greater numbers than before the fires. In contrast, Rowley et al. (2020) found fewer observations of Lit. peronii in areas impacted by fires, which differs from our results which showed this species to have no relationship with extreme fire extent. This discrepancy can be explained by the time span of the two studies, as Rowley et al. (2020) focused on immediate short-term responses 4 months after the fires (summer and autumn), which does not include the breeding season of Lit. peronii (spring). This highlights the importance of sampling amphibian communities for at least a year after fire so that the breeding seasons of all species present in the community are considered.
The resilience of arboreal amphibians found in our analysis suggests that the refugia they use is buffered from the impacts of fire. Currently, there is a paucity of records on the refugia used by tree frogs compared with birds and arboreal mammals (Gibbons & Lindenmayer, 2002). Eucalypt trees have deep hollows that provide valuable refuge habitat for amphibians during drought and wildfire.
There is anecdotal evidence from forestry operations of aggregations of frogs when timber is felled and that smaller bodied frogs (e.g. Lit. dentata) use hollows created by timber boring insects (Mo, 2015).
It is likely these arboreal refuges provide tree frogs with a physiological buffer to extremes of dryness and heat encountered during a drought or the passing of a wildfire (O'Connell & Keppel, 2016).
There is an urgent need to understand the use of refuges and the physiological constraints of amphibians with increased wildfire risk.

| Severe fire impacts both threatened and common frogs
Several threatened species that were already under stress from other threatening processes were impacted by fire, including H. australiacus, Litoria littlejohni, Lit. watsoni, P. pughi and P. sphagnicola.
While wildfire has previously been considered as a threat to all these species in conservation assessments, this has not been previously quantified (except see Heard et al., 2021). We recommend that wildfire and climate change impacts are prioritized in future conservation assessments of these threatened species, especially since they often occur in isolated populations where there may be limited opportunities for postfire recolonization without human assistance Mahony et al., 2021;Stock et al., 2022).
Focus has been placed on understanding and predicting the impacts of fire on threatened species (Penman et al., 2015), while little attention has been paid to species that are thought to be common. However, unprecedented weather anomalies driven by climate change may cause sudden and severe contractions in amphibian abundance, recruitment and distribution (Heard et al., 2021;Hossack et al., 2006;Rochester et al., 2010), even for those species buffered by their large population sizes and/or ranges .
Our analysis shows that amphibian species not currently of conservation concern were negatively impacted by the 2019-2020 fires, including Limnodynastes dumerilii, Lit. nudidigitus and Lit. phyllochroa.
These results highlight the utility of community-level monitoring under a multispecies occupancy model framework, which accounts for common species within conservation assessments.

| Caveats and cautions
While we did control for drought in our models, we are cautious in interpreting our results as strictly revealing the impacts of the wildfires alone, as it is likely that some species were also affected by the long-term preceding drought. An unprecedented drought was a prequel to the fire events in 2019-2020 in eastern Australia and caused reductions in an amphibian population. This was clearly shown for a coastal population of the green and golden bell frog (Litoria aurea) in the southern portion of the study area, which shrunk 10-fold during the 2019-2020 drought before the wildfires occurred . Drought can affect amphibians through mortality of tadpoles due to desiccation of drying water bodies (Beranek et al., 2020;, desiccation of adults and temperatures surpassing physiological thresholds Cayuela et al., 2016). These observations underscore the importance of long-term monitoring programmes to account for the cumulative impact of several weather events, especially as we enter a time where more extreme weather anomalies may occur in close succession or even simultaneously . It would be possible to unravel the influence of drought and fire on amphibian communities if long-term monitoring data were available for extended periods of time before, during and after these events occurred.
Another caveat of this study is that the field surveys targeted only threatened species, which subsequently had equal numbers of burnt and unburnt sites for comparison. In contrast, nonthreatened species were not targets for these surveys, and hence, some may have had unequal sample coverages and low sample sizes. We consider that the estimates of fire impact obtained for most of the nontarget species are reliable given the large survey effort and the use of covariates to account for other sources of variation that could explain occupancy. There are some species that were detected infrequently, and hence, the models of these may not meaningfully predict their responses to fire (e.g. Adelotus brevis, Lit. dentata and Lit. revelata). It is likely that the few detections in these common species are due to a lack of sampling in habitats they prefer during targeted surveys for threatened species. For example, Lit. dentata breeds in ephemeral forest ponds in the northern region, a habitat not included in our targeted surveys in this area. To further understand the impacts of fire on co-occurring nontarget species, we recommend that future studies take a community-based approach into consideration for monitoring programmes.
There are difficulties in interpreting results for species that were only detected via calling and not visually observed. A. darlingtoni, Philoria spp. and H. australiacus were all exclusively detected via calls either in transect surveys or on audio recorders. This raises the question whether there was a reduction of their site occupancy due to local extinction, or alternatively, whether the fires reduced breeding habitat suitability which resulted in a lack of calling activity. The alternative hypothesis remains worthy of attention for conservation managers, as this would lead to reduced breeding output and increased chance of local population declines. Long-term monitoring is needed to elucidate which hypotheses best explain these results for species detected exclusively by their call. Manipulative experiments may be useful to gain knowledge on physiological and ethological changes in these species in a simulated postfire landscape.
Lastly, these results should be considered with a short-term lens that gives insight to first-order fire impacts and short-term postfire indirect impacts as the surveys were conducted within a 1.5-year timeframe after the fires. This constitutes at least one postfire breeding season for each species in the study area. Follow-up studies are required to determine factors influencing long-term persistence and recolonization within the fire-and drought-affected study sites.

| CON CLUS ION
It is of international concern that many Australian forest amphibians, both threatened and those of least conservation concern, were severely affected by the climate change-driven wildfires experienced in 2019-2020, as it suggests similar impacts of fire are being experienced by amphibians in other countries. In recent history, there have already been notable extinctions of Australia amphibians, and indeed extinctions globally, caused by other, human-related threats, such as the introduction of the pathogenic chytrid fungus. It would be unconscionable to lose species to a new threat which we should have the capacity to manage at local and global scales. We underscore the importance of monitoring with survey techniques that can account for amphibian metacommunities combining data from targeted surveys, passive acoustics and citizen science monitoring to gain a complete understanding in future potential severe weather events. It may be possible to create a future where species impacted by unnatural wildfires due to anthropogenic climate change can be protected (Shoo et al., 2011). We encourage research exploring options to bolster climate change adaptations within the paradigms of genetic management (Cummins et al., 2019), cryopreservation (Clulow & Clulow, 2016), reintroduction (Klop-Toker et al., 2021) and habitat restoration . However, these efforts may only offer a temporary solution because without global action to reverse or halt further climate change, wildfires are predicted to become larger and more severe, reducing the distributions of many forest-dependent amphibians.

ACK N O WLE D G E M ENTS
We acknowledge significant in-kind contributions from partner or-

FU N D I N G I N FO R M ATI O N
Funding for fieldwork assessment and analysis was provided by a grant from the bush fire recovery programme for wildlife and habitat, received from the Federal Department of Agriculture, Water, and the Environment (grant number: GA-2000241), and the NSW NPWS Saving Our Species Program.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflicts of interest.

PEER R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/ ddi.13700.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are available on the Dryad data repository: https://doi. org/10.5061/dryad.jm63x sjfd.

Dr Chad T.
Beranek is a postdoctoral researcher and teaching associate at the University of Newcastle, Australia. His passion for promoting and conserving biodiversity is the foundation of his work. His research focusses on developing practical solutions for on-ground conservation and restoration, particularly for fauna such as amphibians. Dr Beranek examines scientific questions with multiple lenses, from the individual with behavioural biology, to the metacommunity ecology scale. He is currently involved in several major projects that are centred on optimizing surveys and examining the impacts of climate change and urbanization.