Just 3 questions about this paper. they have to be comprehensive and interesting because I am going to have to discuss them in class.

Biological Conservation 160 (2013) 97–104

Contents lists available at SciVerse ScienceDirect

Biological Conservation

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b i o c o n

Predicting post-release establishment using data from multiple reintroductions

0006-3207/$ - see front matter � 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biocon.2013.01.013

⇑ Corresponding author. Tel.: +64 6 3569099; fax: +64 6 3505623. E-mail address: [email protected] (E.H. Parlato).

Elizabeth H. Parlato ⇑, Doug P. Armstrong Wildlife Ecology Group, Institute of Natural Resources, Massey University, Palmerston North, Private Bag 11 222, New Zealand

a r t i c l e i n f o

Article history: Received 22 October 2012 Received in revised form 17 January 2013 Accepted 20 January 2013 Available online 28 February 2013

Keywords: Establishment probability Return rate Bayesian modeling Post-release Survival Dispersal

a b s t r a c t

For any reintroduction it is important to maximise the probability of released individuals establishing in the target area (settling and surviving to breed). Factors influencing establishment have typically been studied at single sites, making it impossible to assess factors that vary at the site level (e.g. connectivity) or quantify unpredictable variation among sites. Using data from 14 reintroductions of the North Island robin (Petroica longipes) to native forest reserves, we show how Bayesian modelling can be used to iden- tify general drivers of establishment and to account for site-to-site variation when making predictions for new sites. High landscape connectivity and high rat tracking rates (a density index) at reintroduction sites were key factors associated with lower individual establishment probabilities. Habitat similarity between source and release sites was also important, as robins sourced from native forest had higher establishment than those from exotic pine forest. Previous predator experience appeared to affect estab- lishment in sites with mammalian predators, as founders sourced from sites with these predators had higher establishment than those from other sites. Our approach can be applied to a wide range of species that are being reintroduced to multiple sites, providing guidance on source and release site selection, effi- cacy of management interventions, and the numbers of individuals to release to achieve desired initial population sizes. The results are not only applicable to these particular species, but can be used to predict site suitability for reintroductions of species with similar dispersal behaviour or other ecological characteristics.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Reintroduction is increasingly used to re-establish populations of threatened species within their historical ranges (Sarrazin and Barbault, 1996; Seddon et al., 2007). However, many reintroduc- tion attempts are unsuccessful (Griffith et al., 1989; Sarrazin, 2007; Wolf et al., 1996) and the underlying causes of failure are rarely well understood (Dickens et al., 2010; Fischer and Linden- mayer, 2000; Letty et al., 2007). Analysis of factors influencing reintroduction outcomes is therefore important to improve the success of future reintroduction programmes (Ewen and Arm- strong, 2007; Le Gouar et al., 2012; Sarrazin and Barbault, 1996; Sutherland et al., 2010).

The two key phases affecting the dynamics of reintroduced pop- ulations are establishment and persistence (Armstrong and Sed- don, 2008). While the ultimate goal of any reintroduction is population persistence (Seddon, 1999), this is only achievable if the population survives the establishment phase. There is often elevated mortality (e.g. Calenge et al., 2005; Kreger et al., 2006) and dispersal (e.g. Moehrenschlager and Macdonald, 2003; Tweed

et al., 2003) immediately after release, meaning that reintroduc- tions can fail during the establishment phase even if conditions at the new site would enable persistence once established (Arm- strong and Seddon, 2008). Dispersal and mortality can have similar costs because individuals who disperse and settle away from the reintroduction area will not contribute demographically or genet- ically to the population (Le Gouar et al., 2012).

Because individuals are lost soon after release, the effective ini- tial population size, commonly defined as the number of individu- als that survive to the breeding season, is often much lower than the number of individuals released (Armstrong and Seddon, 2008; Armstrong and Wittmer, 2011). This in turn can exacerbate problems faced by small populations, including demographic sto- chasticity, environmental stochasticity, Allee effects and loss of heterozygosity. Maximising initial population size is therefore an important consideration for any reintroduction.

The most obvious approach to increase the initial population size is to release more individuals. The benefit of larger release groups is widely cited in the literature (e.g. Deredec and Cour- champ, 2007; Griffith et al., 1989; Wolf et al., 1998). However, releasing more individuals has a trade-off with impact on the source population (Armstrong and Wittmer, 2011) and can also have financial and logistical repercussions. There may also be a

98 E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104

trade-off at an individual and ethical level, as larger founder groups can result in more individuals being lost due to post-release dis- persal or mortality.

An alternative to releasing more individuals is taking measures to reduce post-release mortality or dispersal, thereby increasing the probability of founders settling in the reintroduction area. Pop- ulation establishment is dependent on the probability of reintro- duced individuals establishing at the new site, so understanding the key determinants of individual establishment is important for reintroduction success. Post-release survival and dispersal can be affected by various aspects of a reintroduction; including the trans- location process (e.g. release strategy, Devineau et al., 2011), char- acteristics of the individuals involved (e.g. age or sex, Masuda and Jamieson, 2012; Moehrenschlager and Macdonald, 2003), condi- tions at the reintroduction site (e.g. predator levels, Moorhouse et al., 2009), similarity between release and source sites (Lawrence and Kaye, 2011; Roe et al., 2010; Stamps and Swaisgood, 2007), and the habitat matrix surrounding the reintroduction site (La Morgia et al., 2011). Establishment of reintroduced individuals can therefore be facilitated at various levels; although the most appropriate and effective measures will depend on the species in question. For example, riparian brush rabbits (Sylvilagus bachmani riparius) held longer in enclosures before release had higher post- release survival (Hamilton et al., 2010), whereas delayed release of stitchbirds (Notiomystis cincta) lowered survival compared to birds released immediately (Castro et al., 1995).

Analysis of data collected after reintroduction can provide cru- cial information about factors affecting establishment of individu- als post-release. Importantly, modelled relationships can then be used to make predictions before new reintroductions take place, providing guidance to managers about site suitability and appro- priate measures to improve reintroduction success. However, iden- tification of factors influencing post-release establishment is often based on data from single sites (e.g. Bernardo et al., 2011; Jõgar and Moora, 2008; Roe et al., 2010; Tweed et al., 2003). While these studies can provide valuable insights for the site in question, fac- tors influencing success throughout a species’ range may not be apparent in results from a single site (Jachowski et al., 2011). Using data from reintroduction attempts at multiple sites provides more certainty that identified relationships are general (Johnson, 2002) and therefore applicable to other sites. Analyses of data from single sites are also limited to factors that can be manipulated within that site (for example, release techniques or supplementary feeding). Potentially more important factors, such as habitat quality or con- nectivity, only vary among sites so analysing data from multiple sites is necessary to evaluate their influence on reintroduction outcomes.

There are numerous examples where single species have been released into multiple sites for conservation purposes. In New Zea- land and Australia, more than 40 vertebrate species have each been translocated to at least five different sites (e.g. http://rsg-ocea- nia.squarespace.com/nz/; Short, 2009). In southern Africa, most large herbivores (e.g. Linklater et al., 2011; Van Houtan et al., 2009) and carnivores (e.g. Hayward et al., 2007) have been reintro- duced to multiple sites. There are also examples from other parts of the world, including Griffon vultures (Gyps fulvus) in France (Le Gouar et al., 2008) and black-footed ferrets (Mustela nigripes) in North America (Jachowski et al., 2011). These multiple releases create a unique opportunity to integrate data among sites to iden- tify the key influences on reintroduction outcomes, while also accounting for any unexplained site-to-site variation in population parameters. The results obtained would not only be applicable to the species that have already been reintroduced to multiple sites, but could be used to predict site suitability for reintroductions of species with similar dispersal behaviour or other ecological characteristics.

We present an approach whereby data from multiple reintro- duced populations are integrated into a Bayesian hierarchical mod- el to identify important factors influencing post-release establishment. We model establishment data for North Island rob- ins (Petroica longipes) reintroduced to 14 sites, and show how the resulting model can be used to make predictions for a candidate reintroduction site under alternative management scenarios. The strength of our approach is the ability to model the general influ- ences on establishment while accounting for site-to-site variation, thereby enhancing predictive capability and enabling targeted management to improve reintroduction success.

2. Methods

2.1. Species and reintroductions

The North Island robin is a small (26–32 g) insectivorous forest passerine endemic to New Zealand. The species was historically found over the entire North Island, but is now restricted to native forest remnants and exotic plantations in the central North Island, as well as some offshore islands (Higgins and Peter, 2002). Robins are susceptible to predation, primarily by exotic ship rats (Rattus rattus) (Brown, 1997; Powlesland et al., 1999), but also other exotic mammals such as stoats (Mustela erminea) and native avian preda- tors such as morepork owls (Ninox novaeseelandiae). Their breeding season is generally from early September to February, and juve- niles become sexually mature by the start of the breeding season after that in which they fledge.

North Island robins were reintroduced to 15 different sites (31– 1100 ha forested area) between 1997 and 2007 and analysable data were available for 14 of these (Table 1). Thirteen of the sites were on the North Island and two (Glenfern, Windy Hill) were on Great Barrier Island, a ca. 28,500 ha island off the north-east of the North Island. Reintroductions always occurred between March and August. Pre-release monitoring was conducted at all sites prior to reintroduction and no robins were found. Birds were caught from the wild and were released immediately on arrival at the re- lease site. Robins typically undergo a period of dispersal post-re- lease, and become sedentary once pairs and territories are established in the breeding season. All sites, including the pro- posed site, were managed to control exotic mammalian predators. At the time of reintroduction, two sites were fenced to exclude mammalian predators, which were eradicated after fencing, hence those species were expected to be absent. Another site was fenced but had openings for vehicle access, so mammalian predators re- mained present. All reintroductions were to areas of native forest, and birds could potentially disperse into unmanaged forest in the surrounding landscape. One site also had an exotic pine forest plantation within its boundary.

2.2. Data collection

We compiled data to assess the probability of released individ- uals establishing at each reintroduction site, where ‘‘establish- ment’’ is defined as surviving and remaining at the site until the start of the breeding season (late August). We specifically modelled return rates, which are the proportions of released individuals that remain at the site and are detected (Cam et al., 2005; Martin et al., 1995), as it was impossible to separately estimate establishment and detection probabilities from the data available for some sites. We included data on return rates from initial reintroduction at- tempts only, so any supplementary translocations in subsequent years were excluded from our analysis. All birds were individually colour banded prior to release, and data on the number of birds re- leased and post-release sightings of individuals were available

Table 1 Characteristics of 14 North Island robin reintroduction sites and one proposed reintroduction site.

Site Month and year reintroduced

Number of robins released

Forested predator- control area (ha)a

Peninsula Rat tracking rate (95% CI)b

Standardised habitat ratioc

Connectivity index

Mammalian predators present

Ark in the park April 2005 53 1100 No 0.05 (0.02–0.11) 0.34 97 Yes Boundary stream April 1998 28 800 No 0.01 (0–0.04) �0.48 68 Yes Bushy park August 2001 28 87 No 0.13 (0.01–0.52) �1.08 15 Yes Cape kidnappers May 2007 35 280 Yes 0.10 (0.07–0.15) �0.98 10 Yes Glenfern April 2005 27 230 Yes 0.16 (0.10–0.23) �0.51 54 Yes Hunua May 2001 30 600 No 0.34 (0.23–0.48) 1.79 99 Yes Kakepuku June 1999 30 198 No 0.68 (0.18–0.98) �1.33 33 Yes Paengaroa March 1999 40 101 No 0.27 (0.20–0.34) �0.32 50 Yes Tawharanui March 2007 25 240 Yes 0 (0–0) �0.60 17 No Trounson April 1997 21 445 No 0.01 (0–0.04) 0.15 71 Yes Waotu May 2001 30 31 No 0.54 (0.10–0.93) 1.55 40 Yes Wenderholm March 1999 21 60 Yes 0.37 (0.05–0.85) 0.93 25 Yes Windy hill April 2004 30 267 No 0.27 (0.20–0.35) 1.68 64 Yes Zealandia May 2001 40 225 No 0 (0–0) 0.08 91 No Pukaha (proposed) NA NA 942 No NA NA 60 Yes

a Area of forest managed to control exotic mammalian predators at time of reintroduction. b Tracking tunnel rates estimated from observed data. Imputed values are shown in italics for sites where data were missing (estimated from the modelled relationship

between return rates and tracking tunnel rates for the other sites). c Standardised (mean 0, variance 1) area of accessible forest habitat within 2 km of perimeter of reintroduction site divided by the forested predator-control area.

E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104 99

from site managers, reports, field notebooks or theses (Pattemore, 2003; Small, 2004). Nine of the sites were systematically searched in September using robin lure tapes at regular distances to identify individuals present at the start of the breeding season. Less tar- geted monitoring was undertaken at 5 sites, where field staff re- corded birds sighted as they carried out other work in the site. Intensity of post-release monitoring is likely to influence the prob- ability of detecting individuals that establish, so we took this into account in our analysis. We expected detection probability to be close to 1 at intensively monitored sites, meaning return rates are equivalent to establishment probabilities, and test this by esti- mating detection probabilities at sites where this is possible.

We also compiled data on variables that were potentially useful predictors of return rates based on our knowledge of the species. These fell into three main categories: (1) Reintroduction site char- acteristics, which included the size of forested predator control area, presence/absence of mammalian predators (ship rats and stoats), rat tracking rate (an index of rat density), and three land- scape variables potentially influencing robin emigration post-re- lease; (2) Translocation process, which included monitoring intensity (moderate or high, as described above) and time (number of months) from release to the start of the first breeding season; and (3) Source site characteristics, including forest type (exotic or native) and presence/absence of mammalian predators.

Rat tracking rates are used throughout New Zealand to monitor effectiveness of rat control. Rat tracking data were collected at 10 of the reintroduction sites. Usually 10–25 tracking tunnels were placed at 50 m intervals along transects (1–14 transects per site for sites with rats present). Tunnels were set by baiting them with peanut butter, and ink pads and paper were placed inside to record the prints of a rat if it moved through a tunnel. The papers were usually collected the next day but at two sites the papers were left out for more than one night. We used data collected between the date of reintroduction and the start of the breeding season at each site to estimate the rat tracking rate, which is the nightly probabil- ity of a rat moving through a tunnel. Tunnels were set 1–3 times at each of the sites with rats present over this timeframe.

The first landscape variable was a binary measure of whether sites were on a peninsula. We considered that peninsularity could be important for establishment of reintroduced birds, as peninsular sites were found to adversely affect apparent survival of juvenile North Island robins at 10 reintroduction sites, probably due to higher dispersal rates (Parlato and Armstrong, 2012). The second

landscape variable was a connectivity index based on maps of the land-cover within 2 km of site perimeters, which we manually digitised (5 m cell resolution) from aerial photographs and satellite imagery using ArcGIS 9.3 (ESRI, Redlands, California) and Imagine 9.2 (ERDAS, Atlanta, Georgia). Vegetation was classified as mature native forest, mature exotic forest, native/exotic shrubland or pas- ture/bareland. We assigned permeability values to each cell, reflecting the extent to which the different vegetation types facil- itated robin movements. Permeability values were based on an in- verse scale of the resistance values estimated for dispersing juvenile robins by Richard and Armstrong (2010). They found that resistance increased progressively from mature native forest to pine plantations, shrubland and pasture, and inferred that robins did not cross pasture gaps >110 m. Values of zero were therefore assigned to any woody vegetation that could only be reached by crossing >110 m of pasture. The connectivity index (C) for each site was calculated as:

C ¼ PNc

c¼1Pc 100ðNcÞ

where Pc is the permeability value of each cell and Nc is the total number of cells within 2 km of the site perimeter. The third land- scape variable was an alternative index of connectivity where we used the land-cover maps to calculate the area of mature forest within 2 km of site perimeters (again excluding areas only reach- able by crossing >110 m of pasture). We then calculated the ratio of forest area outside the site to the site’s forest area and standard- ised these ratios for the 14 sites. This standardised variable is here- after termed ‘‘habitat ratio’’.

2.3. Modelling

Data were analysed using generalised linear models (logit link function, binomial error distribution) fitted in WinBUGS 1.4 (Spie- gelhalter et al., 2003) using Markov Chain Monte Carlo (MCMC) methods. Alternative models were compared based on the Devi- ance Information Criterion (DIC). DIC is a Bayesian criterion for model comparison that can be interpreted similarly to AIC (Akaike’s Information Criterion). DIC will be approximately equal to AIC in models with negligible prior information (Spiegelhalter et al., 2002). All models had uninformative priors and were run with 2 chains for 110,000 samples, with the first 10,000 samples

100 E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104

discarded as burn-in. We visually checked for convergence using the Brooks–Gelman–Rubin and auto-correlation plots.

We initially created a full model that included an intercept and fixed effects of peninsula, site area (log transformed), mammalian predators at reintroduction site, time from release to breeding sea- son, monitoring intensity, and source site effects of forest type and presence of mammalian predators (which we applied only to rein- troduction sites with those predators present). We assessed whether DIC was reduced by sequentially substituting the penin- sula effect with connectivity (logit transformed), then substituting the connectivity and area effects with habitat ratio. After identify- ing the best landscape metric (that which provided the lowest DIC), we assessed whether DIC was further reduced by substituting rat tracking rate (logit transformed) for presence of mammalian predators at the reintroduction site. Effects identified as important from the best full model were then used to create a simplified model to estimate return rates at the 14 reintroduction sites and the proposed reintroduction site. We always included the effect of monitoring intensity to account for differential detection of established birds. We ran the simpler model with and without a random effect among sites to assess whether there were differ- ences in return rates among sites caused by random variation or unknown factors. We assumed the random effect was normally distributed.

To estimate the rat tracking rate (rat.nightly) for each site, we first sampled the number of tunnels tracked from a binomial distri- bution where the sample size was the total number of tunnels set. Because tunnels were set for more than one night at two sites, we used the modelled probability of a rat passing through a tunnel over t nights (rat.total) to estimate the nightly rat tracking rate (rat.nightly = 1-(1-rat.total)1/t). For the four sites where tracking tunnel data were missing, we imputed rat tracking rates based on the relationship with return rate modelled from the other data.

Return rates of reintroduced robins are the product of both indi- vidual establishment probabilities and detection probabilities (sensu Cam et al., 2005). We expected detection probabilities of established birds would be close to 1 (return rate � establishment probability) for sites with systematic searches post-release, as rob- ins are relatively easy to find due to their inquisitive and friendly nature (Armstrong, 2000). To check this assumption, we used the live recaptures model in MARK (White and Burnham, 1999) to ob- tain monthly survival and re-sighting estimates using individual encounter histories (Lebreton et al., 1992) of birds released into intensively monitored sites. Zealandia was excluded from this analysis as we did not have individual sightings data. The encoun- ter histories reflected four surveys, at the start (September) and toward the end (January) of the first two breeding seasons post-re-

Table 2 Comparison of establishment models fitted to data for North Island robins at 14 reintrodu

Modela

logit(r) = a + bclogit(C) + brtlogit(rat.nightly) + bpsmps � mpr + bfspi + bmii

logit(r) = a + bclogit(C) + brtlogit(rat.nightly) + bpsmps � mpr + bfspi + bmii + re

logit(r) = a + bclogit(C) + brtlogit(rat.nightly) + bpsmps � mpr + bfspi + bmii + baarea + btmth

logit(r) = a + bclogit(C) + bprmpr + bpsmps � mpr + bfspi + bmii + baarea + btmths

logit(r) = a + brratio + bprmpr + bpsmps � mpr + bfspi + bmii + btmths

logit(r) = a + bpenn + bprmpr + bpsmps � mpr + bfspi + bmii + baarea + btmths

a r, Return rate (probability of reintroduced individuals establishing and being detected index (C); brt, effect of nightly rat tracking rate (rat.nightly); bps, effect of mammalian pr present at source site, mps = 0 if not present at source site; mpr = 1 if present at reintrod source site (pi = 1 if pine forest, pi = 0 if native forest); bmi, effect of monitoring intensity time from release to start of breeding season (mths = number of months); bpr, effect of ma peninsula, n = 0 if peninsula); br, effect of standardised habitat ratio; re, random effect a

b Effective number of parameters (mean of the posterior deviance minus the mean of c Deviance Information Criterion, indicating the model’s level of support. d difference in DIC from that of the best model.

lease. We used the survival and re-sighting estimates to calculate the probability of an individual being detected in at least one sur- vey (p0), which is given by:

p0 ¼ 1 �ðð1 � p1Þð1 � s2p2Þð1 � s2 s3p3ÞÞ

where s2 and s3 are survival probabilities for the second and third intervals (first breeding season and subsequent non-breeding sea- son), and p1, p2, p3 are re-sighting probabilities for the first three surveys, respectively.

3. Results

Comparison of landscape variables gave strong support for con- nectivity as the best predictor of return rate, lowering DIC by 12.9 and 10.2 relative to the peninsula and habitat ratio effects, respec- tively (Table 2). The DIC was further reduced when rat tracking rate was substituted for mammalian predator presence at the rein- troduction site (DDIC = 9.8). Parameter estimates from the best full model (Table 3) suggest effects of connectivity and rat tracking rates at the reintroduction site, and forest type and mammalian predator presence at the source site, were all useful predictors of return rate (95% credible intervals did not incorporate zero). Including these four effects with monitoring intensity in a simpler model resulted in better predictive capability than the full model (DIC lowered by 1.6) (Table 2). Adding a random effect among sites did not further improve the model, and instead reduced model per- formance (DIC increased by 1.3).

Robin return rates were higher at sites with lower connectivity and rat tracking rates (Fig. 1, Table 3). Sourcing founders from na- tive forest with mammalian predators also resulted in higher re- turn rates than sourcing birds from a predator-free site (Fig. 1) or pine forest (Table 3). The effect of monitoring intensity was more ambiguous although, as expected, more intense monitoring was positively associated with return rate. Detection probabilities for established birds at intensively monitored sites were between 0.99 and 1 indicating that established robins had a high probability of being encountered. As such, we were able to estimate individual establishment probabilities from return rates modelled with high monitoring intensity. Including the effect of monitoring intensity also allowed us to account for lower detection of established indi- viduals at the five sites that were not intensively monitored, and directly compare modelled and observed return rates. In general, the model provided a very good fit to the data (Fig. 2).

Sites varied greatly in their connectivity to the surrounding landscape (Table 1), and this had a strong influence on estimated establishment probabilities. The two most connected sites, Ark in

ction sites.

pDb DICc DDICd

7.23 72.58 0.0 9.43 73.90 1.32

s 8.64 74.22 1.65 8.09 84.07 11.49 7.07 94.29 21.71 8.11 97.00 24.43

in the reintroduction site); a, intercept term for return rate; bc, effect of connectivity edator presence at source site if predators present at reintroduction site (mps = 1 if uction site, mpr = 0 if not present at reintroduction site); bfs, effect of forest type at (i = 0 if moderate intensity, i = 1 if high intensity); ba, effect of site area; bt, effect of mmalian predator presence at reintroduction site; bpen, peninsula effect (n = 1 if non- mong sites. the posterior distribution).

Table 3 Means and credible limits (CLs) for parameters in best full model and simplified model (Table 2) of return rates for North Island Robins at 14 reintroduction sites.

Full model Simplified model

Nodea Mean SD 2.5% CL Median 97.5% CL Mean SD 2.5% CL Median 97.5% CL

a 0.18 1.62 �2.93 0.15 3.43 �0.76 0.36 �1.52 �0.75 �0.09 bc �0.34 0.10 �0.55 �0.34 �0.14 �0.36 0.08 �0.52 �0.36 �0.22 brt �0.32 0.07 �0.48 �0.32 �0.19 �0.32 0.07 �0.47 �0.32 �0.20 bps 0.96 0.45 0.11 0.94 1.89 1.29 0.39 0.56 1.28 2.08 bfs �0.96 0.51 �2.01 �0.94 �0.01 �1.22 0.46 �2.15 �1.21 �0.35 bmi 0.47 0.44 �0.40 0.47 1.32 0.28 0.28 �0.27 0.28 0.84 ba �0.03 0.28 �0.59 �0.02 0.50 bt �0.21 0.15 �0.49 �0.21 0.09

a a, intercept term; bc, effect of connectivity (C); brt, effect of rat tracking rate (rat.nightly); bps, effect of mammalian predator presence at source site if predators present at reintroduction site; bfs, effect of forest type at source site (pi = 1 if pine forest, pi = 0 if native forest); bmi, effect of monitoring intensity (i = 0 if moderate intensity, i = 1 if high intensity); ba, effect of site area; bt, effect of time from release to start of breeding season.

(a)

(b)

Fig. 1. Modelled relationship between connectivity of reintroduction sites to surrounding forest and establishment probability (return rates with intensive post-release monitoring) for reintroduced North Island robins sourced from native forest with mammalian predators (a) or without mammalian predators (b). Black and grey lines are estimated probabilities at 5% and 25% rat tracking rates, respectively (dashed lines are 95% credible intervals).

E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104 101

the Park and Hunua, were estimated to have the lowest probabili- ties of establishment (0.32 (95% CI 0.21–0.45) and 0.35 (95% CI 0.21–0.50), respectively) whereas Tawharanui, a relatively isolated mammalian predator-free site, had the highest establishment probability (0.93, 95% CI 0.85–0.98). Nevertheless, the importance of rat tracking and source site characteristics was also apparent, with moderately connected Boundary Stream estimated to have similarly high establishment (0.90, 95% CI 0.80–0.97) to Tawhara- nui due to low rat tracking (1%, 95% CI 0–3%), and because birds were sourced from native forest with mammalian predators present.

Predictions for the proposed reintroduction site Pukaha were dependent on the level of predator control achieved and character- istics of the source site. Predicted establishment probabilities were similar if the founder population was sourced from pine forest with mammalian predators or native forest without mammalian preda- tors. Our model predicted that reintroduced individuals captured

from predator-free native forest would have 0.43 (95% CI 0.32– 0.54) probability of establishing at Pukaha if rat tracking rates were 25%, or 0.45 (95% CI 0.31–0.59) probability if sourced from pine forest. This probability increased to 0.58 (95% CI 0.48–0.68) (0.59 when sourced from pine forest) if rat tracking rates were reduced to 5%, which might be expected with high intensity predator con- trol. Sourcing founders from native forest with mammalian preda- tors maximised predicted establishment probabilities for any level of rat control. For instance, robins were estimated to have 0.73 (95% CI 0.60–0.84) or 0.83 (95% CI 0.72–0.92) probability of estab- lishing at Pukaha with 25% or 5% rat tracking, respectively (Fig. 2). These results in turn have implications for the initial population size at Pukaha. For example, if 40 robins were caught from mam- mal-free native forest and released into Pukaha when rat tracking rates were 25%, we would expect 17 (95% CI 13–22) to remain in the reintroduction area and survive to the start of the first breeding season. If the birds were instead sourced from native forest with

Fig. 2. Observed return rates (white bars) and modelled establishment probabilities (light grey bars) for North Island robins at 14 reintroduction sites. Diagonal hatching represents modelled return rates for robins reintroduced to sites without intensive post-release monitoring. Dark grey and medium grey bars show predicted establishment probabilities for a proposed reintroduction site (Pukaha) at 5% and 25% rat tracking rates, respectively, assuming founders sourced from native forest with mammalian predators present.

102 E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104

mammalian predators, this initial population size is expected to be 29 (95% CI 24–34) at 25% rat tracking or 33 (95% CI 29–37) at 5% rat tracking.

4. Discussion

Maximising the probability of reintroduced individuals remain- ing in the target area and surviving to breed is an important con- sideration for any reintroduction. Only by understanding the key drivers of post-release establishment can we hope to identify effec- tive management interventions and make useful predictions for fu- ture reintroductions. In recent years, there have been numerous calls for quantitative modelling to become part of reintroduction evaluations (Armstrong and Reynolds, 2012; Le Gouar et al., 2012; Osborne and Seddon, 2012; Seddon et al., 2007). The chal- lenge is that reintroductions, by their very nature, often involve small data sets from distinct locations, so data from individual reintroduction attempts may be inadequate for thorough evalua- tion of factors affecting reintroduction success across a species’ range (Jachowski et al., 2011). Species are increasingly being re- leased into multiple sites as part of recovery programmes, provid- ing a valuable opportunity to move beyond the inferences that can be derived from single-site studies. Our study demonstrates the benefits of integrating data from multiple reintroductions into a single model to identify important influences on establishment across sites. Our methods allowed us to make predictions of initial population size for a candidate reintroduction site while simulta- neously quantifying uncertainty in those predictions. This ap- proach also potentially allows unexplained variation among sites to be taken into account through inclusion of random effects, although no such variation was detected in this study.

Our results showed that landscape connectivity and rat tracking rates at the reintroduction site, and forest type and mammalian predator presence at the source site, were all important for post- release establishment of North Island robins. Lower establishment probabilities were associated with greater connectivity to sur-

rounding forest, probably due to differential dispersal of robins out of sites. Post-release dispersal is increasingly being recognised as an important influence on establishment of reintroduced popu- lations (Le Gouar et al., 2012); however, to date, there has been lit- tle consideration given to the biological implications of landscape structure in reintroduction biology (La Morgia et al., 2011). Identi- fication of release sites where the surrounding landscape is more likely to inhibit dispersal out of the target area is essential to avoid dispersal-related failure of reintroductions (Le Gouar et al., 2012). Nevertheless, the effects of dispersal are not always detrimental. For example, if a reintroduced population can maintain positive population growth despite emigration or the habitat matrix out- side the target area is of sufficient quality to allow dispersing indi- viduals to survive and successfully breed, then there are unlikely to be negative consequences (Le Gouar et al., 2012). Understanding the implications of post-release dispersal is clearly important for improving reintroduction success. Our model enables us to predict individual establishment probabilities for a target area in relation to the surrounding landscape, providing vital information for assessments of site suitability.

Rat tracking rate was negatively correlated with robin estab- lishment, indicating that improvements in predator control at re- lease sites will benefit reintroduced populations. This finding was somewhat unexpected, because although ship rats are known to prey on nesting female robins (Brown, 1998), survival of adult males and non-breeding females in established populations is not markedly affected by rat densities (Armstrong et al., 2006; Par- lato and Armstrong, 2012). Translocation-induced stress is proba- bly responsible for the relationship found here, given that reintroduced individuals are subject to a number of stressors as part of the translocation process (Dickens et al., 2009; Teixeira et al., 2007) and are especially vulnerable to predation immedi- ately after release into a new location (Letty et al., 2007). This vul- nerability can be particularly relevant if there are new predators at the release location (Dickens et al., 2010), and predator-naïve rob- ins could be more susceptible to predation if they are unable to recognise predators as a potential threat. Previous studies have

E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104 103

found that robins in mammal-free environments are less likely to recognise model mammalian predators than robins coexisting with such predators (Jamieson and Ludwig, 2012; Maloney and McLean, 1995), raising questions about the appropriateness of reintroduc- ing naïve individuals to sites where mammalian predators are present (Jamieson and Ludwig, 2012). Our results suggest predator experience may indeed be important for robin establishment in sites with mammalian predators, with founders sourced from sites with mammalian predators estimated to have a higher probability of establishing than individuals captured from mammal-free sites. The lower establishment rates associated with sourcing robins from exotic pine forest could also be linked to predation vulnera- bility, as rats are known to be generally less abundant in pine than native forest (King et al., 1996). In addition to predator exposure, prior habitat experience could influence post-release survival if robins sourced from pine forest were less able to forage effectively in unfamiliar native forest or were more stressed by release into native habitat than robins sourced from native forest; for example, leading to compromised foraging ability or predator evasion (Dick- ens et al., 2010). Robins sourced from pine forest may also have greater propensity to disperse out of reintroduction areas, as post-release movements of reintroduced individuals tend to be more extensive in unfamiliar release environments (Biggins et al., 2011; Roe et al., 2010). The importance of forest type and mamma- lian predator presence at the source site suggest choosing a source population from habitat that best matches the ecological charac- teristics of the reintroduction site can be important for reintroduc- tion success (Letty et al., 2007; Rittenhouse et al., 2008). However, these factors need to be weighed against impact on source popula- tions, as populations at mammal-free sites may be more resilient to harvesting than populations coexisting with mammalian predators.

The ultimate purpose of building models for reintroductions is to make predictions that can be used to inform management (Arm- strong and Reynolds, 2012). Our approach provides practical guid- ance for managers when determining appropriate management strategies. Model predictions for the proposed reintroduction site Pukaha indicated that sourcing robins from native forest with mammalian predators would attain the highest establishment rates for any level of rat control, giving a simple way to improve establishment probabilities. Predator control was also an impor- tant contributor to establishment (Fig. 2) with an on-going influ- ence on the long-term growth of the reintroduced population (Parlato and Armstrong, 2012). Estimated establishment probabil- ities can also guide decisions on the number of individuals to re- lease. Despite the common focus on release group size in the literature, the relationship between the number of individuals re- leased and initial population size at the first breeding season is of- ten unknown, making it difficult to determine how many individuals should be released to meet programme objectives. For example, Tracy et al. (2011) developed a useful framework for deciding how many individuals to release to maintain a desired level of genetic diversity, but a key assumption was how many founders would remain to contribute to the gene pool. Our meth- ods therefore move beyond the educated guesses about expected initial population size often necessary when planning reintroduc- tions, and provide a quantitative basis for management decisions.

Our approach can easily be extended to other species; incorpo- rating any factors considered potentially important for establish- ment (for example, age or body mass of founders). We do, however, caution against perfunctory inclusion of release group size as an explanatory variable due to potential confounds associ- ated with the implicit, though probably unrealistic, assumption that numbers of individuals are chosen at random with respect to the probability of success (Armstrong and Wittmer, 2011). Not- ing this, release group size can be an important determinant of

establishment in its own right; for example, when Allee effects pose a non-trivial threat to the survival of reintroduced individuals (Armstrong and Reynolds, 2012).

Factors influencing establishment of reintroduced populations are typically identified using data collected from single sites. Here we present an approach that integrates data from multiple reintro- ductions, providing confidence that identified relationships are general and allowing predictions to be made for a new population in a new situation. The resulting model gives useful guidance for managers at a number of levels, including source and release site selection, efficacy of management interventions and, ultimately, the number of individuals to release to achieve a desired initial population size. Nevertheless, taking steps to ensure successful establishment of a reintroduced population is only worthwhile if conditions at the release site are sufficient to allow long-term pop- ulation growth and persistence. Establishment models should therefore be considered one of a suite of tools for assessing project feasibility. With the value of modelling reintroduced populations becoming increasingly recognised, we expect to see greater emphasis on the development of quantitative models to inform management and guide future reintroductions. For species reintro- duced to multiple sites, integrated models provide an ideal oppor- tunity to develop understanding over time of the key drivers of reintroduction success.

5. Role of the funding source

Our research was funded by Massey University and Landcare Research. The funding sources were not involved in the study de- sign, in the collection, analysis or interpretation of data, in the writing of the report or in the decision to submit the article for publication.

Acknowledgements

We thank Ark in the Park, Tony Bouzaid, Natasha Coad, Raewyn Empson, Denise Fastier, Judy Gilbert, Janice and Laurie Hoverd, Tim Lovegrove, Terry O’Connor, Kevin Parsons, Gordon Stephenson, Tamsin Ward-Smith, and the Windy Hill field team for providing information on robin post-release monitoring, Peter Newsome and Janice Willoughby for assistance with geographic information systems, Terralink International Limited, New Zealand Department of Conservation, Auckland Regional Council and Hawkes Bay Regio- nal Council for aerial imagery, and Jay Gedir, Ian Jamieson, Kate Richardson, and two anonymous reviewers for comments on the manuscript.

References

Armstrong, D.P., 2000. Re-introductions of New Zealand robins: a key component of ecological restoration. Reintrod. News 19, 44–47.

Armstrong, D.P., Reynolds, M.H., 2012. Modelling reintroduced populations: the state of the art and future directions. In: Ewen, J.G., Armstrong, D.P., Parker, K.A., Seddon, P.J. (Eds.), Reintroduction Biology: Integrating Science and Management. Wiley-Blackwell, Oxford, pp. 165–222.

Armstrong, D.P., Seddon, P.J., 2008. Directions in reintroduction biology. Trends Ecol. Evol. 23, 20–25.

Armstrong, D.P., Wittmer, H.U., 2011. Incorporating Allee effects into reintroduction strategies. Ecol. Res. 26, 687–695.

Armstrong, D.P., Raeburn, E.H., Lewis, R.M., Ravine, D., 2006. Modeling vital rates of a reintroduced New Zealand robin population as a function of predator control. J. Wildlife Manage. 70, 1028–1036.

Bernardo, C.S.S., Lloyd, H., Olmos, F., Cancian, L.F., Galetti, M., 2011. Using post- release monitoring data to optimize avian reintroduction programs: A 2-year case study from the Brazilian Atlantic Rainforest. Anim. Conserv. 14, 676–686.

Biggins, D.E., Godbey, J.L., Horton, B.M., Livieri, T.M., 2011. Movements and survival of black-footed ferrets associated with an experimental translocation in South Dakota. J. Mammal. 92, 742–750.

Brown, K.P., 1997. Predation at nests of two New Zealand endemic passerines; implications for bird community restoration. Pac. Conser. Biol. 3, 91–98.

104 E.H. Parlato, D.P. Armstrong / Biological Conservation 160 (2013) 97–104

Brown, K.P., 1998. Identifying predators at nests of small birds in a New Zealand forest. Ibis 140, 274–279.

Calenge, C., Maillard, D., Invernia, N., Gaudin, J.C., 2005. Reintroduction of roe deer Capreolus capreolus into a Mediterranean habitat: female mortality and dispersion. Wildlife Biol. 11, 153–161.

Cam, E., Cooch, E.G., Monnat, J.Y., 2005. Earlier recruitment or earlier death? On the assumption of equal survival in recruitment studies. Ecol. Monogr. 75, 419–434.

Castro, I., Alley, J.C., Empson, R.A., Minot, E.O., 1995. Translocation of hihi or stitchbird Notiomystis cincta to Kapiti Island, New Zealand: transfer techniques and comparison of release strategies. Reintroduction Biology of Australian and New Zealand Fauna, pp. 113–120.

Deredec, A., Courchamp, F., 2007. Importance of the Allee effect for reintroductions. Ecoscience 14, 440–451.

Devineau, O., Shenk, T.M., Doherty Jr., P.F., White, R.H., Kahn, R.H., 2011. Assessing release protocols for Canada lynx reintroduction in colorado. J. Wildlife Manage. 75, 623–630.

Dickens, M.J., Delehanty, D.J., Romero, L.M., 2009. Stress and translocation: alterations in the stress physiology of translocated birds. Proc. R. Soc. B: Biol. Sci. 276, 2051–2056.

Dickens, M.J., Delehanty, D.J., Michael Romero, L., 2010. Stress: an inevitable component of animal translocation. Biol. Conserv. 143, 1329–1341.

Ewen, J.G., Armstrong, D.P., 2007. Strategic monitoring of reintroductions in ecological restoration programmes. Ecoscience 14, 401–409.

Fischer, J., Lindenmayer, D.B., 2000. An assessment of the published results of animal relocations. Biol. Conserv. 96, 1–11.

Griffith, B., Scott, J.M., Carpenter, J.W., Reed, C., 1989. Translocation as a species conservation tool – status and strategy. Science 245, 477–480.

Hamilton, L.P., Kelly, P.A., Williams, D.F., Kelt, D.A., Wittmer, H.U., 2010. Factors associated with survival of reintroduced riparian brush rabbits in California. Biol. Conserv. 143, 999–1007.

Hayward, M.W., Adendorff, J., O’Brien, J., Sholto-Douglas, A., Bissett, C., Moolman, L.C., Bean, P., Fogarty, A., Howarth, D., Slater, R., Kerley, G.I.H., 2007. The reintroduction of large carnivores to the Eastern Cape, South Africa: an assessment. ORYX 41, 205–214.

Higgins, P.J., Peter, J.M., 2002. Handbook of Australian, New Zealand and Antarctic birds. Volume 6: Pardalotes to Shrike-thrushes. Oxford University Press, Melbourne. <http://rsg-oceania.squarespace.com/nz/> (accessed 10.10.12).

Jachowski, D.S., Gitzen, R.A., Grenier, M.B., Holmes, B., Millspaugh, J.J., 2011. The importance of thinking big: large-scale prey conservation drives black-footed ferret reintroduction success. Biol. Conserv. 144, 1560–1566.

Jamieson, I.G., Ludwig, K., 2012. Rat-wise robins quickly lose fear of rats when introduced to a rat-free island. Anim. Behav. 84, 225–229.

Jõgar, Ü., Moora, M., 2008. Reintroduction of a rare plant (Gladiolus imbricatus) population to a river floodplain – how important is meadow management? Restor. Ecol. 16, 382–385.

Johnson, D.H., 2002. The importance of replication in wildlife research. J. Wildlife Manage. 66, 919–932.

King, C.M., Innes, J.G., Flux, M., Kimberley, M.O., Leathwick, J.R., Williams, D.S., 1996. Distribution and abundance of small mammals in relation to habitat in Pureora Forest Park. New Zeal. J. Ecol. 20, 215–240.

Kreger, M.D., Hatfield, J.S., Estevez, I., Gee, G.F., Clugston, D.A., 2006. Behavioral profiles of the captive juvenile whooping crane as an indicator of post-release survival. Zoo Biol. 25, 11–24.

La Morgia, V., Malenotti, E., Badino, G., Bona, F., 2011. Where do we go from here? Dispersal simulations shed light on the role of landscape structure in determining animal redistribution after reintroduction. Landscape Ecol. 26, 969–981.

Lawrence, B.A., Kaye, T.N., 2011. Reintroduction of Castilleja levisecta: effects of ecological similarity, source population genetics, and habitat quality. Restor. Ecol. 19, 166–176.

Le Gouar, P., Robert, A., Choisy, J.P., Henriquet, S., Lecuyer, P., Tessier, C., Sarrazin, F., 2008. Roles of survival and dispersal in reintroduction success of griffon vulture (Gyps fulvus). Ecol. Appl. 18, 859–872.

Le Gouar, P., Mihoub, J., Sarrazin, F., 2012. Dispersal and habitat selection: behavioural and spatial constraints for animal translocations. In: Ewen, J.G., Armstrong, D.P., Parker, K.A., Seddon, P.J. (Eds.), Reintroduction Biology: Integrating Science and Management. Wiley-Blackwell, Oxford, pp. 138–164.

Lebreton, J.D., Burnham, K.P., Clobert, J., Anderson, D.R., 1992. Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecol. Monogr. 62, 67–118.

Letty, J., Marchandeau, S., Aubineau, J., 2007. Problems encountered by individuals in animal translocations: lessons from field studies. Ecoscience 14, 420–431.

Linklater, W.L., Adcock, K., du Preez, P., Swaisgood, R.R., Law, P.R., Knight, M.H., Gedir, J.V., Kerley, G.I., 2011. Guidelines for large herbivore translocation simplified: black rhinoceros case study. J. Appl. Ecol. 48, 493–502.

Maloney, R.F., McLean, I.G., 1995. Historical and experimental learned predator recognition in free-living New Zealand robins. Anim. Behav. 50, 1193–1201.

Martin, T.E., Clobert, J., Anderson, D.R., 1995. Return rates in studies of life history evolution: are biases large? J. Appl. Stat. 22, 863–875.

Masuda, B.M., Jamieson, I.G., 2012. Age-specific differences in settlement rates of saddlebacks (Philesturnus carunculatus) reintroduced to a fenced mainland sanctuary. New Zeal. J. Ecol. 36, 123–130.

Moehrenschlager, A., Macdonald, D.W., 2003. Movement and survival parameters of translocated and resident swift foxes Vulpes velox. Anim. Conserv. 6, 199–206.

Moorhouse, T.P., Gelling, M., Macdonald, D.W., 2009. Effects of habitat quality upon reintroduction success in water voles: evidence from a replicated experiment. Biol. Conserv. 142, 53–60.

Osborne, P.E., Seddon, P.J., 2012. Selecting suitable habitats for reintroductions: variaton, change and the role of species distribution modelling. In: Ewen, J.G., Armstrong, D.P., Parker, K.A., Seddon, P.J. (Eds.), Reintroduction Biology: Integrating Science and Management. Wiley-Blackwell, Oxford.

Parlato, E.H., Armstrong, D.P., 2012. An integrated approach for predicting fates of reintroductions with demographic data from multiple populations. Conserv. Biol. 26, 97–106.

Pattemore, D., 2003. Translocations of North Island robins (Petroica australis longipes) to Mainland Reserves: Breeding Success, Dispersal and Habitat Preference. MSc Thesis. University of Auckland, Auckland, p. 128.

Powlesland, R.G., Knegtmans, J.W., Marshall, I.S.J., 1999. Costs and benefits of aerial 1080 possum control operations using carrot baits to North Island robins (Petroica australis longipes), Pureora Forest Park. New Zeal. J. Ecol. 23, 149–159.

Richard, Y., Armstrong, D.P., 2010. Cost distance modelling of landscape connectivity and gap-crossing ability using radio-tracking data. J. Appl. Ecol. 47, 603–610.

Rittenhouse, C.D., Millspaugh, J.J., Hubbard, M.W., Sheriff, S.L., Dijak, W.D., 2008. Resource selection by translocated three-toed box turtles in Missouri. J. Wildlife Manage. 72, 268–275.

Roe, J.H., Frank, M.R., Gibson, S.E., Attum, O., Kingsbury, B.A., 2010. No place like home: an experimental comparison of reintroduction strategies using snakes. J. Appl. Ecol. 47, 1253–1261.

Sarrazin, F., 2007. Introductory remarks: a demographic frame for reintroductions. Ecoscience 14, iii–v.

Sarrazin, F., Barbault, R., 1996. Reintroduction: challenges and lessons for basic ecology. Trends Ecol. Evol. 11, 474–478.

Seddon, P.J., 1999. Persistence without intervention: assessing success in wildlife reintroductions. Trends Ecol. Evol. 14, 503.

Seddon, P.J., Armstrong, D.P., Maloney, R.F., 2007. Developing the science of reintroduction biology. Conserv. Biol. 21, 303–312.

Short, J., 2009. The Characteristics and Success of Vertebrate Translocations within Australia: A Progress Report to Department of Agriculture, Fisheries and Forestry. pp. 95.

Small, D., 2004. Survival, Breeding Success and Habitat Selection of North Island Robins Petroica australis longipes Reintroduced to Karori Wildlife Sanctuary. MSc Thesis. Victoria University, Wellington, p. 216.

Spiegelhalter, D.J., Best, N.G., Carlin, B.P., Van Der Linde, A., 2002. Bayesian measures of model complexity and fit. J. R. Stat. Soc. Ser. B: Stat. Meth. 64, 583–616.

Spiegelhalter, D., Thomas, A., Best, N., Lunn, D., 2003. WinBUGS User Manual, Version 1.4.3. <http://www.mrc-bsu.cam.ac.uk/bugs>.

Stamps, J.A., Swaisgood, R.R., 2007. Someplace like home: experience, habitat selection and conservation biology. Appl. Anim. Behav. Sci. 102, 392–409.

Sutherland, W.J., Armstrong, D., Butchart, S.H.M., Earnhardt, J.M., Ewen, J., Jamieson, I., Jones, C.G., Lee, R., Newbery, P., Nichols, J.D., Parker, K.A., Sarrazin, F., Seddon, P.J., Shah, N., Tatayah, V., 2010. Standards for documenting and monitoring bird reintroduction projects. Conserv. Lett. 3, 229–235.

Teixeira, C.P., de Azevedo, C.S., Mendl, M., Cipreste, C.F., Young, R.J., 2007. Revisiting translocation and reintroduction programmes: the importance of considering stress. Anim. Behav. 73, 1–13.

Tracy, L.N., Wallis, G.P., Efford, M.G., Jamieson, I.G., 2011. Preserving genetic diversity in threatened species reintroductions: how many individuals should be released? Anim. Conserv. 14, 439–446.

Tweed, E.J., Foster, J.T., Woodworth, B.L., Oesterle, P., Kuehler, C., Lieberman, A.A., Powers, A.T., Whitaker, K., Monahan, W.B., Kellerman, J., Telfer, T., 2003. Survival, dispersal, and home-range establishment of reintroduced captive-bred puaiohi, Myadestes palmeri. Biol. Conserv. 111, 1–9.

Van Houtan, K.S., Halley, J.M., van Aarde, R., Pimm, S.L., 2009. Achieving success with small, translocated mammal populations. Conserv. Lett. 2, 254–262.

White, G.C., Burnham, K.P., 1999. Program MARK. Survival estimation from populations of marked animals. Bird Study 46, 120–138.

Wolf, C.M., Griffith, B., Reed, C., Temple, S.A., 1996. Avian and mammalian translocations: update and reanalysis of 1987 survey data. Conserv. Biol. 10, 1142–1154.

Wolf, C.M., Garland, T., Griffith, B., 1998. Predictors of avian and mammalian translocation success: reanalysis with phylogenetically independent contrasts. Biol. Conserv. 86, 243–255.