Frog Survival and Population Viability in an Agricultural Landscape with a Drying Climate

Received: 18 December 2017 Revised: 15 June 2018 Accepted: 24 September 2018 Published on: 25 December 2018 DOI: 10.1002/1438-390X.1001

ORIGINAL ARTICLE

Frog survival and population viability in an agricultural landscape with a drying climate

Robert A. Davis1,2 | Cheryl A. Lohr3 | J. Dale Roberts2,4

1School of Science, Edith Cowan University, Joondalup, Western Australia, Australia Abstract 2School of Biological Sciences, University of Amphibians are the most threatened class of vertebrate in the world. Although a Western Australia, Perth, Western Australia, number of causes of the amphibian decline phenomenon are emerging, there is a Australia need for robust demographic data to be able to monitor current and future threats 3 Department of Biodiversity, Conservation and such as climate change. Despite this, few studies on amphibians have the life- Attractions, Science and Conservation Division, Perth, Western Australia, Australia history data available to undertake these analyses and fewer still have looked at the — 4Centre of Excellence in Natural Resource challenges to population viability posed by fragmentation a feature inherent in Management, University of Western Australia, agricultural landscapes where the matrix is highly modified. Our aim was to inves- Perth, Western Australia, Australia tigate the population viability of a large burrowing frog in an agricultural land- Correspondence scape. Specifically, we aimed to investigate the future persistence of populations Robert A. Davis, School of Science, Edith Cowan University, 270 Joondalup Drive, Joondalup, WA under a range of scenarios including populations connected by various levels of 6027, Australia. dispersal and reduced rainfall. We used the life-history parameters of Heleioporus Email: [email protected] albopunctatus, a frog species widely distributed in the extensively cleared agricul- Funding information tural regions of south-western Australia. We investigated the viability of 24 partially Australian Geographic; Peter Rankin Trust fund for connected populations under a range of scenarios using the program Vortex herpetology; University of Western Australia Version 10.1.6.0. Metapopulations were consistently more robust to extinction than isolated local populations. Both meta- and local populations were more susceptible to increases in age-specific mortality rates than to variation in the estimated ability of H. albopunctatus to disperse between breeding ponds, the survival rate of dispersers, or the frequency of drought. Our results reinforce the importance of metapopulations for survival in fragmented landscapes and point to the need to manage amphibian breeding ponds across landscapes to ensure high survival rates, particularly for juveniles.

KEYWORDS amphibian, climate change, drought, population viability analysis, PVA

1 | INTRODUCTION found little evidence of large-scale drivers of decline and stated the importance of identifying and understanding local Amphibians are the world's most threatened class of verte- threats and appropriate solutions (Grant et al., 2016). brate that has been completely assessed (IUCN, 2016), and Underlying all investigations of local decline is the need there has been great concern about both the cause and scale for both robust demographic data on populations, and long- of declines of amphibians globally (Stuart et al., 2008). Vari- term monitoring of populations. In the case of the Harlequin ous causes of declines have been proposed, the most signifi- frog, Atelopus cruciger (Lichtenstein and Martens, 1856), cant being the role of the chytrid fungus (Lips, 2016), careful demographic monitoring challenged the perceived climate change (Croteau, Davidson, Lean, & Trudeau, 2008) declining status of a population and proved that visual and habitat modification (D’Amore, Hemingway, & encounter survey techniques that had been used for monitor- Wasson, 2010). Despite this, a recent extensive analysis ing were not accurate (Lampo, Celsa, Rodriguez-Contreras,

Rojas-Runjaic, & Garcia, 2011). Because many species exist indicate the presence of small, subdivided populations, as metapopulations, understanding dispersal processes and evidence of high larval mortality and instances of no or population viability at larger scales is important, particularly very low recruitment. This is counterbalanced by a small if attempting to model the viability of species in fragmented number of populations with positive demographic fea- landscapes (Crawford, Peterman, Kuhns, & Eggert, 2016; tures, which may indicate the possibility of a source–sink Heard, Scroggie, & Malone, 2012). Complete demographic metapopulation structure (Harrison, 1991). If this is the datasets are generally rare for amphibians and where such case, it would place management importance on maintain- datasets exist; they can, and should, be usefully applied to ing “core” breeding habitats with successful annual determining the long-term viability of populations recruitment to maintain this species at a landscape scale (Conroy & Brook, 2003). (Davis & Roberts, 2005a). Habitat loss has been proposed as the leading cause for Our objective was to model the population viability of the decline of Australian frogs (Hero et al., 2006), yet we H. albopunctatus. We investigated the sensitivity of three have a very limited understanding of amphibian persistence different metapopulation structures to a decline in breeding- in fragmented landscapes. Some studies have found that season rainfall and associated breeding-pond hydroperiods. metapopulations are essential for regional persistence Both are scenarios associated with climate change but also (Greenwald, 2010) and that juvenile survival and dispersal features of the historic climate patterns in the region may be particularly important in increasing population per- (Cullen & Grierson, 2009). We aimed to predict the likely sistence (Di Minin & Griffiths, 2011). persistence of this species in a highly modified landscape Here, we used a comprehensive demographic dataset for that still presents opportunities for successful persistence. the Western spotted frog Heleioporus albopunctatus H. albopunctatus occurs across all of its known range (Myobatrachidae; Gray) (Davis, 2004), to model population (Burbidge et al., 2004) and uses a wide range of water bod- processes and future survival at a landscape scale. Over the ies for breeding that have been generated by human modifi- majority of the range of H. albopunctatus, there has been cation of the landscape (Davis, 2004). large-scale habitat loss and fragmentation caused by agricultural activities (Hobbs, 1993). A detailed life-history provides an ideal opportunity to model the viability of an 2 | MATERIALS AND METHODS amphibian metapopulation in a fragmented landscape, and to test the robustness of populations to multiple real-world sce- 2.1 | Study site narios, including a projected decline in rainfall. We used data from 24 H. albopunctatus sub-populations The Western spotted frog H. albopunctatus, is a large found in remnant native vegetation near Kellerberrin, West- burrowing dryland frog, distributed widely throughout the ern Australia (Figure 1). The whole metapopulation had a semi-arid region of Western Australia from near Kalbarri in total of 14 connections between sub-populations. Isolated the north-west to Esperance in the south-east (Tyler & sub-populations have zero connections. Connected sub- Doughty, 2009). Much of this distribution coincides with the populations have one to four neighboring sub-populations. wheat-sheep farming region of Western Australia, and The area may be characterized as semi-arid arable land with accordingly, much of the range of this species has been sub- low, unpredictable rainfall and localized thunderstorms ject to habitat loss, fragmentation and subsequently, saliniza- (Hughes, Guttorp, & Charles, 1999). The major habitats of tion (e.g., George, Clarke, & English, 2008; Saunders, the study area include woodlands dominated by Salmon Hobbs, & Arnold, 1993) involving the clearance of Eucalyptus salmonophloia 2 Gum ( , F. Muell.), York Gum 140,000 km in less than 150 years (Saunders, 1989). In (Eucalyptus loxophleba Benth.) and Gimlet (Eucalyptus most areas of the wheatbelt, less than 10% of the native veg- salubris F. Muell.) interspersed with Jam Wattle (Acacia etation now remains, mostly as small, isolated remnants acuminata, Benth.) woodlands and proteaceous sandplains (Hobbs, 1993). Such rapid and widespread landscape change (shrubby sandplains dominated by nectar-rich species of the has had a great impact on endemic fauna and flora. Habitat family Proteaceae). All frog breeding sites comprised low- fragmentation and salinity caused by clearing are the key lying winter-wet depressions and included anthropogenic mechanisms responsible for the loss of plant and animal spe- ponds, farm water interceptor ditches (for overland runoff ), cies in this region and will have ongoing future impacts roadside ditches and abandoned sand quarries. Details of (Saunders, 1989; Saunders et al., 1993). populations sampled are in Table 1. Although H. albopunctatus currently appears widespread and secure (Burbidge, Rolfe, McKenzie, & Rob- 2.2 PVA modeling erts, 2004; Roberts, Conroy, & Williams, 1999), detailed | studies of genetic structuring (Davis & Roberts, 2005a), Population viability analysis (PVA) is an effective, trans- adult survival (Davis & Roberts, 2011), larval survival parent and widely used process (Brook, Cannon, Lacy, (Davis, 2004) and egg survival (Davis & Roberts, 2005b) Mirande, & Frankham, 1999; Brook, Lim, Harden, & 104 DAVIS ET AL.

FIGURE 1 Map of H. albopunctatus metapopulation included in this population viability analysis. Boxes provide smaller scale view of connected sub- populations. Non-boxed populations are isolated. Letter codes refer to population sampled (see Table 2)

Frankham, 1997; Lindenmayer & Lacy, 1995; Towns, Par- the number of progeny produced by each female each year, rish, & Westbrooke, 2003) for synthesizing demographic and to determine which of the two alleles at a genetic locus data and exploring the long-term impacts of stochastic and are transmitted from each parent to each offspring (Lacy, deterministic factors on the persistence of populations 1993, 2000). Fecundity is assumed to be independent of age (Reed, O’Grady, Brook, Ballou, & Frankham, 2003). There after an animal reaches reproductive age. Mortality rates are has been contention about the accuracy and role of PVA in specified for each pre-reproductive age-sex class and for predicting extinction probabilities, particularly with scant reproductive-age animals. Population carrying capacity is data (e.g., Coulson, Mace, Hudson, & Possingham, 2001); imposed by a probabilistic truncation of each age class when however, comprehensive studies have unequivocally advo- the population size exceeds the user defined carrying capac- cated the importance and robust nature of PVA in produc- ity. Environmental variation is incorporated into VORTEX ing unbiased predictions (Akçakaya & Sjögren-Gulve, by sampling birth rates, death rates and the carrying capacity 2000; Brook et al., 2000; Hamer et al., 2016). One PVA from a user-defined binomial or normal distribution. Catas- simulation model, VORTEX, is frequently used as a predic- trophes are modeled as sporadic random events that reduce tive tool by the Species Survival Commission of the Inter- survival and reproduction for 1 year (Lacy, 1993, 2000). national Union for Conservation of Nature (IUCN) Users define the probability of a catastrophe occurring and (e.g., Estrada et al., 2014). the probable effect of the catastrophe on species survival and We used Vortex Version 10.1.6.0 (Chicago Zoological reproductive rates. VORTEX allows multiple subpopula- Society, Brookfield, Illinois, USA) (Lacy & Pollak, 2014) tions to be tracked, with user-specified migration among the for PVA of populations of H. albopunctatus. VORTEX is an units. The model it incorporates is most applicable to species individual-based simulation program that models population with low fecundity and long lifespans, such as mammals, processes (fecundity, mortality) as discrete, sequential birds and reptiles (Lacy, 1993). We did not include inbreed- events, with probabilistic outcomes (Lacy, 1993; Lacy, ing in our simulations, as this is unlikely to be a major factor 2000). VORTEX uses random numbers to determine in the viability of populations of H. albopunctatus (Davis & whether each animal in the model lives or dies, to determine Roberts, 2005a). DAVIS ET AL. 105

TABLE 1 Sub-population specific parameters

Population ID Location UTM zone UTM WGS 84 E UTM WGS 84 N Population size Initial abundance MM Mount Moore 50 623,089 6,547,065 Large 65 BR Billyacatting Rock 50 595,074 6,565,046 Large 65 CH Chingha Hills 50 637,029 6,495,041 Large 65 HD Hunt's Dam 50 623,257 6,520,351 Medium 35 TNR Tandegin Rock 50 634,072 6,502,074 Large 65 TR Totagin Rock 50 615,460 6,506,351 Large 65 MS Mount Stirling 50 556,092 6,477,758 Large 65 YR2 Yoting Rock 2 50 556,051 6,477,728 Large 65 DW Deep Well Road 50 568,684 6,512,292 Medium 35 RS Roger Scott's property 50 562,973 6,499,860 Small 12 CKL Creekline in Woollering NR 50 569,436 6,512,680 Medium 35 MR McNeil Rd 50 560,114 6,488,686 Medium 35 HMD Hammond 50 571,085 6,522,828 Small 12 MC Mount Cramphorne 50 662,014 6,477,315 Medium 35 KR2 Kokerbin Rock 2 50 567,399 6,472,528 Small 12 KR Kokerbin Rock 50 566,241 6,472,176 Small 12 HW Hunt's Well 50 546,772 6,496,202 Medium 35 YKR Yorkrakine Rock 50 548,383 6,523,246 Large 65 H2 Hills Site 2 50 456,057 6,428,063 Medium 35 H1 Hills Site 1 50 456,049 6,429,910 Small 12 WI Wilkins 50 569,603 6,513,652 Medium 35 MO Morley 50 569,487 6,513,693 Medium 35 MW Morley West 50 568,865 6,513,185 Large 65 LK Leake 50 571,325 6,522,828 Large 65

Population sizes are classified based on field experience and comparison with the largest and smallest known sites (see Davis & Roberts, 2005b). Initial abundance was based on the number of active burrows with calling males.

2.3 | Estimation of life-history traits opportunities to breed and thus potentially is a best-case sce- All key life-history data (Table 2) for PVA were derived nario for populations. A stable age distribution was from our population studies on H. albopunctatus (Davis, assumed. 2004; Davis & Roberts, 2005a, 2005b, 2011), except for 2.3.2 | Reproduction juvenile survival and longevity, for which we were not able to secure field data and have estimated parameters from the Field observations over 4 years suggest that there is no evi- literature. We used the results of our studies to define realis- dence that H. albopunctatus mate more than once per breed- tic demographic parameters (described below) for ing season or that matings are polyandrous. We assumed a 24 H. albopunctatus populations from a similar, but slightly 1:1 sex ratio and full availability of both sexes at breeding broader study area in VORTEX (Table 2). sites. All females captured in each population were assumed to be breeding or have just bred, as very few frogs have ever 2.3.1 | Age been sighted at breeding sites outside the breeding season. Lee (1967) states that all Heleioporus species attain sexual We varied the proportion of females breeding within each maturity and breed only after 2 years of age. In our analysis season across modelled scenarios and tested the impact of ages 0–1 year were defined as metamorphs and 1–2 years this variation on the probability of population persistence are assumed to be juvenile. The upper age limit for during our sensitivity analysis. The average number of off- H. albopunctatus in this PVA is derived in part from an spring per female (hereafter fecundity) was derived from unpublished honors study (Thom, 1995) that estimated ages data obtained during studies of clutch size (Davis, 2004), of up to 7 years for this species based on skeletochronology. embryonic mortality (Davis & Roberts, 2005a) and larval There have been few studies on longevity in Australian frogs survivorship (Davis, 2004) and calculated as the average against which to compare, but a smaller species of Myoba- number of metamorphs successfully recruited per single egg trachid studied by Driscoll (1999) lived for 6 years. We clutch (34 8). The maximum number of progeny per elected to use 10 years as the maximum age based upon the brood (132) was derived from the best-case scenario using margin of error associated with skeletochronology studies maximum observed values for clutch size (876), probability and the notion that a longer-lived frog has more of egg survival (1), probability of embryonic survival to 106 DAVIS ET AL.

TABLE 2 Model input parameters for Vortex simulations of our model, but there is little alternative where such data is H. albopunctatus population viability lacking. Our sensitivity analysis is designed to identify demo- Realistic Parameter graphic parameters that require additional refinement. Parameter model variation Reproductive system Monogamous – 2.3.4 | Population size – Maximum number of egg clutches per year 1 We included 24 H. albopunctatus populations studied – First breeding age of males and females 2 between 2000 and 2004 (Davis, 2004) in this analysis. Initial – Maximum age 10 population sizes varied from 12 to 65 (Table 1) and were Carrying capacity 300 – based on maximum breeding burrow counts at each site Average number of metamorphs per year 34 (8) – (Davis & Roberts, 2011) which have been shown to provide Maximum number of metamorphs per year 132 – a reliable estimate of population size (Davis & Roberts, Sex ratio at birth 1:1 – 2005b). Carrying capacity was based on the largest potential Males in breeding pool % 100 – population size estimated for H. albopunctatus in the region Females reproducing per year/site % 100 60–100 (Davis, 2004) and set at a consistent value across the Percent annual mortality (SD) 24 populations (Table 2). Davis (2004) derived carrying Adult females (> 2 years of age) 44.25 (38.79) 25–65 capacity from the availability of suitable breeding sites Adult males (> 2 years of age) 38.32 (32.28) 18–58 (sandy depressions with sub-surface soil moisture and hold- Juveniles (age 1–2) 38.32 (32.28) 20–60 ing water in winter). This model does not include sites with Metamorphs (age 0–1) 77.5 (N/A) 55–95 zero recruitment and thus represents the integrated set of Infidelity % (I)1010–90 “source” populations in a source–sink metapopulation model Survival of dispersers % 62 10–90 Age range of dispersers 1–7 – (e.g., Harrison, 1991). Frequency of drought % 33 13–53 2.3.5 | Dispersal among sub-populations Influence of drought as a proportion of normal values on Reproduction 0.45 – Because genetic and mark-recapture studies both suggested Survival 1 – dispersal between subpopulations (Davis & Roberts, 2005a, Frequency of high rainfall % 27.5 7.5–47.5 2011), a metapopulation model (rather than a single popula- Influence of high rainfall as a proportion of normal values on tion) was used to realistically represent the spatial structure Reproduction 1.55 – of 24 populations across the landscape, and investigate the Survival 1 – impacts of habitat fragmentation. Mark-recapture data sug- Environmental variation (corr EV) 0.5 0.5–1 gest that H. albopunctatus have a low-moderate site fidelity correlation among populations (F =10–26%) and that on average 1.67% of frogs survive Stable age distribution? Yes dispersing between connected sub-populations LK and Trend in K?NoHMD and the connected sub-populations WI, MO and MW, Numbers in parentheses indicate SDs. The parameter variation column denotes despite travelling up to 873 m across agricultural land values used during sensitivity analyses. (assuming straight path of travel; Figure 1; Davis, 2004). The congener, the Moaning frog Heleioporus eyrei (Gray) is metamorphosis (0.66) and the probability of ponds retaining known to move up to 2 km from breeding sites (Bamford, sufficient water to allow metamorphosis (0.23). 1992) and so we used this as a maximum dispersal distance. We used the hypothetical Equation (1), 2.3.3 | Survival − Mortality estimates for each life-stage (Table 2) were derived Dispersal probability ¼ Ie 0 005D ð1Þ from mark-recapture studies (Davis & Roberts, 2011). where I is percent infidelity (I = 100 − F), and D is distance Although no field data were available for juvenile survival between sites (m) to estimate the dispersal probability due to zero recapture of 500 metamorphs marked, an estimate among 24 populations of H. albopunctatus. With our hypo- of 22.5% was used for metamorph age class 0–1, and the thetical equation we assume the probability of dispersal is adult survival rate of 61.68% was used for the juvenile age somehow influenced by the distance to other potential breed- class 1–2, based on a literature review of anuran survival rates ing sites, but not to the probability of frogs surviving dis- presented in Conroy (2001, p. 52). We have assumed that persal. VORTEX treats the probability of dispersal and the frogs that do not expose themselves to predators via mating survival of dispersers as separate parameters. The probability behaviors but are nearing maturity and adult size will have of dispersal is specific to any pair of sub-populations, and higher survival rates, as seen in other anuran species (Lode, hence thought to be related to the distance between sub- Holveck, Lesbarreres, & Pagano, 2004; Ryan, Tuttle, & populations, whereas the survival of dispersers is a single Rand, 1982). We acknowledge that the lack of field data on number applied to each sub-population. We assumed that juvenile survival for our species may affect the accuracy of the survival of dispersers would be consistent across sub- DAVIS ET AL. 107 populations within each modeled scenario. If our hypotheti- 2.4 | Scenario projections and sensitivity analysis cal equation is correct, when 10% of frogs disperse (I = 10) Once we constructed the realistic model described above, we the survival of dispersers must equal 62% for the average investigated the response of metapopulation persistence to a probability of dispersal among the five populations listed variety of demographic and environmental scenarios that above to equal 1.67% as calculated by Equation (1). Con- were likely to impact populations (Table 2). We projected all versely, if 90% of frogs disperse (I = 90) only 6.9% of dis- stochastic simulation scenarios for 30 years, and 1000 itera- persers must survive for the average probability of dispersal tions of each model were run to ensure stabilization of the to equal 1.67%. We used sensitivity analysis to assess the results (Brook et al., 1997). Sensitivity analysis was under- impact of variation in the survival of dispersers on the proba- taken by systematically changing the inputs for percentage bility of population persistence (Table 2). of females breeding per year, dispersal probability, age- and sex-specific mortality rates, the frequency of drought and 2.3.6 | Catastrophes high-rainfall years and the correlation of environmental vari- For all models, the catastrophic effects of drought or high ation (corr EV) among sub-populations, five times in step- rainfall on the reproductive success of H. albopunctatus wise increments of 10 around the mean (Table 3), while were based on long-term (1962–2002) rainfall records col- holding all other variables constant (Beissinger, 1995; Di lected weekly by landholders in the study area. This period Minin & Griffiths, 2011). We report the simulation results realistically reflects long-term variation in rainfall in this for our realistic model and the results of our sensitivity anal- region based on analyses of tree rings going back several ysis in the form of general linear model with binomial distri- hundred years (Cullen & Grierson, 2009). In 2001 there was bution, comparing variation in the demographic input a drought resulting in a 55% reduction in population sizes or parameter to the probability of population persistence for attendance at breeding sites (Davis & Roberts, 2011). Using isolated sub-populations. The primary scenarios investigated rainfall data for 2001 (averaged 39 mm for the March–May were as follows: breeding period across five sites) and the long-term rainfall records, we calculated a drought (rainfall in March–May 2.4.1 | Pond drying ≤39 mm) frequency of approximately 1 year in three (0.33). The most obvious cause of recruitment failure from field The impact of high rainfall years (>100 mm in the breeding studies (Davis, 2004) was pond drying. Based on field stud- period and >400 mm per annum) on populations was calcu- ies of mean larval survival during drought years (Davis, lated from rainfall records as 1 year in 3.5 (frequency of 2004), we estimated that offspring production per female 0.28). Thus drought years were almost as frequent as high was reduced by 55%. The normal frequency of drought was rainfall years for reproduction. The positive impacts of good once every 3 years or 33%. During sensitivity analysis we years on recruitment were estimated as being of the same varied the frequency of drought, and hence pond drying order of magnitude as the negative impacts of drought on from 13 to 53% (Table 2). populations. Drought does not appear to reduce the probability of adult, metamorph or juvenile survival beyond the 2.4.2 | Survival influence of standard environmental variation, hence the Two important influences on population size are juvenile severity factor for survival was one (Davis, 2004; Table 2). and adult survival. Scenarios for juvenile or adult population

TABLE 3 Sensitivity analysis of results of population viability analysis (PVA) simulation scenarios for H. albopunctatus given variation in input parameters

Metapopulation (n = 1) Connected populations (n = 5) Isolated populations (n = 11) Range input Range Range Range 2 2 2 Varied input parameter values Sensitivity Adj. R Ps % Sensitivity Adj. R Ps % Sensitivity Adj. R Ps % Correlation of environmental 0.5–0.9 −4.18 0.99 7.5–29.9 −3.64 0.55 1.1–11.3 0.05 2.34−4 0.1–1.2 variation (corr EV) among populations Drought (% frequency) 13–53 −0.14 0.70 28.8–30.3 −0.12 9.27−4 1.8–14.3 0.14 2.10−3 0.1–1.2 Dispersal (F; infidelity rate) 10–90 3.37−3 0.16 25.7–43 9.64−3 0.31 2.4–28.8 2.01−3 0.02 0.2–1 Survival of dispersers % 5–62 0.01 0.88 14–29.9 0.03 0.41 0.3–11.3 −2.35−3 0.01 0.1–1.2 Metamorph age 0–1 mortality % 55–95 −0.17 0.96 0.1–98.5 −0.17* 0.93 0–88 −0.23 0.98 0–30.3 Juvenile age 1–2 mortality % 20–60 −0.16 0.91 7.7–95.4 −0.13* 0.86 0.1–72 −0.12 0.83 0–17.1 Adult mortality % 25–65 −0.16 0.85 13.5–1 −0.21* 0.91 0.7–99.4 −0.32** 0.99 0–90.2 Females breeding % 60–100 7.723 0.95 23.3–29.9 0.01 0.10 1.3–11.3 −8.16−3 0.05 0.1–1.2

Presented values are the slope of general linear model with binomial distribution. Ps ~ input parameter + intercept; n = number of sub-populations used in the analysis. Connected populations include populations with at least one connection to another subpopulation. *p < 0.05; **p < 0.01. 108 DAVIS ET AL. size were modeled by increasing or decreasing metamorph 3 | RESULTS mortality by 10 or 20%. 3.1 | Realistic model 2.4.3 | Dispersal In the realistic model, the metapopulation was declining but

The importance of dispersal to metapopulation persistence had a probability of population persistence (Ps) of 29.9% was investigated by increasing the infidelity rate (I; Equa- over 30 years (Figure 2). Isolated sub-populations (n = 11) tion (1)) of H. albopunctatus by increments of 10% from the had an insignificantly lower Ps (average = 5%) than con- realistic model. nected sub-populations (Ps = 11.7%; p = 0.25; Adj R2 = 0.54). The mean time to extinction for isolated popula- 2.4.4 | Environmental variation tions was 7.4 years, whereas the metapopulation lasted Climatic and environmental influences on populations are 18.8 years. The deterministic growth rate of the H. albopunc- particularly important in areas of low and unpredictable rain- tatus metapopulation was positive (r = 0.58). fall such as occurs in the range of H. albopunctatus. It is possible that there will be local variation in the probability 3.2 | Sensitivity analysis of variation in input and quantity of rainfall occurring (e.g., in localized thunder- parameters storms) in the southwest of Western Australia (Hughes et al., The results of the general linear models of each series of 1999) and that localized environmental variation creates per- parameter variations are shown in Table 3. The amount of sistently high recruitment success in some H. albopunctatus corr EV among H. albopunctatus populations had an insig- populations over others (Davis, 2004). The realistic model nificant effect on the Ps for the metapopulation with Ps rang- assumed that all populations, which were within an area with ing from 29.9% when the correlation was 0.5, to a Ps of less a radius of approximately 150 km, were unequally impacted than 7.5% when corr EV equal to 0.9 and all populations by weather conditions (i.e., corr EV = 0.5) which were experienced very similar environmental conditions based on drought and rainfall frequencies derived from long- (p = 0.64; Table 3). Evidently, sub-populations of H. albo- term rainfall data. To investigate the impact of environmen- punctatus are prone to extinction, but metapopulation tal variation in model simulations, we varied the corr EV dynamics allow the species to persist when environmental among populations from 0.5 to 1 (Table 2). conditions vary across the landscape. All subsequent sensitivity analyses used corr EV = 0.5 to determine which Number of connections 1.00 0 demographic parameters have the greatest influence on the 1 persistence of the H. albopunctatus metapopulation under 4 realistic weather conditions. 14 When corr EV was equal to 0.5, the frequency of drought did not have a significant effect on either the meta- 0.75 population (p = 0.698; Table 3) or isolated populations (p = 0.99). While we did increase the frequency of drought from 33% to 53% and simultaneously decreased the frequency of high-rainfall years, we did not alter the frequency 0.50 of average years. Meteorological research predicts that the annual average rainfall may decrease in the south-west of Australia, but that summer rain events and flood events may become more common (Hughes, 2003). 0.25 Increasing the infidelity rate (I; Equation (1)) had a direct but insignificant effect on the Ps for isolated sub-populations (p = 0.97; Table 3) and an insignificant effect on Ps for the

Probability of population persistence metapopulation (p = 0.90). Further analysis reveals that for connected sub-populations with corr EV = 0.5 the effect of 0.00 increasing the infidelity rate for H. albopunctatus was best − 0102030described by a quadratic equation (Ps = 0.22 2 5 2 Year x + 0.24x + 0.18; p < 1.51 ; Adj R = 0.64; vertex = 54.8, 0.24) in which increasing the infidelity rate up to 55% signif- FIGURE 2 Population viability over time, based on connectivity of sub- icantly increases the Ps, but further increases in the infidelity populations within the metapopulation. The connected metapopulation had P a total of 14 connections between sub-populations and a probability of rate reduce s. In contrast, increasing the survival rate for persistence (Ps) of 29.9%. Isolated sub-populations have zero connections. dispersers did not have a significant effect on the Ps for the Connected sub-populations have one to four neighbouring sub-populations metapopulation (p = 0.77). DAVIS ET AL. 109

The Ps for connected sub-populations was significantly breeding populations when sites do not fill with water during affected by the mortality rate for juvenile and adult frogs winter. The relevance of this to the current study is that (p < 0.05; Table 3). If the annual adult mortality rate future consideration may need to be given to designing dropped from the 44% measured in the field to 25% then the breeding wetlands to ensure the presence of core sites that

Ps for the metapopulation increased from 29.9% to 100%. continue to provide opportunities for recruitment, even in a Similarly, reducing the mortality rate for juveniles aged series of low rainfall years. 1–2 years from the average of 38.3% which was measured in When corr EV was equal to 0.5, increased infidelity rates the field to 30% increases the probability of persistence for up to 55% improved the probability of persistence for con- the metapopulation from 29.9% to 90.1%. Reducing the mor- nected sub-populations. Further increases reduced the Ps for tality rate for metamorphs aged 0–1 year from the literature sub-populations. The movement of migrants among popula- based estimate of 77.5–65% increases the probability of per- tions supports metapopulation dynamics (Harrison, 1991) sistence for the metapopulation from 29.9% to 64.2%. replenishing sub-populations that otherwise may be prone to extinction. Dispersers however, are subject to additional dispersal mortality factors. Hence, when the infidelity rate is 4 | DISCUSSION very high, the Ps for sub-populations declines as potential breeding adults are drawn away from the population. In the The existence of a metapopulation for H. albopunctatus wheatbelt of Western Australia H. albopunctatus have to appears to increase the long-term persistence of this species disperse across agricultural land, which provides less vegeta- (compared to local populations), even with dispersal at low tive cover and hence protection against the elements or pre- levels. This is an outcome reinforced by a number of similar dation. This is particularly so during the breeding season studies, though there are few comparable PVA models of when the annual cropping cycle is such that fields are barren frogs with similar life histories. Conroy and Brook (2003) or being ploughed in preparation for seeding, thus leaving used a stage-structured metapopulation model of two Geo- little crop cover for protection. A number of studies have crinia species, although they did not specifically examine Felis catus the impacts of variations in dispersal because this is shown predation of frogs by introduced cats extremely low in the species modeled. Heard, McCarthy, (Linnaeus) (Doherty et al., 2015; Martin, Twigg, & Robin- Scroggie, Baumgartner, and Parris (2013) used Bayesian son, 1996) and foxes Vulpes vulpes (Linnaeus) (Read & PVA and found that metapopulations of Litoria raniformis Bowen, 2001) in Australia and both species are a major pest (Keferstein) in close proximity to each other were less likely in Australia's southern agricultural regions. As cover from to suffer extinction. predators is so limiting in agricultural regions, particularly In our study, the metapopulation facilitated population when fields are fallow, dispersal across open fields may persistence, despite increases in the frequency of drought. carry a higher predation risk for frogs than in landscapes Atmospheric research predicts a drying climate for the with more intact ground-cover. southwest of Western Australia (Pitman, Narisma, & Hol- The sensitivity analysis of juvenile and adult mortality brook, 2004). H. albopunctatus is however, a long-lived dry- supports the conclusion that mortality factors, either at a land frog that may not necessarily breed every year (Davis, breeding site or during dispersal may be the most pertinent 2004) and hence may be resilient to dry years. Conse- factor for H. albopunctatus persistence. Field studies of quently, we assumed drought did not have a catastrophic H. albopunctatus (Davis & Roberts, 2011) revealed gener- effect on metamorph survival. However, as the drying cli- ally good adult survival (0.34–1) although little information mate worsens, the predicted reductions in rainfall, runoff and was obtained on juvenile survival despite an intensive mark- associated water-holding capacity of ponds may result in recapture study of 500 metamorphs. Conroy and Brook widespread failure of recruitment for species such as this that (2003) also reported that the most sensitive life-history use temporary ponds. In such a scenario, the maintenance of parameter in a PVA of Geocrinia spp. was juvenile survival a well-connected metapopulation may be even more critical and Di Minin and Griffiths (2011) also found that the PVA to the species viability. Davis (2004) noted the potential of they constructed for Bufo calamita (Laurenti) was most sen- ecological traps resulting from the attraction of breeding sitive to changes in juvenile survival. Besides the direct adults to sub-surface water and sandy substrates such as in demographic importance of juveniles in becoming a repro- salinity interceptor banks and small pastoral soaks. Schlaep- ductively active adult, juveniles may also be an important fer, Runge, and Sherman (2002) define an ecological trap as dispersal stage of the lifecycle. Adult anurans are often less arising when organisms make poor habitat choices based on vagile than juveniles. For example, Breden (1987) found that cues that correlated formerly with habitat quality. In this juvenile dispersal was responsible for most of the gene flow case, frogs are choosing breeding sites based on cues (per- observed between populations of Bufo woodhousei fowleri haps moisture or the presence of a depression) but our data (Girard). show that these sites do not hold water long enough for tad- Data on juvenile dispersal were not available for our poles to develop and there can be a high rate of failure of PVA due to no recaptures of marked juveniles, indicating a 110 DAVIS ET AL. need for further attempts to quantify juvenile survival. able to persist even with 95% adult mortality. This supports Genetic analyses (Davis & Roberts, 2005a) indicated the many other studies that have determined the critical role presence of strong gene flow across the range of this species of dispersal and metapopulation formation in buffering and that even low levels of dispersal may be sufficient to against extinction, particularly in fragmented landscapes eliminate genetic subdivision in this species. Adult dispersal (Griffiths & Williams, 2000; Halley, Oldham, & Arntzen, capacity seems moderate to high, though Davis and Roberts 1996; Sjogren, 1991). (2011) detected little movement between adjacent popula- The life-history attributes of H. albopunctatus, high tions. Tadpole dispersal in times of flood may also be an fecundity, high adult longevity and low to moderate dis- important mechanism driving gene flow. Though this is persal contribute to a regional metapopulation which is somewhat speculative, many breeding sites are located along responsive to changes, but with a low-moderate chance of ancient paleodrainage channels and one of us (J.D.R.) has persistence over the long-term. To increase the accuracy and observed the flooding of these usually dry channels in very resolution of future analyses, further work is required to rare flood events. There are observations of tadpoles of investigate the survival of juveniles due to the importance of H. albopunctatus being found in rivers in metropolitan Perth this life-stage in regulating population dynamics. As a wide- when washed several 100 km along wheatbelt river systems spread species in a fragmented and rapidly drying landscape, after such flood events (A. Main, personal communication). there is some cause for optimism that similar species of The general literature estimate of juvenile survival used amphibian may have the adaptive capacity required to sur- in the PVA may be an overestimate due to the more ephem- vive drying climates, particularly if connectivity is main- eral and variable nature of H. albopunctatus ponds and the tained or enhanced. high rate of recruitment failure observed (Davis, 2004). Poor Our work highlights the importance of furthering our recruitment will have ongoing implications for populations, fundamental knowledge of amphibian species. While recent as adults do not reach reproductive maturity until 2 years of efforts have focused considerable resources on understand- age (Lee, 1967). Consequently successive years of recruit- ing and preventing the spread of chytrid fungus and deter- ment failure could result in potentially fatal outcomes for mining the impact of climate change on amphibians, sound local populations. Regional metapopulations could also be baseline population demographic data is essential to inform affected if there was a high degree of environmental correla- these processes and the complex models that are often used. tion between local populations and low rainfall was the pri- With baseline life history data lacking for most species of mary cause of recruitment failure. amphibian, it is critical that such studies are resourced and While the proportion of females in the breeding pool and continued. fecundity is highly variable from year to year, in response to rainfall (Davis, 2004; Davis & Roberts, 2005a), and changes in the proportion of females in the breeding pool had an ACKNOWLEDGMENTS impact on population sizes and growth rates under all R.D. wishes to acknowledge the private landholders who models (Table 3), it did not have a significant adverse effect granted access to their properties for this research. All field on metapopulation persistence. Similarly, a decrease in the sampling was undertaken by RD under an animal ethics per- frequency of high rainfall years in which reproduction and mit 99/008/E92 and wildlife licences NE002947 and persistence are favoured to only 1 year in 10, did not signifi- SF003756 issued by the Department of Conservation and cantly affect metapopulation persistence. These data are con- Land Management. Funding was provided by the University sistent with several null models of amphibian behaviour of Western Australia, Peter Rankin Trust fund for herpetol- (e.g., Alford & Richards, 1999) and similar studies (Berven, ogy and Australian Geographic. 1990; Pechmann et al., 1991) which have reported high levels of annual variation and multiple years of recruitment failure or reduced recruitment success, interspersed with REFERENCES occasional years of high rainfall and peak reproductive suc- Akçakaya, H. R., & Sjögren-Gulve, P. (2000). Population viability analyses in – cess (Conroy & Brook, 2003). conservation planning: An overview. Ecological Bulletins, 48,9 21. Alford, R. A., & Richards, S. J. (1999). Global amphibian declines: A problem The conservation and management implications of this in applied ecology. Annual Review of Ecology and Systematics, 30, study are that juvenile and adult survival is critical to the 133–165. long-term persistence of H. albopunctatus and this may be Bamford, M. J. (1992). The impact of fire and increasing time after fire upon Heleioporus eyrei, Limnodynastes dorsalis and Myobatrachus gouldii the case for many amphibians with similar life histories. Of (Anura: Leptodactylidae) in Banksia woodland near Perth, Western interest was the finding that a metapopulation comprising Australia. Wildlife Research, 19, 169–178. as few as five populations, was able to decrease the extinc- Beissinger, S. R. (1995). Modeling extinction in periodic environments: Ever- tion risk for H. albopunctatus. All sub-populations in our glades water levels and Snail Kite population viability. Ecological Applica- tions, 5, 618–631. analyses were not sensitive to most changes in parameters Berven, K. A. (1990). Factors affecting population fluctuations in larval and and because they were connected by dispersal, they were adult stages of the wood frog (Rana sylvatica). Ecology, 71, 1599–1608. DAVIS ET AL. 111

Breden, F. (1987). The effect of post-metamorphic dispersal on the population George, R., Clarke, J., & English, P. (2008). Modern and paleographic trends in genetic structure of Fowler's toad, Bufo woodhousei fowleri. Copeia, 1987, the salinisation of the Western Australian wheatbelt: A review. Australian 386–395. Journal of Soil Research, 46, 751–767. Brook, B. W., Cannon, J. R., Lacy, R. C., Mirande, C., & Frankham, R. (1999). Grant, E. H. C., Miller, D. A. W., Schmidt, B. R., Adams, M. J., Comparison of the population viability analysis packages GAPPS, INMAT, Amburgey, S. M., Chambert, T., … Muths, E. (2016). Quantitative evidence RAMAS and VORTEX for the whooping crane (Grus americana). Animal for the effects of multiple drivers on continental-scale amphibian declines. Conservation, 2,23–31. Scientific Reports, 6, 25625. Brook, B. W., Lim, L., Harden, R., & Frankham, R. (1997). How secure is the Greenwald, K. R. (2010). Genetic data in population viability analysis: Case Lord Howe Island Woodhen? A population viability analysis using Vortex. studies with ambystomid salamanders. Animal Conservation, 13, 115–122. Pacific Conservation Biology, 3, 125–133. Griffiths, R. A., & Williams, C. (2000). Modelling population dynamics of great Brook, B. W., O’Grady, J. J., Chapman, A. P., Burgman, M. A., crested newts (Triturus cristatus): A population viability analysis. Herpeto- Akçakaya, H., & Frankham, R. (2000). Predictive accuracy of population logical Journal, 10, 157–163. viability analysis in conservation biology. Nature, 404, 385–387. Halley, J. M., Oldham, R. S., & Arntzen, J. W. (1996). Predicting the persistence Burbidge, A. H., Rolfe, J. K., McKenzie, N. L., & Roberts, J. D. (2004). Biogeo- of amphibian populations with the help of a spatial model. Journal of graphic patterns in small ground-dwelling vertebrates of the Western Applied Ecology, 33, 455–470. Australian wheatbelt. In G. J. Keighery, S. A. Halse, M. S. Harvey, & Hamer, A. J., Heard, G. W., Urlus, J., Ricciardello, J., Schmidt, B., Quin, D., & N. L. McKenzie (Eds.), A biodiversity survey of the Western Australian agri- Steele, W. K. (2016). Manipulating wetland hydroperiod to improve occu- cultural zone. Records of the Western Australian Museum Supplement No 67 pancy rates by an endangered amphibian: Modelling management scenarios. – (pp. 139 202). Kewdale: Museum of Western Australia. Journal of Applied Ecology, 53, 1842–1851. Conroy, S. D. (2001). Population biology and reproductive ecology of Geocrinia Harrison, S. (1991). Local extinction in a metapopulation context: An empirical alba and G. vitellina, two threatened frogs from southwestern Australia. evaluation. Biological Journal of the Linnean Society, 42,73–88. (PhD thesis). University of Western Australia, Perth, Western Australia. Heard, G. W., McCarthy, M. A., Scroggie, M. P., Baumgartner, J. B., & Conroy, S. D., & Brook, B. W. (2003). Demographic sensitivity and persistence Parris, K. M. (2013). A Bayesian model of metapopulation viability, with of the threatened white- and orange-bellied frogs of Western Australia. Popu- application to an endangered amphibian. Diversity and Distributions, 19, lation Ecology, 45, 105–114. 555–566. Coulson, T., Mace, G. M., Hudson, E., & Possingham, H. (2001). The use and Heard, G. W., Scroggie, M. P., & Malone, B. S. (2012). The life history and abuse of population viability analysis. Trends in Ecology & Evolution, 16, decline of the threatened Australian frog, Litoria raniformis. Austral Ecol- 219–221. ogy, 37, 276–284. Crawford, J. A., Peterman, W. E., Kuhns, A. R., & Eggert, L. S. (2016). Altered Hero, J.-M., Morrison, C., Gillespie, G., Roberts, J. D., Newell, D., Meyer, E., functional connectivity and genetic diversity of a threatened salamander in … Shoo, L. (2006). Overview of the conservation status of Australian frogs. an agroecosystem. Landscape Ecology, 31, 2231–2244. Pacific Conservation Biology, 12, 315–320. Croteau, M. C., Davidson, M. A., Lean, D. R. S., & Trudeau, V. L. (2008). Hobbs, R. J. (1993). Effects of landscape fragmentation on ecosystem processes Global increases in ultraviolet B radiation: Potential impacts on amphibian in the Western Australian wheatbelt. Biological Conservation, 64, 193–201. development and metamorphosis. Physiological and Biochemical Zoology, Hughes, J. P., Guttorp, P., & Charles, S. P. (1999). A non-homogenous hidden 81, 743–761. Markov model for precipitation occurrence. Applied Statistics, 48,15–30. Cullen, L. E., & Grierson, P. F. (2009). Multi-decadal scale variability in Hughes, L. (2003). Climate change and Australia: Trends, projections, and autumn–winter rainfall in South-Western Australia since 1655 AD as recon- impacts. Austral Ecology, 28, 423–443. structed from tree rings of Callitris columellaris. Climate Dynamics, 33, IUCN. (2016). IUCN Red List version 2016-3. Table 4a. Retrieved from http:// 433–444. www.iucnredlist.org/about/summary-statistics#Tables_3_4 D’Amore, A., Hemingway, V., & Wasson, K. (2010). Do a threatened native Lacy, R. C. (1993). VORTEX: A computer simulation model for population via- amphibian and its invasive conger differ in response to human alteration of bility analysis. Wildlife Research, 20,45–65. the landscape? Biological Invasions, 12, 145–154. Lacy, R. C. (2000). Structure of the VORTEX simulation model for population Davis, R. A. (2004). Metapopulation structure of the western spotted frog viability analysis. Ecological Bulletins, 48, 191–203. Heleioporus albopunctatus in the fragmented landscape of the Western Lacy, R. C., & Pollak, J. P. (2014). Vortex: A stochastic simulation of the extinc- Australian wheatbelt. (PhD thesis). University of Western Australia, Perth, tion process. Version 10.0. Brookfield, IL: Chicago Zoological Society. Western Australia. Lampo, M., Celsa, S. J., Rodriguez-Contreras, A., Rojas-Runjaic, F., & Davis, R. A., & Roberts, J. D. (2005a). The effects of habitat fragmentation on the population genetic structure of the Western spotted frog, Heleioporus Garcia, C. Z. (2011). High turnover rates in remnant populations of the Har- albopunctatus (Anura: Myobatrachidae) in south-western Australia. lequin frog Atelopus cruciger (Bufonidae): Low risk of extinction? Biotro- – Australian Journal of Zoology, 53, 167–175. pica, 44, 420 426. Davis, R. A., & Roberts, J. D. (2005b). Embryonic survival and egg number in Lee, A. K. (1967). Studies in Australian amphibia II: Taxonomy, ecology and small and large populations of the frog Heleioporus albopunctatus in West- evolution of the genus Heleioporus Gray (Anura:Leptodactylidae). – ern Australia. Journal of Herpetology, 39, 133–138. Australian Journal of Zoology, 15, 367 439. Davis, R. A., & Roberts, J. D. (2011). Survival and population size of the frog Lindenmayer, D. B., & Lacy, R. C. (1995). Metapopulation viability of arboreal Heleioporus albopunctatus in a highly modified, agricultural landscape. marsupials in fragmented old-growth forests: Comparison among species. – Copeia, 2011, 423–429. Ecological Applications, 5, 183 199. Di Minin, E. D., & Griffiths, R. A. (2011). Viability analysis of a threatened Lips, K. R. (2016). Overview of chytrid emergence and impacts on amphibians. amphibian population: Modeling the past, present and future. Ecography, 34, Philosophical Transactions of the Royal Society of London B, 371, 162–169. 20150465. Doherty, T. S., Davis, R. A., van Etten, E. J. B., Algar, D., Collier, N., Lode, T., Holveck, M. J., Lesbarreres, D., & Pagano, A. (2004). Sex-biased pre- Dickman, C. R., … Robinson, S. (2015). A continental-scale analysis of feral dation by polecats influences the mating system of frogs. Proceedings of the cat diet in Australia. Journal of Biogeography, 42, 964–975. Royal Society of London B (Suppl., 271, S399–S401. Driscoll, D. A. (1999). Skeletochronological assessment of age structure and Martin, G. R., Twigg, L. E., & Robinson, D. J. (1996). Comparison of the diet of population stability for two threatened frog species. Australian Journal of feral cats from rural and pastoral Western Australia. Wildlife Research, 23, Ecology, 24, 182–189. 475–484. Estrada, A., Gratwicke, B., Benedetti, A., DellaTogna, G., Garrelle, D., Pechmann, J. H. K., Scott, D. E., Semlitsch, R. D., Caldwell, J. L., Vitt, L. J., & Griffith, E., , Miller, P. S. (2014). The golden frogs of Panama (Atelopus Gibbons, J. W. (1991). Declining amphibian populations: The problem of zeteki, A. varius): A conservation planning workshop. Final report. Apple separating human impacts from natural fluctuations. Science, 253, 892–895. Valley, MN: IUSN/SSC Conservation Breeding Specialist Group. Retrieved Pitman, A. J., Narisma, G. T., Pielke, R. A., Sr., & Holbrook, N. J. (2004). from http://www.cpsg.org/sites/cbsg.org/files/documents/PGF_Workshop Impact of land cover change on the climate of southwest Western Australia. FinalReport_22July2014.pdf Journal of Geophysical Research, 109, D18109. 112 DAVIS ET AL.

Read, J., & Bowen, Z. (2001). Population dynamics, diet and aspects of the biol- Stuart, S. N., Hoffman, M., Chanson, J. S., Cox, N. A., Berridge, R. J., ogy of feral cats and foxes in arid South Australia. Wildlife Research, 28, Ramani, P., & Young, B. E. (2008). Threatened Amphibians of the World. 195–203. Barcelona, Spain, Gland, Switzerland, Arlington, USA: Lynx Editions, Reed, D. H., O’Grady, J. J., Brook, B. W., Ballou, J. D., & Frankham, R. (2003). IUCN, Conservation International. Estimates of minimum viable population sizes for vertebrates and factors Thom, M. D. (1995). Age structure of breeding populations in four Western influencing those estimates. Biological Conservation, 113,23–34. Australian frog species: Heleioporus eyrei, H albopunctatus, Neobatrachus Roberts, J. D., Conroy, S., & Williams, K. (1999). Conservation status of frogs pelabatoides and N kunapalari. (Honours thesis). University of Western in Western Australia. In A. Campbell (Ed.), Declines and disappearances of Australia, Perth, Western Australia. Australian frogs (pp. 177–184). Canberra: Environment Australia. Towns, D. R., Parrish, G. R., & Westbrooke, I. (2003). Inferring vulnerability to Ryan, M. J., Tuttle, M. D., & Rand, A. S. (1982). Bat predation and sexual introduced predators without experimental demonstration: Case study of advertisement in a neotropical anuran. The American Naturalist, 119, Suter's skink in New Zealand. Conservation Biology, 17, 1361–1370. 136–139. Tyler, M. J., & Doughty, P. (2009). Field guide to frogs of Western Australia Saunders, D. A. (1989). Changes in the avifauna of a region, district and remnant (4th ed.). Perth, Western Australia: Western Australian Museum. as a result of fragmentation of native vegetation: The wheatbelt of Western Australia. A case study. Biological Conservation, 50,99–135. Saunders, D. A., Hobbs, R. J., & Arnold, G. W. (1993). The Kellerberrin project How to cite this article: Davis RA, Lohr CA, on fragmented landscapes: A review of current information. Biological Con- servation, 64, 185–192. Dale Roberts J. Frog survival and population via- Schlaepfer, M. A., Runge, M. C., & Sherman, P. W. (2002). Ecological and evo- bility in an agricultural landscape with a drying cli- lutionary traps. Trends in Ecology & Evolution, 17, 474–480. mate. Popul Ecol. 2019;61:102–112. https://doi.org/ Sjogren, P. (1991). Extinction and isolation gradients in metapopulations: The case of the pool frog (Rana lessonae). Biological Journal of the Linnean 10.1002/1438-390X.1001 Society, 42, 135–147.