Slow Life History Leaves Endangered Snake Vulnerable to Illegal Poaching

Global wildlife trade is a multibillion-dollar industry and a signicant driver of vertebrate extinction risk. Yet, few studies have quantied the impact of wild harvesting for the illicit pet trade on populations. Long-lived species, by virtue of their slow life history characteristics, may be unable to sustain even low levels of harvesting. Here, we assessed the impact of illegal poaching on a metapopulation of endangered broad-headed snakes (Hoplocephalus bungaroides) at gated (protected) and ungated (unprotected) populations. Because broad-headed snakes are long-lived, grow slowly and reproduce infrequently, populations are likely vulnerable to increases in adult mortality. Long-term data revealed that annual survival rates of snakes were signicantly lower in the ungated population than the gated population, consistent with the hypothesis of human removal of snakes for the pet trade. Population viability analysis showed that the ungated population has a strongly negative population growth rate and is only prevented from ultimate extinction by dispersal of small numbers of individuals from the gated population. Sensitivity analyses showed that the removal of a small number of adult females was sucient to impose negative population growth and suggests that threatened species with slow life histories are likely to be especially vulnerable to illegal poaching.


Introduction
The global wildlife trade is a multibillion-dollar industry that is a signi cant driver of vertebrate extinction risk 1,2 . Annually, billions of plants and animals are tra cked across the globe to meet the burgeoning demands of consumers 3,4 , and the pro ts reaped from the illegal wildlife trade make it one of the world's leading illegitimate businesses 5,6 . Whilst much of this trade consists of the tra cking of wildlife products (e.g., skins, ivory, horns), another major component comprises the tra c of wild caught animals for the Unfortunately, because we only have a limited understanding of the population dynamics of most Australian reptiles, and the threatening processes that affect them 19,20 , there is often insu cient information to assess the effects of illegal poaching on local populations.
Although populations of some geographically widespread reptile species may be relatively unaffected by moderate levels of wild harvesting (e.g. 21 ), long lived species with slow life history traits are vulnerable to over harvesting 22,23 . Species at a high trophic level, low population density, small geographic range size and/or slow life history are already predisposed to increased risk of extinction 24 , and long-lived species with slow growth rates, infrequent reproduction, and small litter sizes are likely highly susceptible to overharvesting. Unfortunately, we know so little about the threats to most Australian reptiles that it is di cult to ascertain which species may be most vulnerable to the impacts of illegal poaching. One notable exception is the threatened broad-headed snake (Hoplocephalus bungaroides), a species endemic to the Sydney region of south-eastern Australia, which has been extensively studied over the last few decades (e.g. [25][26][27][28][29] ). This spectacularly coloured venomous elapid snake is highly valued by reptile keepers 27 , and the underground trade in broad-headed snakes is thought to have contributed to its decline 30 . Unfortunately, the life history traits mentioned above may make this threatened species particularly susceptible to illegal poaching for the pet trade 27,28 .
In this study, we assessed the impact of illegal poaching on a metapopulation of endangered broadheaded snakes (Hoplocephalus bungaroides) in the southern part of their range. This population represents one of two isolated and deeply divergent clades that compose this threatened snake species 31 . Broad-headed snakes are long-lived and produce few offspring, which develop slowly and mature relatively late in life, potentially making populations incredibly vulnerable to population-level disturbances such as wild re or illegal poaching of mature individuals 27,28,32 . Our study population comprises two sandstone plateaux, one that has locked gates on re trails to restrict access ('gated'), and another that has no locked gates on re trails ('ungated'). Since 2008, study sites on the ungated plateau have been disturbed every year by humans searching for snakes, whereas sites on the gated plateau have been disturbed less frequently 33 . To assess the impact of poaching at ungated sites, we used MARK software to develop models of survival to test the hypothesis that survival rates would be lower at ungated sites than gated sites. Next, we used population viability analysis (PVA) to assess the effect of illegal poaching on the growth rate and trajectory of the ungated population and the better protected gated population. By using sensitivity analyses, we identi ed the parameters that explained most of the variation in population persistence. We predicted that harvesting of mature adults at ungated sites would likely have the greatest impact on modelled population persistence, resulting in a much higher probability of extinction over the coming decades.

Study species
Broad-headed snakes (Hoplocephalus bungaroides) are small (< 90 cm snout-vent length), live-bearing, venomous elapid snakes. They are nocturnal, and juveniles feed mostly on lizards that they ambush beneath rocks. Adults have a broader diet that includes lizards, birds, and small mammals 29,34 . In Morton National Park, the location of an isolated and deeply diverged southern clade of the species 31 , broadheaded snakes grow slowly, reaching maturity at 5-6 years 35 . The species is long lived (up to 28 years), and they have long generation lengths (10.4 years 28 ). Mating occurs between autumn and spring, females ovulate in late spring, offspring are born in March and April, and females rarely reproduce annually 28,29 . During the cooler months (May-October), broad-headed snakes occupy the western edge of exposed sandstone plateaux where they shelter under small, exposed rocks that have exfoliated away from underlying sandstone substratum 29 . During the warmer months (November-April), broad-headed snakes leave the exposed outcrops for the shelter of old growth eucalypt forests where they use tree hollows as refugia 36 . It is during the cooler months, when the majority of snakes in the population shelter beneath small, thin, exfoliating rocks on easily accessible ridgelines, that they are most vulnerable to illegal poaching 27 .

Field sites and methods
We studied a metapopulation of broad-headed snakes located approximately 160 km south of Sydney, New South Wales, Australia. One population was located on a sandstone plateau (roughly 20 km long by 2 km wide) inside Morton National Park (henceforth, 'gated'). The other population was located 6 km east on a sandstone plateau (roughly 27 km long by 9 km wide) on crown land (i.e. public land without tenure; henceforth, 'ungated'). The two populations are separated by a steep valley dissected by a river, but there is some gene ow from the gated population to the ungated population 37 . At the gated population, there are two sets of locked gates at the entry and at access points of the re trail that traverses the plateau.
These gates were installed in 2008 and have since made it more di cult for people to access and disturb rock outcrops, including reptile poachers and rock collectors. By contrast, the ungated population is easily accessible to poachers because there are no locked gates on the re trails which crisscross the plateau. Many re trails throughout the ungated population terminate at rock outcrops, so poachers do not have to walk far (often just metres) to nd broad-headed snakes. Because previous publications detailing the speci c location of these populations have led to increased habitat disturbance and poaching of snakes 27 , we opted not to include location details or maps.
Since 1992 (gated population) and 2007 (ungated population), one of us (JKW) has carried out annual mark-recapture studies of these populations of broad-headed snakes. Each year during late winter and/or spring, a team of herpetologists (usually 3-5 people) surveyed four study sites at the gated population and three study sites at the ungated population. At each site, the team lifted all sun-exposed rocks that could be lifted without risking a back injury. All reptiles found under rocks were identi ed and recorded, and any broad-headed snakes or small-eyed snakes (Crytophis nigrescens) were hand captured and brie y held with thick welding gloves. If a broad-headed snake or small-eyed snake was captured, the researcher measured the snout-vent length and tail length (with a ruler, to nearest mm), determined the sex (via tail shape), and weight (with spring balance, to nearest g). We recorded the snakes' microchip number, and if the snake was unmarked, we injected a miniature transponder (Trovan Midichip 8 mm x 1.4 mm) under the skin. During surveys, the team noted whether rocks had been disturbed by humans, and if so, the nature of the disturbance (i.e. whether rocks were overturned, displaced, or broken). Disturbed rocks were easily identi ed because aside from being displaced or overturned, they often had the remains of squashed invertebrates or vertebrates (lizards and frogs) beneath them. Any unmarked rocks were given a unique identi cation number (with a paint pen, underneath the rock) to enable us to assess long-term usage of rocks by snakes, and disturbance to rocks. After processing each snake, the rock was returned to its exact original location, and the snake was released under the rock.
All procedures were approved by the University of Technology Sydney Animal Ethics Committee and were carried out in accordance with relevant guidelines and regulations under licence from state and federal wildlife agencies.
Evidence for removal of snakes from the ungated population Because sites at the ungated population had higher levels of disturbance than sites at the gated population 33 , we hypothesised that humans were removing snakes for the illegal pet trade. If poaching has occurred, broad-headed snakes from ungated sites should have lower survival rates than snakes from gated sites. To test this hypothesis, we analysed mark-recapture data collected from 2007-2019 using Cormack-Jolly-Seber (CJS) models in Program MARK v9.0 38 . Because previous studies have shown that survival rates vary with age and size 28,35 , we allocated each snake to one of two size classes (subadults and adults, SVL > 349 mm [henceforth 'adults'], and juveniles, SVL < 350 mm). To investigate whether survival rates differed among populations, we allocated each snake to one of two populations (gated or ungated). Thus, there were four groups in the input le: gated adults, gated juveniles, ungated adults, and ungated juveniles. Next, we ran a series of models in MARK to test the following a priori hypotheses: (1) survival rates are higher in gated than ungated sites; (2) survival rates of adults and juveniles are higher in gated than ungated sites; (3) survival rates vary through time independent of location; and (4) survival rates are constant independent of location. For these hypotheses, we ran equivalent survival models in which the probabilities of recapture were constant, time-dependent, or group-dependent. We then used the Akaike Information Criterion corrected for small sample size (AICc) to identify the most parsimonious model from the candidate model set 39 .

Estimation of life history parameters
Life history parameters were either estimated in this study or were taken from previously published studies of these populations (Table 1). Because good estimates of parameter uncertainty are necessary to construct informative stochastic demographic models, we used program MARK to obtain estimates of environmental (process) variation around survival rates. We did this using the variance components subroutine in MARK (appendix D, MARK Book v 19, see 40 ). For this analysis, we used the mark-recapture data set for the gated population, with two groups (juveniles and adults). We then ran CJS models, and estimated variance components for each age class separately from the model . Initial population size estimates were obtained from previous analyses that extrapolated estimates from study sites to the plateau (Table 1 28 ). Although our initial population sizes are extrapolations upon estimates and are potentially imprecise, we accounted for this uncertainty by varying initial population size to see what effect it has on population growth rates (see below). Using previously published literature from these populations and parameters estimated in this study, we obtained the following life history parameters required for population viability analysis: (1) age at maturity for males and females; (2) maximum age of reproduction; (3) maximum litter size; (4) sex ratio at birth; (5) percentage of females breeding annually; (6) mean number of offspring per female; (7) biological survival rates calculated by age class; and (8) dispersal rates (Table 1).
Population growth rate is central to our ability to predict population dynamics 41 and is essentially governed by rates of fecundity, survival, immigration and emigration. Our ability to estimate fecundity and survival is relatively robust due to the long-term demographic data available for these populations.
Although we cannot differentiate between mortality and permanent emigration, adult broad-headed snakes show site delity, and are often recaptured underneath the same rocks where they were originally captured 29,42 . Furthermore, there is virtually no suitable habitat outside of the national park into which snakes can permanently emigrate 31,43 . Population genetics for our study area showed that dispersal occurred in the juvenile life stage and was unidirectional, with juvenile dispersal movements occurring only from the gated population into the ungated population 37 . Thus, we have reasonable justi cation to treat the metapopulation as closed.

Population trajectories using population viability analysis
To determine the effects of poaching on the ungated population, we used the program Vortex 10.2.5.0 44 to conduct 100-year baseline scenario population viability analyses (PVA), using 1000 iterations to obtain a mean population growth rate and a probability of population extinction. The baseline scenario ( Fig. 1) uses our best estimates for life history parameters, all of which are either estimated in this study (see Table 1 & Results) or are taken from previously published results from these populations of broad-headed snakes ( Table 1). Where we were uncertain about any life history parameters, we employed subsequent sensitivity tests that arti cially varied the parameters to determine their effect on population growth rate (see below; Table 2).
Using our mark-recapture data we calculated the age-speci c survival rates of each population (see Results) and found that the annual survival of our ungated population was substantially lower than that of our gated population for both juveniles and adults (Table 1). These survival rates, however, included the impacts of illegal poaching activities. To investigate how poaching affects population viability, we made the assumption that without this impact both populations would have the same survival rates as the gated population. Therefore, for all subsequent analyses, we assumed all life history parameters to be equal between both ungated and gated populations, with unidirectional dispersal from the gated population into the ungated population.

Sensitivity analysis
Whilst useful for predicting population trajectories 45 , population viability analysis does not provide quantitative information about which parameters most in uence population trajectories through time, nor does it account for uncertainties in estimated life history parameters 46 . To account for these uncertainties and directly assess the in uence of different parameters, we arti cially varied the values of ve parameters for sensitivity testing in Vortex, including: (1) immigration; (2) initial population size (n); (3) carrying capacity (k); (4) harvest rate of adult females; and (5) harvest rate of adult males ( Table 2). All parameters of the gated population were maintained at base scenario levels for sensitivity analysis.
Since we were interested in the relative importance of each parameter on the population growth rate, we limited variation around the parameter to values that we as experts considered biologically plausible. Unidirectional dispersal of juvenile snakes (1-3 years) from the gated population to the ungated population was estimated to be at a rate of ~4% of the juvenile population of origin (Table 1 37 ), and this value was allowed to vary between 0-8%. The size of our gated population had been previous estimated as 595 (95% CI: 389-808) snakes 27 , and since the available habitat of both ridges is similar in extent, we allowed initial ungated population to vary between 100-1200 snakes. We assumed the lowest possible carrying capacity for both populations to be close to the estimated initial population size and allowed the carrying capacity of the ungated population to double (n = 600-1200). Since poaching of broad-headed snakes is illegal, we have little knowledge about how many snakes are removed each year. We do, however, know the difference in survival between the gated and ungated populations. Since annual adult survival in the ungated population is about ~22% lower than that of the gated population, we can estimate that about 22% of adults (n ≈ 48) at the initial total population size (n = 600) are being removed from the population, or at least are not persisting in the population, each year. Given this uncertainty, to investigate the effect of poaching on population viability, we varied the rate of harvesting of male and female adults between 1 -30 individuals, respectively (Table 2). In each iteration of the sensitivity analyses, stochastic variation was maintained at the same level as the base scenario. To optimise sampling of the available parameter space for each parameter, we ran 250 sensitivity samples in Vortex using the Latin hypercube sampling method.
For analysis, the summary output le from the sensitivity test in Vortex was read into the statistical program R version 4.0.2 47 . To allow for comparisons between the effects of each parameter on the stochastic population growth rate, all predictor variables were scaled and centred. To ensure there was limited correlation between the arti cially manipulated predictor variables, we created pairwise scatterplots and calculated correlation coe cients using the pairs.panels function using the 'psych' package 48 . A generalised linear model was constructed with the stochastic population growth rate as the response variable and all arti cially varied parameters as predictor variables. Model predictions for the population growth rate were plotted against the main effects to visualise the relative slopes of the predictor variables. The anova function was used on the model object to produce an analysis of deviance table, and the percentage of deviance explained by each of the predictor variables was recorded.

Effects of poaching on population viability
To investigate and visualise how poaching may be affecting the population viability of the ungated population, we maintained the assumption that without the impact of poaching both populations would have the survival rates of the gated population. To investigate whether this was a sound assumption, we ran a second population viability analysis, where survival rates were kept at the observed levels for the gated population and harvesting of adult females (n = 22) was imposed on the ungated population. We also included unidirectional dispersal from the gated population into the ungated population.

Results
Evidence for removal of snakes from the ungated population Survival analyses in program MARK provided strong support that survival rates () differed between snake size classes and population ( Table 3). The best-supported model , was one in which survival rates varied with snake size class and population, and recapture rates varied from year to year (Table 3). From this model, survival rates were lower for the ungated population than for the gated population. Annual survival rates were 0.67 (± 0.11 SE) for adults and 0.39 (± 0.16) juveniles in the ungated population, and 0.86 (± 0.05) for adults and 0.57 (± 0.08) juveniles in the gated population. A second model had almost equivalent support from the data. From this model, the annual survival rate of snakes was 0.84 (± 0.05) for the gated population and 0.55 (± 0.08) for the ungated population. A third model had some support, but none of the other models we tested had any support from the data ( Table 3).

Baseline population viability analysis
The results of the baseline population viability analysis showed that the gated population has a positive growth rate (r = 0.084) and rapidly grows to the carrying capacity which we imposed upon it (Fig. 1). This may suggest that the carrying capacity of this population is actually likely to be closer to our estimated initial population. The ungated population, however, has a negative growth rate (r = -0.246) and only avoids population extinction because it is sustained by the small amount of juvenile dispersal from the gated population (Fig. 1).

Sensitivity analysis
Sensitivity analyses showed that three of the ve factors we varied had a signi cant effect on stochastic growth rate. Carrying capacity (k; X 2 (1) = 11.26; p < 0.001) and harvest rate of adult males (X 2 (1) = 7.17; p < 0.01) signi cantly modi ed the stochastic growth rate, but both accounted for < 5% of the total deviance explained by the model (Table 4). Harvest rate of adult females, however, had a large effect on the stochastic growth rate (X 2 (1) = 226.92; p < 0.0001) and accounted for 91% of total deviance explained by the model (Table 4). Importantly, the analysis showed that populations were extremely sensitive to even small changes in adult female survival (Fig. 2). When all other parameters were set to their mean values ( Table 2), harvest of greater than ~15 females per year resulted in a negative stochastic growth rate (Fig. 3).

Effects of poaching on population viability
When we ran a population viability analysis using the same population parameters for each population, with unidirectional dispersal (Table 1) and a harvest rate of 22 adult females per year, we observed that the ungated population has a negative growth rate (r = -0.025) and only avoids population extinction because it is sustained by the small amount of juvenile dispersal from the gated population (Fig. 4).

Discussion
The global demand for reptiles for the exotic pet trade is growing, and each year millions of reptiles are tra cked 2 . However, determining the impacts of poaching on wild populations remains a formidable challenge 12 . Here, we provide one of the rst attempts to model the impact of illegal poaching on a population of threatened reptiles. Our results provide strong evidence that human poaching of endangered broad-headed snakes from wild populations to supply the underground, illegal pet trade is capable of affecting population-level impacts. Over the 12-year period from 2007-2019, humans disturbed rock outcrops on the ungated plateau every year, and they disturbed more ungated sites than gated sites 33 . Consistent with the hypothesis that humans removed snakes from the ungated population, we found that annual survival rates of broad-headed snakes were ~22% lower in the ungated population than the gated population. Population viability analysis using the life history parameters of each population show that the ungated population is on a trajectory towards local extinction, presumably due to the impacts of illegal poaching for the wildlife trade.
Our sensitivity analyses showed that increases in adult female mortality rates had the most signi cant effect on population growth rates of broad-headed snakes, consistent with the results of demographic studies on other long-lived species (e.g. 49,50 ). However, what was most surprising was that, without dispersal, the removal of less than 20 adult females per year from the ungated population was su cient to essentially drive this population to local extinction in the next few decades. While few studies have modelled the effects of poaching on population growth, other studies that have modelled the effects of small increases in mortality rates on long-lived vertebrates have found similar effects. For example, in a population of the Egyptian vulture Neophron percnopterus, small increases in adult mortality from wind turbine collisions decreased population sizes and time to extinction 51 . Similarly, the current levels of illegal poaching of broad-headed snakes are not sustainable. During annual surveys of the ungated population, we found 0.46 adult female broad-headed snakes per hour of survey effort (J. K. Webb unpubl. data). It would, therefore, take approximately 5 days for a team of two people to remove 20 adult females from this population. Because broad-headed snakes are easy to locate on these plateaux during the cooler months, it would not be di cult for poachers to remove a high proportion of females from this population. Studies on other reptiles show that even a single poaching event can dramatically impact small populations. For example, in February 2010, three poachers removed almost half of the breeding population of jewelled geckos from a small population in New Zealand 52-54 .
Our viability analyses suggest that the only factor rescuing the ungated population from imminent extinction is immigration of juveniles from the adjacent and better protected gated population. However, the model ignores habitat quality, which is also an important determinant of population viability 55 . Numerous studies have shown that all else being equal, populations are more likely to persist in patches of higher quality habitat [56][57][58] . With the exception of our study sites, which were restored with arti cial rocks 59,60 , the habitat quality of rock outcrops on the ungated plateau is lower than on rock outcrops on the gated plateau 33 . Over the last three decades, bush rock collectors have removed sandstone rocks from rock outcrops on the ungated plateau to supply the landscape garden industry 61 . Most rock outcrops on the ungated plateau are accessible via re trails, and consequently, many outcrops have been stripped of their natural rocks 26,33 . The rocks stolen by rock thieves are similar in size to those selected by broad-headed snakes, and the velvet gecko Amalosia lesueurii 61 , which forms a large component of the diet of juvenile broad-headed snakes. The removal of surface rocks will decrease prey abundance 61 and the availability of thermally suitable shelter sites, which may render juvenile snakes more vulnerable to predation from birds or sympatric small-eyed snakes 62 . Thus, poor habitat quality may contribute to lower survival rates of immigrants, so that the rescue effect of immigration may be negligible 63 .
There are a number of biological attributes that are thought to predispose a species to risk of extinction.
Although some species, including some reptiles (e.g. 21 ) can clearly sustain high levels of harvesting from the wild without suffering negative growth rates 64 , some species cannot sustain even low levels of wild harvesting. Factors such has high trophic level, low population density, slow life history, small geographical range size and ecological specialisation are all signi cantly associated with high extinction risk 24,50 , and may make some species exceptionally vulnerable to the impacts of the wildlife trade.
Unfortunately, broad-headed snakes are affected by a number of these attributes and are threatened with extinction across much of their restricted geographic distribution, a 250km radius of Sydney, Australia's largest city. Because of the habitat specialisation of this species, they are vulnerable to destruction of the fragile sandstone habitats they are dependent on. Unfortunately, these ridgetops also provide desirable locations for housing developments, and because of this, the species has gone extinct from nearly all locations outside of national parks. Unfortunately, illegal activities, such as bush rock removal, habitat disturbance by amateur herpetologists, and illegal poaching, continue to further threaten the species even in areas where they are protected from habitat clearance 33,65 . The results of the current study suggest that the slow life history of broad-headed may make them more vulnerable to threats that reduce adult female survival than had previously been recognised.
To date, most studies investigating the persistence of broad-headed snakes have focused on their ecological specialisation as the cause of their engenderment (e.g. 27,61,66 ). In this study, however, we show that the slow life history of this species results in even small amounts of illegal poaching threatening its persistence independently of habitat destruction. It is, however, possible that the lower rates of survival we detected in the ungated population is the compounding effects of both poaching of snakes for the pet trade and destruction of non-renewable habitat features in the process of searching for snakes. While our models ignore habitat quality, previous studies of this species have shown that disturbance to shelter rocks signi cantly changes their thermal pro le, making them unsuitable for broad-headed snakes and their prey 55 . Potentially, our ungated population does not require poachers to remove ~15 reproductive females a year to cause the population to trend towards negative growth. If poachers simply recklessly disturb rocks in their pursuit of the snakes, they may be impacting the survival of far more snakes that they are actually taking. Similarly, the impacts of direct habitat disturbance by humans and the impacts of illegal poaching are not operating in isolation from other anthropogenic disturbances. For example, the massive 2019-2020 wild re season in south-eastern Australia resulted in substantial proportions of the distributions of many threatened species being burnt at high severity, including the broad-headed snake 32 . Impacts such as these can potentially compound the effects of poaching and bush-rock removal and signi cantly reduce dispersal between plateaux via old growth forests 36,37,42 , resulting in higher levels of mortality and increasing uncertainty for population viability.
Concerningly, species that are perceived to be rare in the wild may actually be at greater risk to exploitation because their perceived rarity increases demand for them as a commodity [67][68][69][70] . Traditional economic theory predicts that wild harvesting alone is unlikely to cause species' extinctions due to diminishing return-the escalating cost involved in nding an increasingly di cult to nd species 71 . Fisheries, for instance, often suffer from the impacts of over shing, however, over shing rarely result in the global extinction of species, because the cost of shing for declining species typically increases with their rarity while their value remains static 64 . However, if disproportionate value is placed upon a species as a consequence of its rarity, this may result in a situation where increased exploitation reduces the population size, which in turn increases its value and ultimately leads to extinction 72 . Unfortunately for broad-headed snakes, even when population densities are very low, they are easy to detect because of their habitat speci city and proclivity to shelter under small sun-exposed rocks during winter. Because they attract high demand from snake keepers and photographers, and are easy to detect, illegal poaching of broad-headed snakes will likely remain a future threat to the viability of extant populations.
Globally, we face a rapidly accelerating extinction crisis. In many cases, we lack the detailed understanding of species' life histories and ecologies required to determine what is causing their extirpation. Our study highlights the value of detailed long-term demographic studies not only for identifying threats to populations, but for modelling the impacts of such threats on population viability. We also show that management actions aimed at deterring poachers and restricting their access to sensitive populations can help reduced those threats. In the current example, the erection of locked gates and installation of hidden cameras has helped reduce disturbance to the gated population, and this population is not at imminent risk of local extinction. By contrast, the ungated population is on a trajectory to extinction, and urgent measures are necessary to reverse this decline. The installation of locked gates and cameras, and restoration of degraded habitats, may well help reverse the current decline, but tougher penalties are also required for poachers caught collecting or selling broad-headed snakes. If we can help authorities to enact such actions, we may yet be able to prevent an evolutionarily signi cant population of endangered species from going extinct.

Declarations
Acknowledgements We thank our colleagues, students and volunteers who have assisted us with annual broad-headed snake surveys over the last three decades. We are indebted to Ben Croak, who generously provided us with data from his habitat restoration study, and Rick Shine, who supported and participated in the research during the early years of the project. We particularly thank Les Mitchell, Phil Craven and Bruce Gray (NSW National Parks and Wildlife Service) for providing logistical support and access to gated sites, and the Department of Lands who provided access to lands under their care. We thank Megan Hinds (NSW Department of Planning, Industry and Environment) for managing and supporting the broad-headed snake recovery program. The research was supported nancially by the Australian Research Council and a Saving Our Species grant administered by the NSW Government.
Author contributions JKW collected the data and conceived the ideas; CJJ and BvT conducted data analysis; CJJ led the writing with assistance from all authors; all authors contributed to framing and intellectual content.
Competing interests We declare no con ict of interest.
Ethics approval Animal ethics approval was granted by University of Technology Sydney Animal Ethics Committee.
Availability of data and material All data available upon request from authors.
Code availability All code available upon request from authors.  Table 3. Results of Cormack-Jolly-Seber analyses in MARK that was used to model rates of survival (S) and recapture (p) of broad-headed snakes. Each snake was assigned to one of four groups depending on its size at rst capture (sub-adults and adults, or juveniles) and population (gated or ungated).