Identifying factors affecting captive breeding success in a critically endangered species

ABSTRACT

Captive breeding programs are an increasingly important tool for species’ conservation efforts, but not all species reproduce well in captivity. Identifying factors that affect the reproductive success of captive populations is crucial to improving the performance and management of conservation-breeding programs, both by providing individuals for release and informing decision making. We examined breeding records collected from the long-running conservation-breeding program for the critically endangered Orange-bellied Parrot Neophema chrysogaster over an 11-year period. We examined egg hatching rate, nestling survival rate, and offspring sex ratio in response to a wide range of variables related to characteristics of individual birds, breeding events, and the captive environment. The hatch rate of eggs was higher in first clutches compared to second clutches and was lower than the wild population. The survival rate of nestlings through to fledging was variable between years but became higher and more consistent over the last five years of the study period. Variation in brood sex ratio was not related to any of the potential explanatory variables that we examined. This is one of the first studies to examine reproductive data in a long-running conservation-breeding program and shows that many common metrics do not explain reproductive variation. Our approach provides a framework for managers to investigate factors affecting reproductive success in conservation breeding programs more broadly.

KEYWORDS:

• Animal husbandry
• conservation breeding
• hatching success
• nestling success
• Psittacidae
• zoo

Introduction

Species have been bred in captivity for many reasons (D’Elia 2010), but captive breeding programs are increasingly being implemented with the explicit goal of assisting the recovery of wild populations (Cohn 1988). Such programs might contribute to species’ recovery actions by breeding animals for release to supplement wild populations, facilitating education/outreach initiatives, providing opportunities for research, and/or maintaining insurance populations against extinction (Saint Jalme 2002). While captive breeding programs have undoubtedly saved some species from extinction (Butchart et al. 2006), these programs can be difficult to implement, time consuming, and resource intensive (D’Elia 2010; Hoffman et al. 2010; Collar and Butchart 2014). Furthermore, there is no guarantee of success, as many programs are limited by animals reproducing less successfully in captive environments compared to the wild (Lees and Wilcken 2009; Mason 2010; Farquharson et al. 2018).

Optimising the reproductive success of captive animals is crucial to maximise the number of individuals available for release, minimise stress to the animals, and utilise available resources most efficiently (Mason 2010; Canessa et al. 2016b; Tripovich et al. 2021). However, there is often considerable uncertainty about the factors that drive variation in reproductive success in captivity. There are several reasons for this, including data deficiencies that hinder statistical analysis (Conde et al. 2019), resource limitations to study the issue (Fa et al. 2014), and the difficulties of breeding some species in captivity (D’Elia 2010). Overcoming these limitations and optimising the reproductive success of captive animals based on robust statistical evidence is an elusive goal of many captive breeding programs.

Understanding the sources of variation in the breeding output of captive animals is important for optimising conditions to improve their reproductive success (Griffith et al. 2017). Intrinsic traits like age (Ricklefs et al. 2003), experience (Imlay et al. 2017), genetic heterozygosity (Rabier et al. 2021), and personality (McCowan et al. 2014) have all been linked to variation in reproductive success in captive animals. There is also strong evidence that a captive environment itself can influence reproductive output. Wild birds brought into captivity have shown immediate changes to incubation behaviours (Gilby et al. 2013) and critical hormonal responses (Dickens and Bentley 2014; Jensen et al. 2019). Environmental variables including enclosure size (Ali et al. 2016), outdoor access (Dickens and Bentley 2014; Griffith et al. 2017) and autonomy of mate choice (Massa et al. 1996) can all influence individual stress responses, activity levels, and/or social behaviours, which can affect reproductive outputs (Clubb et al. 2015).

While optimal reproductive success is often important to achieving conservation goals (Cohn 1988; Collar and Butchart 2014; Slade et al. 2014), disentangling variables contributing to reproductive variation in captive environments is difficult. Confounding variables and resource constraints can make designing experiments to understand variation challenging or impractical. Furthermore, conservation breeding programs usually involve small, and/or vulnerable populations (Martin et al. 2012), for which few data have been collected, inherently limiting the statistical rigour of quantitative analyses. For these reasons, correlational studies of accumulated records provide a starting point to narrow the range of variables that could be later investigated in small, targeted experiments. While correlational studies have limitations, they can help identify factors that contribute to variation in reproductive success for captive populations (Canessa et al. 2016a; Griffith et al. 2017).

We use a correlational approach to identify predictors of breeding productivity in the critically endangered Orange-bellied Parrot Neophema chrysogaster, making use of data collected from a long-running captive breeding program. The Orange-bellied Parrot is a small (~45 g) parrot endemic to south-eastern Australia that breeds in southwest Tasmania and migrates to the south-eastern coast of Australia during the austral winter (Brown and Wilson 1980; Higgins 1999). Despite decades of interventions, ongoing population decline has reduced the current breeding range to a single location and Orange-bellied Parrots remain critically endangered (DELWP 2016; Stojanovic et al. 2017; Geyle et al. 2018). The remaining wild population is closely monitored: each individual bird is uniquely identifiable via coloured leg bands, nesting boxes and supplemental food are provided, and prescribed burning is undertaken to improve habitat and foraging opportunities (Stojanovic et al. 2020).

As part of recovery efforts, a captive breed-for-release program to supplement the declining wild population has existed since 1986 (Smales et al. 2000; Morrison et al. 2020). The captive population of Orange-bellied Parrots is closely managed, and all breeding attempts are carefully monitored. Thus, detailed demographic information is available and comprehensive husbandry records are maintained (Morrison et al. 2020). These long-term and comprehensive records have allowed us to investigate the effects of a range of common metrics on reproductive outcomes. Whilst many conservation breeding programs suffer from inherently low population sizes or data deficiency, this is not the case for the Orange-bellied Parrot, making this an unusually informative model system.

We first compiled data on three components of Orange-bellied Parrot reproductive success (hatch rate, nestling survival rate, and offspring sex ratio) for all breeding attempts from two captive breeding institutions over eleven breeding seasons. We then investigated how these measures of reproductive success related to (i) intrinsic differences among individual parrots, (ii) characteristics of each breeding event, and (iii) variation in the captive environment. Finally, we discuss how our observations of the drivers of reproductive success in captive Orange-bellied Parrots shed light on the intricacies of captive breeding programs that aim to support threatened wild populations.

Methods

Background: captive breeding program

A captive breeding program for the Orange-bellied Parrot was established by the Tasmanian Government (NRE) in 1986 when ten parrots were collected from the wild (Smales et al. 2000; Morrison et al. 2020). Further collections were made in subsequent decades, and at present, the captive population comprises more than four hundred birds spread between five breeding institutions across Australia (BirdLife Australia 2018; Morrison et al. 2020; Pritchard et al. 2021). The Orange-bellied Parrot captive breeding program is managed through a studbook and mean kinship strategy to maximise genetic heterozygosity (Morrison et al. 2020). Breeding birds are paired in early spring (Aug-Oct), and usually comprise one male and one female housed together in a separate enclosure. Enclosure and nest box design vary slightly between institutions but are based on recommendations outlined in the species’ husbandry manual (Gowland and Everaardt 2015; DELWP 2016).

The intensive management of the captive Orange-bellied Parrot population means information is collected on all breeding attempts and routine husbandry actions. We compiled these data from the two largest and longest running captive breeding institutions (Zoos Victoria, ZV, and the Department of Natural Resources and Environment Tasmania, NRE; (Smales et al. 2000; BirdLife Australia 2018)) for every recorded Orange-bellied Parrot breeding attempt from the 2011/12–2021/22 breeding seasons. While we were not able to incorporate data from all five captive breeding institutions, these records can be considered a good representative sample as ZV and NRE have historically held the largest populations of captive Orange-bellied Parrots (BirdLife Australia 2018). We considered any birds paired by an institution for breeding as a breeding record.

Defining the response variables

We identified three response variables because they each represented a different component of reproductive output:

1. Hatch rate of eggs - Hatching failure in birds has been associated with a wide array of factors (Clubb et al. 2015) including; inbreeding depression (Jamieson 2011; Rabier et al. 2021), parental age (Ricklefs et al. 2003; Griffith et al. 2017), stress hormone levels (Khan et al. 2016), social environment (Bolund et al. 2010; Clark et al. 2012; Hemmings et al. 2012), lack of mate choice (Massa et al. 1996), and inadequate enclosure designs (Dickens and Bentley 2014; Malek and Haim 2019). We defined hatch rate as the proportion of eggs that hatched in each clutch.

2. Nestling survival rate - Nestling survival can be influenced by the parents (e.g. health (Sheridan et al. 2004), age (Khan et al. 2016), condition (Tollington et al. 2018), behaviour (Rabier et al. 2021), personality (McCowan et al. 2014), experience level (Imlay et al. 2017)), or factors related to that breeding attempt, such as timing within the season (Ortiz-Catedral and H. Brunton 2008) and clutch/brood size (Smith et al. 1989; Dijkstra et al. 1990; Saino et al. 2018). External factors relating to the environment including enclosure aspect (Griffith et al. 2017), nest box design (Larson et al. 2016) or level of disturbance (Butler and Dufty 2007) also have the potential to impact nestling survival. We defined nestling survival rate as the proportion of hatchlings in a brood that survived through to fledging.

3. Offspring sex ratio – Females of some bird species have some control over the sex ratio of their offspring, often in response to physical condition or environmental stressors (Trivers and Willard 1973; Heinsohn et al. 1997; Clout et al. 2002; Taylor and Parkin 2008; Bowers et al. 2017). Alternatively, skewed offspring sex ratios in a population can arise from sex-biased mortality of juveniles or nestlings (De Kogel 1997; Bowers et al. 2011; Heinsohn et al. 2021). We defined offspring sex ratio as the proportion of male fledglings in a brood out of the total number of fledglings of known sex.

Defining the predictor variables

Drawing upon some of the research on variables which can affect avian reproduction, we identified three main themes: individual-level variation (e.g. age (Ricklefs et al. 2003; Imlay et al. 2017); physical condition (Murphy 2007); provenance (Gilby et al. 2013); attributes of a particular breeding event (e.g. timing within season (Naef-Daenzer et al. 2001); brood size (Ortiz-Catedral and Brunton 2008)); and the environment (e.g. enclosure size (Ali et al. 2016) or disturbance levels (Butler and Dufty 2007; Dickens and Bentley 2014)). We summarised data from routine husbandry records for variables related to each theme which could influence captive Orange-bellied Parrot reproduction. We selected a total of 19 variables including: basic demographic information for each individual bird (e.g. age, cohort, body mass, and provenance); data for each breeding event (e.g. year, laying dates, clutch number, and clutch size); and information on the breeding environment (e.g. breeding institution, building, proximity to disturbance and aspect). A full list of predictor variables and corresponding definitions are provided in the Supplementary Material.

Statistical analysis

As not all variables of interest were available or applicable for all records, we first undertook exploratory analysis to evaluate if predictor variables that were incomplete over the full dataset were influential in affecting any of the three response variables. In this exploratory step, we fitted generalised linear mixed effect models with a binomial distribution and logit-link function using ‘glmmTMB’ v. 1.1.3 (Brooks et al. 2017) to subsets of the data where incomplete predictors were available to evaluate the effects on the three response variables. We fitted each predictor as a fixed effect and included both sire and dam ID as random effects in all models to account for repeat observations of the same individual. We compared these main effects models against a corresponding null model based on ∆AIC> 2 (Burnham and Anderson 2002). In all cases, the null model was best supported by the data, so we excluded all incomplete predictor variables from downstream analysis. We further refined the list of predictor variables by excluding the least informative correlated variables from the analysis. We evaluated pairwise correlation between variable using ‘GGally’ v 2.1.2 (Schloerke et al. 2021) and excluded variables with a strong correlation (Spearman’s coefficient> 0.8 or < -0.8). Correlations always involved year of breeding attempt, so we only retained year as a variable. We excluded two records of third clutches due to low sample size.

Using the refined list of predictors, we followed the same procedure for each of the three response variables. We fitted three sets of generalised linear mixed effect models (corresponding to hatch rate, nestling survival rate, and offspring sex ratio) with a beta-binomial (hatch rate) or binomial (survival rate, sex ratio) distribution and logit-link function. We fitted predictor variables as fixed effects, and again included both sire and dam ID as random effects in all models to account for repeated observations of the same individuals. We compared the main effects models for each predictor both against one another and a corresponding null model based on ΔAIC> 2. We excluded an interactive term of year*institution due to models failing to converge. We used the model select function in ‘MuMIn’ v. 1.46.0 (Barton 2020) to rank models, and evaluated model fit using ‘DHARMa’ v.0.4.5 (Hartig 2022). We estimated the effect sizes for different variables using ‘emmeans’ v. 1.7.5 (Lenth 2021) and visualised them using ‘ggplot’ v. 3.3.6 (Wickham 2016). All analysis was conducted in R v. 4.2.1 (R Core Team 2021).

Results

We compiled 722 breeding records of captive Orange-bellied Parrots between the 2011/12 and 2021/22 breeding seasons from two captive-breeding institutions. The data set included detailed information for 521 different individuals (272 dams and 249 sires). Most records (n = 628) involved a nesting attempt being initiated by the pair (defined by females laying > 1 egg), with an overall participation rate of 87.0% of birds that were paired for breeding. Females produced a mean of 1.22 clutches annually, with an average clutch size of 4.74 eggs (range 1–11). The overall hatching rate of eggs was 50.5%, with a mean brood size of 3.29 nestlings (range 1–7). Mean survival rate of nestlings through to fledging was 75.5%, with 1126 fledglings produced by these breeding institutions over the study period: 550 males, 526 females, and 50 of unknown sex; an overall sex ratio of 51.1% male.

We provide the top five models based on AIC values for each response variable in Table 1. For the hatch rate of eggs, we found equivalent support (ΔAIC< 2) for the model with the main effect of clutch number and a model that included an interactive effect of clutch number and laying date (Table 1). The most parsimonious model included the main effect of clutch number; the effect sizes and confidence intervals from this model are presented in Figure 1. First clutches had a higher estimated hatch rate (53.5% CI: 49.2–57.8%) compared to second clutches (28.2% CI: 22.2–35.0%).The hatching rate of first clutches was lower for females who went on to lay a second clutch (36.5% CI: 29.9–43.6%) compared to females that only laid one clutch in a season (60.0% CI: 55.4–64.4%).

Figure 1. Model estimates and 95% confidence limits for the preferred model showing the difference in hatch rates of Orange-bellied Parrot eggs laid in captivity between first (53.5% CI: 49.2–57.8%) and second clutches (28.2% CI: 22.2–35.0%).

Table 1. Model outputs for the top five generalised mixed effect models for each response variable: hatch rate of eggs, nestling survival rate, and offspring sex ratio. Sire and dam ID have been added as random effects to each model. Models are ranked by AIC values, with the preferred model highlighted in bold.

Response variable	Fixed Effect	df	Log-likelihood	AIC	∆AIC	Akaike weight
Hatch rate of eggs	Clutch number	5	−1018.931	2047.863	0.000	0.566
	Clutch number * Laying date	7	−1017.249	2048.498	0.635	0.412
	Laying date	5	−1022.199	2054.397	6.534	0.022
	Dam mass at breeding	5	−1038.288	2086.576	38.713	0.000
	Dam average adult mass	5	−1038.638	2087.277	39.414	0.000
Nestling survival rate	Year	13	−491.195	1008.390	0.000	1
	Sire provenance	4	−517.512	1043.025	34.635	0
	Dam mass at breeding	4	−518.741	1045.483	37.093	0
	Sire home institution	6	−516.891	1045.783	37.393	0
	Dam average adult mass	4	−519.064	1046.129	37.739	0
Offspring sex ratio	Null	3	−443.463	892.925	0.000	0.109
	Sire age	4	−442.555	893.109	0.184	0.099
	Laying date	4	−442.813	893.626	0.701	0.077
	Dam provenance	4	−442.813	893.626	0.701	0.077
	Difference in sire mass metrics (mass at breeding – average adult mass)	4	−443.104	894.208	1.283	0.057

The best supported model for nestling survival rate based on ΔAIC included the effect of year of breeding attempt; effect sizes and confidence intervals estimated from this model are presented in Figure 2. Nestling survival ranged from a low of 61.1% (CI:49.3–71.7%) in 2015 to a high of 91.8% (CI: 85.9–95.4%) in 2020 (full model estimates are presented in Supplementary Materials S2). For offspring sex ratio, the best supported model based on ΔAIC was the null model (Table 1) indicating that offspring sex ratios were independent of the predictor variables we considered.

Figure 2. Model estimates and 95% confidence limits for the preferred model showing the relationship between nestling survival rate and year of breeding attempt for Orange-bellied Parrots breeding in captivity between the 2011/12 and 2021/22 seasons.

Discussion

Captive breeding programs are a globally important tool for species recovery programs, but many species fail to breed successfully in captive environments (Lees and Wilcken 2009). We used the Orange-bellied Parrot captive breeding program as a model to evaluate factors contributing to reproductive success in captivity. The hatch rate of Orange-bellied Parrot eggs was higher in first clutches compared to second clutches (Figure 1), and the rate of nestling survival through to fledging varied between years (Figure 2). Offspring sex ratio was not explained by our predictor variables, however the sample sizes for this analysis are small and conclusions should be viewed cautiously. Taken together, our results indicate that many measurable aspects of captive husbandry and individual variation among animals are independent of observed variation in the reproductive success of captive Orange-bellied Parrots.

Second clutches had a lower hatch rate compared to first clutches. This could be a reflection of physiological stress on females who have already laid a first clutch that season (Saino et al. 2005; Norte et al. 2010; Travers et al. 2010), or it could be an artefact of management practices. Pairs of captive parrots are often encouraged to re-nest immediately if their first attempt fails (e.g. by keeping a pair housed together, providing new nesting box(es) or an altered diet) whereas successful pairs are sometimes prevented from laying a second clutch for management reasons (achieved by separating birds or removing access to nesting boxes). As a result of these practices, that second clutches could be disproportionately laid by less fecund pairs. However, we were unable to explicitly test this hypothesis due to confounding effects of management practices which varied between institutions and over time.

The mean hatching rate of eggs in this study was 50.5%, which is comparable to other data from captivity (48%; Penrose 2016) but substantially lower than reported hatch rates in the wild population (79.5%; Holdsworth 2006); 74.95%, (Troy and Lawrence 2022),However, these estimates might not be strictly comparable as second clutches are extremely rare in the wild population (Holdsworth 2006; Stojanovic et al. 2017), thus the inclusion of second clutches could be lowering the overall estimated hatch rate for the captive population. Nevertheless, even when second clutches are excluded, the hatch rate estimates from only first clutches laid in captivity (53.5%) are still much lower than the wild population.

Captive populations often suffer inbreeding depression or low genetic diversity, which can reduce fertility (Hemmings et al. 2012), but Morrison et al. (2020) demonstrated that wild- and captive-bred Orange-bellied Parrots have a similar level of genetic diversity. Thus, lowered hatching success of the captive population could be due to the captive environment itself rather than a trait of individual birds. Clubb et al. (2015) summarised causes of infertility in birds including medical/physical (e.g. inbreeding, age, obesity, disease, malnutrition), environmental (e.g. enclosure design, nest box design, lack of visual cues, disturbance, lack of environmental cues,) or behavioural factors (e.g. pair incompatibility, experience levels, inappropriate social interactions, breeding behaviours or imprinting). These factors could all partly explain our observations, but there are inadequate data to explore these possibilities within the scope of this study. Nevertheless, low hatching success is an area in need of further research to improve captive breeding efficacy of Orange-bellied Parrots.

The rate of nestling survival varied between years, and the higher rates of survival in the last five years matches estimates from the wild population (87.1–96.7%; Holdsworth 2006; Stojanovic et al. 2017; Troy and Lawrence 2022). The observed variation in nestling survival rates could be partially explained by stochastic events. For example, a disease outbreak at one breeding institution in 2016 prompted a review of biosecurity protocols (Stojanovic et al. 2017; BirdLife Australia 2018). Subsequent changes to biosecurity and husbandry practices could be contributing to the increase in nestling survival from 2017 onwards (Figure 2). However, this higher and more stable rate of nestling survival in recent years could also be a reflection of increased resources directed into the captive program (BirdLife Australia 2018). Unfortunately, factors such as disease outbreaks or changes to husbandry practices (e.g. altered diets) are all confounded with year in our study. Distinguishing between causal factors is often challenging in captive-breeding programs because entire small populations can experience the same variation at the same time. Furthermore, many programs have limited resources to implement experiments and are limited by small sample sizes. Nevertheless, some degree of experimental manipulation or critical review is important for evidence-based decision making (Canessa et al. 2019; Fogell et al. 2019).

Captive breeding programs are becoming an increasingly important strategy for conservation around the world. These programs represent a substantial investment of scare resources, and they often need to contend with small population sizes, lack of knowledge, and limited resources (Mccleery et al. 2007; Fa et al. 2011; Collar and Butchart 2014). The challenges we faced in this study are typical of captive breeding programs; limited data on specific areas of interest, correlations between variables, and confounding effects of year. However, we can confidently rule out some common intrinsic traits (e.g. mass, age, early life history) as significant drivers of variation in reproductive success for Orange-bellied Parrots in captivity. This is important as individuals that might have previously been excluded from breeding or release events (e.g. very young or old, over- or under-weight, hand-reared) could potentially be included, thus increasing the proportion of the population available for breeding or release. Additionally, our results on hatching rates highlight future areas to focus and direct research to increase reproductive output of the captive population. As many captive breeding programs aimed at species recovery do not have the time, resources, or population sizes for experiments, an approach like the one undertaken here could help programs quantify variation in reproductive success, highlight influential variables, and identify areas to direct further research to improve the efficacy of conservation efforts.

Acknowledgments

We would like to thank all staff at Zoos Victoria’s Healesville Sanctuary and the Tasmanian Government’s (NRE) Orange-bellied Parrot Tasmanian Program, specifically; Rachael Alderman, Paul Black, Clare Lawrence, Darren Page, Shannon Troy, and Lisa Vassos for making data available and answering questions. Thanks to Dr Teresa Neeman and the ANU BDSI for statistical advice and thanks to two anonymous reviewers whose thoughtful comments improved this manuscript. We acknowledge the traditional custodians of country upon which this research was conducted. Funding was provided to L.T.B. through an Australian Government Research Training Program Scholarship. D.S. received funding from the Australian Government Department of Agriculture, Water, and the Environment.

Disclosure statement

L.T.B. is currently employed with the Tasmanian Government (NRE) Orange-bellied Parrot Tasmanian Program. No other potential conflicts of interest were reported by the authors.

Data availability statement

Data were provided through research agreements with the Department of Natural Resources and Environment Tasmania Orange-bellied Parrot Tasmanian Program (https://nre.tas.gov.au/conservation/threatened-species-and-communities/lists-of-threatened-species/threatened-species-vertebrates/orange-bellied-parrot/the-obp-tasmanian-program), and Zoos Victoria (https://www.zoo.org.au/fighting-extinction/local-threatened-species/orange-bellied-parrot/) and are used with permission.

Supplementary data

Supplemental data for this article can be accessed at https://doi.org/10.1080/01584197.2023.2194541.