A Genetic Management Decision-Support Framework Implementable with Limited Genetic Data: A Case of a Widespread Arid-Adapted Bird in Fragmented Fire-Prone Ecosystem

1. Introduction

Global biodiversity is declining at unprecedented rates, driven primarily by habitat loss, fragmentation, climate change, altered disturbance regimes, and overexploitation (Díaz et al. 2019; IPBES 2019). Increasingly, this biodiversity crisis is being recognised not only as a loss of species, but also as a profound erosion of genetic diversity—the foundational level of biodiversity that underpins population viability and adaptive capacity (Sgrò, Lowe, and Hoffmann 2011; Mualim et al. 2026). Widespread declines in genetic diversity have been demonstrated across taxa, including in historically widespread and numerically common species, raising concerns that genetic erosion may be occurring well before demographic collapse becomes apparent (Shaw et al. 2025).

Habitat loss and fragmentation are among the major drivers of genetic diversity decline because they reduce landscape connectivity and restrict gene flow among populations, while also reducing carrying capacity and effective population size (Haddad et al. 2015; Radford et al. 2021). Small and isolated populations are particularly vulnerable to extinction due to the combined effects of environmental, demographic and genetic stochasticity (Frankham et al. 2017). Extreme environmental events, such as droughts and high-intensity wildfires, can cause rapid population declines, and reduced population size accelerates genetic drift, leading to the loss of genetic diversity and ability to adapt to environmental changes, increased population differentiation, increased inbreeding and associated loss of fitness (Frankham et al. 2017). In time, insufficient gene flow can lead to extinction of small populations (Saccheri et al. 1998).

Fire regimes are changing globally, with increases in fire frequency, extent, and severity documented in many fire-prone landscapes, resulting in longer ecosystem recovery times and greater ecological impacts across biomes (Bowman et al. 2020; Canadell et al. 2021; Lv et al. 2025; Biswas, Dowdy, and Chand 2026). In fragmented landscapes, where limited dispersal constrains demographic and genetic recovery between disturbance events, fires can disproportionately impact small and isolated populations, increasing the risk of local extinctions (Driscoll et al. 2021; Plumanns-Pouton et al. 2025). Maintaining or restoring habitat connectivity is increasingly recognised as a key strategy for conserving genetic diversity, facilitating gene flow, and reducing extinction risk in landscapes subject to frequent or large-scale disturbance (Haddad et al. 2015; Keeley et al. 2019). When restoring physical landscape connectivity is impractical, gene flow can be increased through human-assisted translocations (Weeks et al. 2015; Pavlova et al. 2025).

Development of practical and implementable genetic management strategies is often constrained by limited genetic and demographic data, particularly for less-studied species. However, an extensive body of empirical studies, theoretical work, and population viability modelling has demonstrated that, when combined with basic knowledge of species biology and life history, even relatively simple genetic metrics can provide sufficient information to guide management decisions (Frankham et al. 2017; Frankham et al. 2019). Decision-support frameworks that explicitly weigh the risks and benefits of gene flow from various sources while also considering risk of inaction (Frankham et al. 2011; Liddell, Sunnucks, and Cook 2021) allow practitioners to make informed decisions even when data are incomplete, by applying well-established principles and global data (e.g. concerning impacts of inbreeding, genetic drift, and outbreeding depression). As a result, genetic management can be proactively implemented to improve genetic health and enhance population persistence (Pavlova et al. 2014; Harrisson et al. 2016; Pavlova et al. 2017; Zilko et al. 2020), with positive outcomes emerging (Lutz et al. 2021; Pavlova et al. 2023). In this study, we synthetize a generalised, decision-support framework for genetic management under data-limited conditions, using an iconic Australian bird as a case example, providing a transferable protocol applicable to other species facing similar conservation challenges.

The malleefowl Leipoa ocellata, a long-lived mound-building Australian megapode occurring in fire-prone southern and central Australian arid and semi-arid woodlands and shrublands, provides a valuable model for examining the genetic consequences of habitat fragmentation and isolation in a fire-prone ecosystem. Known to many Indigenous communities as Ngaṉamara, Gnow, Lowan, Ngamarla, and other names (DCCEEW 2024), malleefowl is a culturally significant and often totemic species for Australian Traditional Owners, holding deep ecological, spiritual, and cultural importance across the mallee and southern desert bioregions (https://www.indigenousdesertalliance.com/what-we-do/desert-projects/significant-species/malleefowl, https://www.bushheritage.org.au/species/malleefowl). Once widespread across semi-arid mallee vegetation, malleefowl populations have declined greatly over the last century due to land clearing, altered fire regimes, predation, and habitat degradation, resulting in a highly fragmented distribution (DCCEEW 2024) (Figure 1).

Malleefowl is listed as Vulnerable under the Environment Protection and Biodiversity Conservation Act 1999 and under IUCN 2002 criteria (DCCEEW 2024). Species-wide population size has been estimated at about 25,000 breeding individuals (Benshemesh et al. 2021) based on breeding densities recorded in the National Malleefowl Monitoring Program (Benshemesh et al. 2018) and extrapolated across species range. Rapid declines in the last three generations have been inferred from habitat loss and recorded in monitoring programs (Benshemesh et al. 2020), with further declines highly likely given predictions of a drying climate and increased probability of more frequent large fires (Benshemesh et al. 2021; Stenhouse and Moseby 2022; DCCEEW 2024).

The distinctive life history of malleefowl shapes dispersal, gene flow, and population dynamics. Malleefowl are megapodes and breed in large incubation mounds that require near-daily maintenance to regulate egg temperature. Breeding adults are socially monogamous and males and females both participate in the construction and maintenance of their incubator mound, although during the egg-laying period (September to January) the sexes lead more separate lives as the female forages for egg production while the male regulates mound incubation temperature (Frith 1956; Frith 1959; Weathers, Seymour, and Baudinette 1993; Neilly et al. 2021). Usually, an active mound represents a single male-female pair, although a case where one male tended two mounds with different females has been recorded (Weathers, Weathers, and Seymour 1990). Reproductive output per female is typically 15–30 eggs per season (Frith 1959; Benshemesh 1992; Priddel and Wheeler 2005) in non-drought years with relatively high hatching success under favourable conditions. Chicks hatch fully independent and receive no parental care, dispersing immediately after emergence. Juveniles and adults rarely fly but are capable of movements over several kilometres by foot, with recorded daily dispersal events exceeding 10 km for adults (Stenhouse and Moseby 2022) and 6 km for newly hatched chicks (Benshemesh 1992); there is some evidence that dispersal may be female-biased (Stenhouse et al. 2022). Cleared land provides a barrier to dispersal (Stenhouse and Moseby 2022), and it is not uncommon for the species to persist in small (<500 ha) remnant patches of suitable habitat surrounded by cleared land where they are somewhat isolated from other populations. Large fires covering tens of thousands of hectares occur frequently in malleefowl habitats, threatening the species by reducing population size and rendering habitats unsuitable for breeding for 15-20 years (Benshemesh et al. 2021).

Despite its broad distribution, vulnerable conservation status and unusual life history, the malleefowl has not yet been subject to a published range-wide population genetic assessment. The only evidence for large-scale genetic structure comes from an unpublished mitochondrial ND2 study, which suggested historical isolation between eastern and western populations separated by the Flinders Ranges (Cope et al. 2014).

Historically high connectivity within each part of the malleefowl range likely facilitated extensive gene flow and accumulation of substantial genetic diversity prior to non-Indigenous human colonisation of Australia beginning ~200 years ago. Due to well-recognized time-lags between demographic change and genetic response, genetic erosion associated with recent habitat loss and fragmentation may not yet be detectable in malleefowl populations, as it is common in species with historically large effective population sizes and long generation time (Landguth et al. 2010; Tang et al. 2024; Mualim et al. 2026). Under such conditions, populations may retain high levels of neutral genetic diversity despite being on a trajectory toward future inbreeding and loss of adaptive potential.

Empirical evidence from the Eyre Peninsula (South Australia west of Flinders Ranges) population of malleefowl supports this interpretation. Using genome-wide SNP data, Stenhouse et al. (2022) detected only weak population genetic subdivision across a highly fragmented agricultural landscape, despite land-use changes, habitat configuration and limited dispersal and low contemporary gene flow through the agricultural matrix. The authors concluded that fragmentation has begun to disrupt connectivity but that substantial genetic erosion has likely not yet occurred, consistent with delayed genetic responses (Stenhouse et al. 2022).

Whereas Stenhouse et al. (2022) suggested that maintaining connectivity among multiple populations is critical, a recent simulation study (Tang et al. 2024) found that, through reductions in population size and increased genetic drift, habitat loss has a stronger negative effect on genetic diversity than does fragmentation per se. These findings led them to conclude that conserving fewer, larger populations may be optimal for maintaining genetic diversity. The differences between these perspectives become relevant under changing disturbance regimes, because large habitat fragments harbouring larger malleefowl populations are vulnerable to extreme, landscape-scale fires, increasing the risk of local or regional population collapse. Under such conditions, smaller, spatially separated populations may function as important reservoirs of genetic diversity, facilitating post-disturbance recovery through natural recolonisation where connectivity is maintained, or through assisted gene flow and translocations where it is not. Also, smaller populations were unexpectedly found to experience relatively high levels of breeding activity, possibly because they are biased towards better-quality habitat (Benshemesh et al. 2020). Thus, it is important to evaluate genetic health of the small, isolated populations and what level of natural gene flow the agricultural matrix around these populations can still support (Amos et al. 2014; Radford et al. 2021). Such data would inform genetic management strategies to prevent genetic problems in isolated populations and increase populations’ capacities to adapt to changing environments (Ralls et al. 2018; Frankham et al. 2019). Even where genetic signals of fragmentation are currently weak, proactive management to maintain or restore gene flow among populations might be essential to buffer against genetic drift and catastrophic demographic loss (Mualim et al. 2026).

In this study, we extend geographic sampling and apply population genetic analyses of genome-wide data to test for historical genetic isolation between western and eastern malleefowl populations. We test for decline in heterozygosity and increased relatedness in small, isolated habitat patches compared to moderately isolated and connected populations. Given the large size of historical malleefowl populations, we predict that the isolation of small populations facilitated genetic drift, expressed as increased population differences among locations, elevated pairwise kinships within locations, and reduced heterozygosity compared to larger and/or more connected populations. We co-analyse populations from Western Australia (WA) and Southern Australia (SA) west of the Eyre Peninsula, and populations from New South Wales (NSW) and Victoria (VIC) east of the Flinders Ranges, providing the first broad-scale assessment of historical genetic structure and connectivity across the species’ contemporary range. Finally, we synthesize a decision-support framework for genetic management of species with limited genetic data, and apply it to malleefowl to provide recommendations for species management for long-term persistence under climate change and altered disturbance regimes.

2. Materials & Methods

2.1. Sampling, Sites and Isolation Categories

Malleefowls’ low density, cryptic behaviour, and reliance on discrete nesting mounds, make individuals difficult to detect and sample. Samples for analysis were chosen from 1140 samples comprising shed feathers collected opportunistically by volunteers monitoring breeding density (83% of samples), membranes from hatched eggs collected during mound excavations and/or artificially incubated eggs (13%), and samples from malleefowl found dead from roadkill or other causes (4% of samples). From this assortment, we selected 282 samples for sequencing (see below), by attempting random selection while avoiding duplicate samples per individual as much as possible (e.g. selecting samples collected in the same season from disparate locations), and with preference to the most recent samples.

Populations (Table 1, Figure 1) were defined by geographic location and landscape connectivity. Population size was estimated from counts of active mounds at 201 sites across Australia from the National Malleefowl Monitoring Program. Across most of the species’ arid range, breeding densities are low (<1 pair per 100 km²) but the potential range is extensive. In more mesic areas, densities can reach 4 breeding pairs per km², although agricultural clearing has greatly reduced the amount of suitable habitat. As malleefowl breeding densities vary depending on environmental conditions, particularly rainfall, we chose the highest breeding density recorded during a five-year period (2015 to 2020) as indicative of the breeding population size at each site.

Isolation was classified from visual assessment of habitat continuity in Google Earth imagery. Because malleefowl are largely ground-dwelling and unlikely to disperse across cleared land (Stenhouse and Moseby 2022), cleared land was treated as a dispersal barrier. Populations were classified as highly isolated if movement to the nearest neighbouring population required crossing at least 10 km of cleared land, moderately isolated if habitat connections were tenuous or uncertain, and connected if populations were linked by continuous habitat.

For sequencing, we used 282 samples collected across 29 sites (hereafter, populations; Appendix A1; Figure 1). Samples included feathers collected from 2013 to 2024 (n=223), hatched egg membranes collected from 2021 to 2024 (n=45), tissue samples ranging from quite fresh to desiccated or degraded remnants (n=10), and buccal swabs (n=4). Fresh (wet) egg membrane samples were preserved in ethanol (n=14), whereas dry membrane samples (of unknown age, ranging from a few days to a few months) were kept in paper envelopes (n=31).

2.2. DNA Extraction, Sequencing, Genotyping and Calculating Autosomal Heterozygosity

Samples were sent to Diversity Arrays Technology (DArT) for DNA extraction, reduced-representation library preparation and sequencing. Raw Illumina sequencing reads were used for reference-genome genotyping (below).

We used the malleefowl draft reference genome (23,112 scaffolds) assembled by The Bird 10,000 Genomes (B10K) Project and made available by Guojie Zhang (Zhejiang University, China) through project website https://b10k.com/index/home/login.html for reference-based genotyping. To detect sex chromosomes in the reference genome for downstream analyses, scaffolds were mapped to chicken genome GCF_016699485.2, using minimap2 (Li 2018). Scaffolds that mapped to chicken chromosome W, Z, or both, were considered as belonging to sex chromosomes (hereafter W-scaffolds, Z-scaffolds, and gametolog-scaffolds, respectively). Repetitive regions were identified using RepeatModeller 2.0 (Flynn et al. 2020) and RepeatMasker 4.1.0 (Smit, Hubley, and Green 2013-2015). Raw DArT reads were mapped to the malleefowl draft reference genome using BWA-MEM 0.7.17 (Li and Durbin 2009; Li 2013). The resulting BAM files were sorted with SAMtools sort (Danecek et al. 2021). The sorted BAM files, the draft genome in fasta format, and a BED file with information on sex-chromosomes and repetitive regions were used as input for Module 1 of JeDi pipeline (Pavlova et al. 2025). This module was used to genotype samples with BCFtools (Danecek et al. 2021), filter out genotypes with fewer than 15 reads (minDP=15) and doubletons (rare alleles that appear in a homozygous state in a single individual per dataset) using VCFtools (Danecek et al. 2011), and calculate per-individual autosomal heterozygosity using piawka (https://github.com/novikovalab/piawka) after excluding sex-chromosomes and repetitive regions. JeDi scales the number of variable sites by the total number of called sites (i.e., variable and invariable ones), while considering 3- and 4-allelic loci, and accounting for missing data—all essential for unbiased estimates of heterozygosity (Sopniewski and Catullo 2024; Pavlova et al. 2025). Assuming DArTseq tags are a random representation of genome diversity, such estimates should be comparable across populations, time scales and species (Pavlova et al. 2025). We note that SNP datasets required as input for other population analyses (see below) excluded non-biallelic loci.

Poor-quality samples, such as old feathers or egg membranes, are expected to have high proportions of missing genotypes per individual, which may result in estimates of heterozygosity samples that could be biased in either direction (Schmidt et al. 2021). On one hand, degraded DNA can increase the prevalence of false alleles due to sequencing and amplification errors, potentially inflating heterozygosity estimates. On the other hand, reduced heterozygosity may arise from technical artefacts associated with low DNA concentration in the samples, which require a large number of PCR cycles to achieve sufficient DNA concentrations, in turn increasing the prevalence of PCR duplicates, leading to heterozygote under-calling. In addition, stringent call rate thresholds tend to disproportionately remove heterozygous sites hence underestimating heterozygosity (Sopniewski and Catullo 2024). To test for biases in our dataset, we plotted individual-based autosomal heterozygosity against proportion of missing loci per individual: reliable estimates of heterozygosity are expected to be independent of the amount of missing data per individual. We also tested whether proportion of missing data per individual is correlated with overall low read depth per sample.

2.3. Filtering of SNP Dataset

Iterative filtering was conducted following (Stenhouse et al. 2022). To generate our SNP datasets, non-biallelic sites, and sites scored for fewer than 20% of individuals were removed from the vcf file produced by JeDi using vcftools. The remaining analyses were performed in R 4.0.5 (R Core Team 2022) using RStudio v1.2.5042 (RStudio Team 2020). The filtered vcf file was imported to R and converted to a genlight object using library vcfR (Knaus and Grünwald 2017). Individuals missing >60% of genotyped loci and loci lacking scores in >60% of individuals were filtered out in R package dartR v2.0.4 (Mijangos et al. 2022). Further filtering was conducted in package wrapper SambaR (de Jong et al. 2021): loci lacking scores in >20% of individuals were removed; to reduce the number of physically linked SNPs, the dataset was thinned by retaining a single SNP per 50,000 bp. The resulting dataset of 4,753 autosomal SNPs scored for 112 individuals was used for kinship analyses. For analyses of population structure, individuals missing >50% of loci were further removed, as were three samples that were inferred a likely duplicate, or likely mislabelled samples (see below); this resulted in 94 unique individuals (Table 1).

2.4. Molecular Sexing

To determine the sex of each individual, we calculated the mean read depth across Z- and autosomal-scaffolds. The rationale is that females, being the heterogametic sex in birds (ZW), carry only one Z chromosome and should therefore exhibit approximately half the read depth on Z-scaffolds relative to autosomal scaffolds, while males (ZZ) should show comparable read depths across both. We used SAMtools bedcov (v1.9) to calculate the summed per-base read depth across Z-linked and autosomal scaffolds for each individual, then divided by total scaffold length to obtain mean read depth estimates. To control for variation in overall sequencing depth across individuals, we calculated a standardized Z-to-autosomal ratio by dividing Z-read depth by autosomal-read depth. This approach yielded the expected bimodal distribution (Appendix A2); individuals with a Z-to-autosomal ratio around 0.5 (i.e., < 0.65) were identified as females, while the remaining individuals with Z-to-autosomal ratio close to 1 (i.e., >0.70) were classified as males.

2.5. Tests of Sex Ratio Differences

We analysed male-to-female (M:F) sex ratio for all genotyped birds, for birds of different ages (chicks vs adults) and for birds sampled in populations with different level of fragmentation (high, moderate and connected). Age was defined based on sample type: chick carcass, fresh chick tissue or hatched egg fragments were considered chick samples; feathers and roadkill tissue were considered adult samples. We used binomial goodness-of-fit tests to test whether sex-ratios deviated significantly from 1:1 for all samples, for chicks only, adults only and adults from each level of fragmentation. We then used Fisher’s exact test to test for significant difference in sex-ratio between chicks and adults, and a chi-square contingency test to test whether adult sex-ratio was statistically different across fragmentation levels. We also used Cochran–Armitage trend test to test for statistical significance of directional change in sex-ratio across fragmentation levels.

2.6. Range-Wide Population Genetic Structure

We used Sambar to assess population structure across Australia, and separately within WA/SA and VIC/NSW populations. Two methods were used: (1) principal coordinates analysis (PCoA) based on Euclidean distances, which are less sensitive to heterozygosity biases and more reflective of recent drift compared to Nei’s distances, which are more reflective of historical divergence and hence were not used here, and (2) Admixture, implemented in the package Lea; both run using SambaR. We applied the cross−entropy criterion in admixture to indicate the most likely number of genetic clusters (Frichot and François 2015). However, variation in population size (i.e. strength of genetic drift) and time since divergence between populations at different hierarchical levels of structure can yield geographically meaningful results at different number of genetic clusters (K), so accordingly we explored other K-value that produced potentially biologically meaningful results.

2.7. Kinship Between Individuals in Habitat Patches with Different Degrees of Isolation

Increased kinship of individuals over time is a sign of increased inbreeding. To identify and remove identical genotypes from further analyses, and to identify highly related pairs of individuals within the same site, pairwise kinship was estimated using SambaR, and King-robust kinship (Manichaikul et al. 2010), designed to be robust to population structure and low-quality sequencing data. King-robust values were used to categorize degree of kinship between pairs of individuals: values that are negative or close to zero indicate unrelated individuals, 0.125- second degree kinship (half sibs or grandparent- grandoffspring), 0.25- first degree kinship (parent-offspring or full sibs), 0.5 represents monozygotic twins or two samples from the same individual. Due to missing data bringing down these benchmarks, we used kinship >0.3 to identify identical genotypes, following an earlier malleefowl study (Stenhouse et al. 2022).

In order to test whether isolated populations have higher values of mean pair-kinship compared to moderately isolated or connected populations, we calculated mean pair-kinship per population by averaging all pairwise values (including self-to-self kinships) within populations. We interpreted results only for populations with 4 or more samples because self-to-self kinship values tend to inflate mean pair-kinship values for populations with few samples.

To assess individual dispersal capacity, we aimed to calculate physical distances between first-degree kin pairs. Because all but one close-kin pairs were within-population (Appendix A3), the attempt is not reported in results. The only exception was a third-degree kinship pair of birds separated by 75 km of straight distance: JB0106 from connected hattah (4 on Figure 1) and JB0845 from nearby isolated cobram (12 of Figure 1).

2.8. A Decision-Support Framework for Genetic Management Recommendations

We synthesized a decision-support framework for genetic management underpinned by the content of two books on genetic management (Frankham et al. 2017; Frankham et al. 2019) (summarized in Table 2, with a full version in Appendix A4). This extends earlier approaches to decision-making applied to several endangered species (Pavlova et al. 2014; Harrisson et al. 2016; Pavlova et al. 2017; Zilko et al. 2020).

3. Results

3.1. Sample Quality

Of a total of 282 samples, 177 (62.8%) produced DNA that resulted in sequencing data. Failed samples included all cheek swab samples and tissue taken from desiccated or rotten carcasses transferred to 100% ethanol, 42% of dry egg membrane samples, 37% of dry feathers and 14% of ethanol-preserved egg membranes (Appendix A5). Sample type was a significant (p<0.05) predictor of missing data, with fresh tissues and egg membranes in ethanol shown to be the best types of samples, followed by dry feather samples (Appendix A5). The final dataset of 94 unique individuals (repeated and mislabelled individuals removed) missing <50% of loci contained 69 feather samples, 12 dry egg membrane samples, 10 membrane in ethanol samples, and 3 tissues in ethanol.

3.2. Mapping, Sexing, Filtering

A total of 454 scaffolds from the malleefowl draft reference genome (out of 23,112 scaffolds) mapped to the sex chromosomes of the chicken genome GCF_016699485.2 (357 to the Z chromosome, 49 to the W, and 48 to both). The remaining scaffolds (i.e., inferred to be autosomal) ranged in length from 162 to 12,625,250 bp (mean = 41,567.4 bp, median = 783.5 bp, N50 = 2,708,757 bp, L50 = 98). Of 177 genotyped individuals, 45 were inferred to be female, and 132 male.

A less-filtered dataset used for kinship analyses comprised 112 samples missing <60% genotypes (including 3 samples inferred to be repeated or mislabelled and later removed) from 27 sites (populations), of which 9 were isolated, 8 had some level of connectivity, and 10 were connected; Table 1). The dataset of 94 unique individuals (23 females, 71 males) with <50% missing data was sampled from 26 populations (Table 1). Only 11 populations in the larger dataset and 8 in smaller dataset had 5 or more samples.

3.3. Dependence of JeDi Autosomal Heterozygosity on Proportion of Missing Genotypes

Estimates of individual autosomal heterozygosity by JeDi (i.e., considering invariable and variable positions) ranged > 5-fold and were negatively correlated to the proportion of missing genotypes per individual (p<0.001, adjusted R-squared=0.519; Figure 2). Samples with higher proportions of missing genotypes also had higher mean and variance of read depths for the genotyped loci. High read depth coupled with high proportion of missing loci is consistent with the expectation of PCR duplicates (i.e. selective amplification of one of the alleles at a locus during library preparation of poor-quality samples) driving heterozygosity downward for some samples. In other cases, read depth remained steady or declined with increased proportion of missing loci, as would be expected from extremely low quantity of DNA template. Thus, read depth is not a reliable indicator of genotype quality, because while high-quality samples are expected to have high read-depth, some poor-quality samples can also yield in high read depth, together with selective allele amplification and/or allelic dropout, and large proportion missing loci. Only a few individuals appear to have heterozygosity that is independent of the proportion of missing data, suggesting that for the great majority of individuals in this dataset, heterozygosity could not be reliably interpreted. Because estimates of inbreeding are tightly linked to heterozygosity, inbreeding coefficients are also likely to be unreliable.

3.4. Population Structure

Our analysis of genetic structure in PCoA using range-wide samples (Figure 3) and VIC/NSW only samples (Appendix A6) separated three small and strongly isolated populations into their own PCA space. PC1 clustered eight individuals from the small and isolated VIC population wychi (30 on Figure 1) from the other individuals in the dataset. PC2 separated eight individuals from the small and isolated NSW population yalgogrin (31 on Figure 1), and to a smaller degree three individuals from the isolated NSW population tallimba (27 on Figure 1), from individuals sampled in all the States. Prominent separation of three isolated populations by PCoA analyses of genetic structure indicates that genetic drift has already strongly affected the most isolated populations. Separate PCoA analysis of western (WA/SA) populations did not detect any inter-population structure (PC1 separated three ravens individuals from the remaining individuals, Appendix A6), consistent with high effective population size and/or ongoing gene flow among the sampled western region.

The cross−entropy criterion for analysis of admixture on range-wide samples showed K=3 as the most likely number of genetic clusters (Figure 4; Appendix A7). The K=3 analysis was most consistent with the PCoA: it assigned the eight individuals from the small and isolated wychi population (VIC) >0.6 membership in their own genetic cluster (green on Figure 4), the eight individuals from the small and isolated yalgogrin population (NSW) >0.8 membership in the second cluster (blue on Figure 4), whereas the remaining individuals from WA, SA and VIC predominantly belonged to the third cluster (blue-green on Figure 4). Interestingly, three individuals–JB0029 from isolated tallimba (NSW; 27 on Figure 1), JB0106 from connected hattah (VIC; 4 on Figure 1), and JB0861 from connected cnsw (NSW; 2 on Figure 1)–shared >0.4 membership in the yalgogrin-dominated cluster (blue on Figure 4), suggesting they each might have had an immigrant ancestor from the yalgogrin population. Additional VIC individuals intermingled with SA/WA populations on the K=3 analysis, suggesting historical connectivity.

The analysis assuming five genetic clusters (K=5) separated SA/WA from both VIC and NSW, suggesting that contemporary connectivity between western and eastern populations might be limited. Given previous unpublished reports of mitochondrial (i.e. relatively old) sequence divergence between western (WA/SA) and eastern (NSW/VIC) populations, it appears that eastern and western populations with historically large effective population size (and weak genetic drift) still share much ancestral diversity, while drifting apart due to isolation-by distance with additional gene flow limitations due to habitat fragmentation. Meanwhile, the strongest evidence of structure (on lower K analyses) is observed for populations that experienced the strongest genetic drift: those that were isolated recently into extremely small habitat fragments that can support only a few breeding adults (Table 1). A similar pattern was observed for other species previously widespread and recently isolated into small populations of different sizes (Pavlova et al. 2017; Pavlova et al. 2025).

Separate Admixture analysis of eastern populations (NSW/VIC) run assuming five genetic clusters (K=5) identified four clusters of individuals (Appendix A8). In addition to previously identified yalgogrin and wychi, it separated two VIC populations: strongly isolated cassin (VIC; 11 on Figure 1) and moderately isolated mali (19 on Figure 1), although the latter appeared to be connected to, or share ancestral variation with, other VIC/NSW populations. Finally, separate Admixture analysis of western populations (WA and SA) did not detect any meaningful structure (not shown).

3.5. Kinship

King-robust scores ranged from -3.25 to 0.35, with the maximum value being for feather samples JB0639 and JB0589 collected from the same WA site (ravens) site in 2023, which we consider the same individual (Figure 5). Consequently, we excluded JB0589 from the mean pair-kinship calculations.

Kinship analysis detected 10 first-degree kin pairs (excluding two pairs involving JB0589), 18 second-degree kin pairs, and 15 third-degree kin pairs (Appendix A3). In all but three cases both members of a pair (i.e. relatives) were sampled from one of six populations: four strongly isolated ones–cassin (6 pairs, up to 3 relatives per individual), cobram (1 pair), wychi (15 pairs, up to 5 relatives per individual), yalgogrin (15 pairs, up to 4 relatives per individual), and two medium-isolated sites–mali (1 pair) and ravens (2 pairs). Four of these populations are in VIC, one (yalgogrin) in NSW, one (ravens) in WA.

Three pairs of relatives were sampled from different populations. One was a pair of third-degree relatives including a bird from connected hattah (JB0106; 4 on Figure 1) and a bird from isolated nearby cobram (JB0845; 12 of Figure 1). This indicates that either some level of cross-generational gene flow still occurs in this area, or–given the longevity of malleefowl–occurred prior to fragmentation. The other two pairs included birds sampled from different States (VIC and WA): a first-degree (JB0648-JB0628) and a third-degree kin pair (JB0266-JB0588). The VIC samples in each pair (JB0648 and JB0266) had a WA genotype on the Admixture K=5 analysis, thus they both likely represent a mislabelled sample. On this basis, JB0648 and JB0266 were removed from the mean pair-kinship calculations. None of the connected sites had close relatives sampled, despite some having up to 7 sampled individuals (Table 1).

Mean within-population pair-kinship supported elevated kinship in small and isolated populations compared to connected or moderately connected ones (Table 1). Two of the three isolated populations with n>4 (yalgogrin and wychi) had mean pair-kinship values >0.05; none of the four moderately isolated populations (bw, mali, nurcoung, ravens), or the four connected populations (sa, annuello, wyperfeld, LD) had a positive value (mean <0).

3.6. Sex Ratio Differences

Sex ratios were significantly male-biased (binomial test, p < 0.001) in the overall dataset and adults (M:F ratio=3.44), considered together, or grouped by different fragmentation levels (Appendix A9). Male-bias in chicks was not significant (p=0.09). Sex-bias of adults and chicks did not differ significantly (Fisher’s exact test p>0.05). Although adult male-bias tended to be stronger in isolated populations (M:F ratio=3.57) compared to moderately isolated (3.14) and connected (3.08) populations, the ratios did not differ significantly (chi-squared test p>0.05) and the trend was not statistically significant (Cochran–Armitage test, p >0.05).

3.7. Genetic Management Recommendations

Applying the decision-support framework (Table 2, Appendix A4) to malleefowl demonstrates how robust genetic management recommendations can be developed despite data limitations, as follows.

Step 1: Define conservation goals. All relevant stakeholders agree on the primary goal: maintaining viable, genetically healthy populations. Reconnecting habitat patches isolated by recent human activities is strongly supported as a means to achieve this goal, given evidence of elevated kinship, pronounced genetic drift, and fragmentation-driven isolation, all of which indicate increasing extinction risk if left unmanaged.

Step 2: Assess population distinctiveness. Genetic analyses suggest there are no distinct subspecies or races and none are currently recognised (DCCEEW 2024). Likewise, the genetic structure analyses in this paper (PCoA in Figure 3, Admixture analysis in Figure 4, kinship analyses) show only weak broad-scale differentiation between eastern and western regions, notwithstanding pronounced differentiation in small, recently isolated populations. This suggests that most populations are not deeply distinct and hence are appropriate candidates for managed gene flow. In contrast, differentiation in small fragments reflects drift in isolation rather than long-term adaptive divergence, which is a genetic problem to solve rather than a reason to manage populations separately (Liddell, Sunnucks, and Cook 2021).

Step 3: Assess reproductive biology & ploidy. Birds are overwhelmingly diploid. Although rare instances of triploidy have been documented in domestic or experimental contexts (Thorne, Collins, and Sheldon 1991), these are not evolutionarily persistent and have no practical relevance for conservation management. The biology of the malleefowl—largely monogamous, with female-biased dispersal over substantial distances—is suggestive of a largely outbreeding genetic system, consistent with modest subpopulation fixation indices (Stenhouse et al. 2022) within local groupings of individuals. As outbreeding diploid species, malleefowl would be expected to adhere to standard expectations that inbreeding and drift will rapidly reduce fitness in small populations, and populations would be expected to benefit from genetic rescue.

Step 4. Create strategy for implementation of genetic management

4.1 Diagnose genetic problems

4.1a Diagnose elevated inbreeding and reduced genetic diversity––In this study, direct estimates of heterozygosity and inbreeding coefficients were unreliable due to sample quality, but see 4.1b for comments on effective population sizes and inbreeding risk. Stenhouse et al. (2022) inferred the onset of loss of genetic variation in small, isolated populations of the species.

4.1b Diagnose altered connectivity––Previous work showed that malleefowl populations are somewhat genetically heterogeneous across the Eyre Peninsula, ascribed to impacts of habitat fragmentation driving differentiation by decreasing opportunities for gene flow (Stenhouse et al. 2022). Similarly, multiple lines of evidence in this paper—including strong PCoA separation particularly of small and isolated populations (Figure 3), elevated within-population kinship (>0.05) in some isolated populations, and clustering in admixture analyses (Figure 4.)—suggest reduced gene flow, small effective population sizes, and emerging inbreeding risk.

4.1c Diagnose reduced adaptive potential––The conclusions summarized under 4.1b indicate that malleefowl are likely to be subject to reduced adaptive potential. As a species occupying in fire-prone arid and semi-arid woodlands and shrublands in a continent experiencing substantial changes in aridity and temperature, and experiencing habitat fragmentation, malleefowl are expected to benefit from increased ability to adapt. Winter rainfall is a strong positive predictor of breeding activity, as is time since fire, but both are expected to reduce under climate change (Benshemesh et al. 2020). For example, winter rainfall decreased by 37% over 30 years in the sites studied by Benshemesh et al. (2020). Enhanced evolutionary potential through managed gene flow would be expected to be beneficial for such a situation.

Step 4.2: Scope source populations. Larger, connected populations in SA/WA and some VIC/NSW populations provide logical candidate sources of gene flow, as they show weaker structure, lower kinship, and signs of retained historical connectivity. If these sources were unavailable for practical reasons, there would still be benefit in translocations among differentiated small populations. Explicit testing of the impacts of sourcing immigrants on the donor populations would be ideal (Mitchell et al. 2022); otherwise, using only modest numbers of individuals and/or swapping individuals between source and recipient locations should have minimal impacts on the source, to be confirmed with monitoring.

Step 4.3: Assess risk of outbreeding depression, loss of local adaptation, or other negative consequences

4.3a Assess risk of outbreeding depression––The presence of admixed individuals and inferred recent migrants indicates that gene flow has occurred without apparent negative effects, suggesting low risk of outbreeding depression; the main genetic risk is therefore continued isolation rather than mixing. Low historical genetic structure across the species suggests that local adaption would be minor, likely eclipsed by the rate of environmental change ((Stenhouse et al. 2022); this study; Weeks, Stoklosa, and Hoffmann (2016)).

4.3b Assess risk of genetic swamping––Genetic swamping is a negligible concern for malleefowl because there are no known desirable local gene pools to maintain in isolation, and levels of gene flow can be controlled by the rate of movement of individuals between habitat remnants, coupled with monitoring of outcomes in the relevant locations.

Step 4.4 Evaluate benefit–risk

4.4a Benefit–risk of gene flow to recipient population––Given the strong signals of genetic drift and elevated kinship in isolated populations, and the absence of evidence for significant risks of gene flow, the expected benefits of genetic rescue (reduced inbreeding and increased genetic variation and adaptive potential) likely outweigh risks. The only PVA available for malleefowl (Bode and Brennan 2011) suggests that an isolated population of 32 adults would likely go extinct in 20-years without management actions due to high adult mortality. But this PVA does not incorporate genetic effect on fitness. The finding that many malleefowl populations are beginning to experience genetic problems but have not suffered considerable genetic erosion favours proactive action (Stenhouse et al. 2022). Furthermore, population trajectories for most malleefowl populations are downwards, despite considerable conservation efforts (Benshemesh et al. 2020), which would be expected to continue if genetic management is not undertaken. Malleefowl breeding activity decreased by several percent per year between 1989 and 2017 in South Australia and Western Australia, consistent with other studies (Benshemesh et al. 2020). Likewise, declines and some local extinctions have occurred in small habitat patches throughout the species range. Thus, the risk of inaction (not conducting genetic management interventions with a strong evidence-base of effectiveness) is high. In contrast, the risk of unfavourable outcomes from gene flow are low, as indicated in Step 4.3.

4.4b Cost-benefit of interventions––The proposed implementation could be readily incorporated into existing activities and would have minor cost implications: for example, the National Malleefowl Recovery Group coordinates the monitoring of malleefowl breeding at over 150 sites across the species range, largely through the efforts of citizen scientists in a national collaboration with land management agencies, and academic researchers (Benshemesh et al. 2020). Moving eggs among sites is the easiest, most efficient and least disruptive translocation strategy for malleefowl. It could be done at scale: a small number (~5) teams of 2-3 trained volunteers or Indigenous rangers under an oversight of a geneticist could feasibly translocate hundreds of eggs in a few days. The cost of such management action would be outweighed by improved population persistence through effective avoidance of inbreeding in isolated populations and augmented gene flow.

Step 4.5: Implement and monitor

4.5a Implementation strategy––We recommend implementing assisted gene flow for all populations with fewer than ~500 breeding pairs, with priority given to the smallest and most isolated populations that already exhibit genetic risk (e.g. elevated kinship or strong genetic drift). Management should prioritise restoring connectivity, either through restoring connectivity through habitat or assisted gene flow (e.g. translocations). An implementation plan should be developed collaboratively with all relevant stakeholders, including Indigenous groups and population genetic experts. Translocations could be integrated into the natural breeding season (November–March). The National Malleefowl Recovery Group routinely monitors over 500 active nesting mounds, each typically containing 8–12 viable eggs, providing a practical opportunity for large-scale interventions. Eggs can be carefully excavated and temporarily placed in controlled incubation conditions for transport to recipient sites. At these sites, eggs can be inserted into active mounds and incorporated into existing clutches for natural incubation. When correctly positioned, the addition of eggs is unlikely to disrupt incubation of existing eggs. Importantly, evidence suggests that such translocations may also confer ecological benefits. Where appropriate, a subset of eggs from donor mounds could simultaneously be redistributed among sites to maximise genetic mixing without compromising local recruitment. Following short transport times, eggs are returned to near-natural incubation conditions, allowing chicks to hatch and disperse naturally. Benshemesh et al. (2020) showed that some small, isolated populations are performing better demographically than larger populations, likely due to higher habitat quality. Translocating eggs into these sites may therefore increase recruitment by placing offspring into environments with higher survival prospects. Because malleefowl are long-lived and may breed for 10–20 years, translocations should be sustained over multiple breeding seasons (e.g. 5–10 years) to ensure that introduced individuals survive to maturity and contribute to the breeding population. The success of translocations can be evaluated through follow-up monitoring, including excavation of recipient mounds to assess hatching rates and short-term outcomes.

4.5b Monitoring and adaptive management––The National Malleefowl Recovery Group (NMRG), under the auspices of the national recovery team, is well placed to coordinate and implement a genetic monitoring program integrated within existing management frameworks (Hauser et al. 2019; DCCEEW 2024). High-quality sampling (blood on FTA cards, or ethanol-preserved tissue, blood, or fresh egg membrane) should be used to establish baseline genetic diversity and enable tracking of relatedness, connectivity, and effective population size. Interventions should follow an adaptive management approach, with key parameters—number of eggs moved, frequency of translocations, and choice of source and recipient populations—refined over time. Targets will vary among populations and should be determined collaboratively with managers, geneticists, ecologists, and Traditional Owners. As an initial guideline, reciprocal (two-way) egg translocations between large and isolated populations could be conducted annually for at least 5 years. Each exchange may involve approximately 10–12 eggs (i.e. a full mound), representing ~50% of seasonal reproductive output while remaining logistically feasible. Annual repetition is recommended because recruitment is variable and often low, meaning single interventions are unlikely to have lasting genetic effects. Translocations at this scale are operationally feasible (e.g. excavation and exchange of mound contents between paired sites within a day), and small populations of a few breeding pairs could be managed effectively. Genetic monitoring should occur 2–3 breeding seasons after translocated individuals reached maturity. Monitoring should also include demographic outcomes where possible to assess combined genetic and ecological benefits. Analyses should assess the success of translocations, modify the intensity and frequency of translocations, and assess which of the gene flow sources are more beneficial. This iterative process of intervention, monitoring, and adjustment will allow optimisation of gene flow strategies and improved conservation outcomes.

4. Discussion

In this study, we synthetized and applied a decision-support framework for genetic management for species persistence in fragmented landscapes, using malleefowl as an example case. We demonstrated how biologically meaningful patterns of connectivity/isolation and genetic health could be produced by a genetic dataset limited in sample sizes, sample quality and resulting genotypic data quality, to inform conservation genetic risk-assessment.

Despite historically large effective population sizes of malleefowl, and expectation of time lag between fragmentation and ability to detect the signals of genetic drift, we found a signal of strong population structure and increased relatedness for multiple small and isolated populations (e.g. wychi, yalgogrin, tallimba, cassin). This likely reflects lack of immigration and strong genetic drift due to extremely small effective population sizes of these populations. Elevated relatedness/kinship observed within these fragments is expected to increase the risk of harmful inbreeding depression and probability of population extinction, unless active translocations are commenced to augment gene flow into these small populations and increase effective population sizes (Frankham et al. 2017). Simulations studies show that a short-term effective population size (e.g. one approximated by the number of unrelated breeders in one generations) of at least 200 is required to prevent inbreeding depression from accumulating in a few generations. Given that 17 out of 26 populations assessed here had fewer than 100 breeding pairs, as indicated by attended mounds, and all isolated populations had fewer than 10 breeding pairs, these populations are likely to dwindle to extinction in few generations, even in benign environments. As climate change increases the frequency and severity of environmental disturbances, this trajectory will only accelerate. In long-lived species, highly inbred populations might linger–and even grow–until the most outbred individuals stop breeding or die (Taylor et al. 2017).

Predominance of close kin pairs in, and positive mean population kinship estimates for, small isolated populations of malleefowl provided evidence of small population size. In time, restricted dispersal might result in loss of genetic diversity by drift, which can lead to loss of populations’ abilities to adapt to future environmental changes. A population needs to have a long-term effective population size of 1000 to prevent loss of genetic diversity through drift and enable accumulation of new variation through mutations to fuel climate adaptation (Frankham, Bradshaw, and Brook 2014). In cases such as malleefowl with current limitations on habitat availability, it is recommended to manage fragments as a metapopulation, regularly reconnecting them by gene flow (Pavlova et al. 2025). Genetic monitoring should be conducted to optimize gene flow and ensure that each population actually attains target levels of new diversity (Pavlova et al. 2025).

Evidence of genetic diversity erosion and increased relatedness and inbreeding in small isolated or moderately-isolated malleefowl populations suggests that maintaining healthy insurance populations of malleefowl in small habitat fragments (e.g. to use as sources of gene flow to rehabilitate habitat destroyed by fires) will require frequent augmented gene flow among populations (e.g. via swapping). Limited carrying capacity in small patches means that some local eggs can be translocated to other populations, necessitating metapopulation-level genetic management.

At broader spatial scales, admixture analyses suggested differentiation between western (SA/WA) and eastern (NSW/VIC) populations, consistent with limited contemporary connectivity, and mitochondrial evidence of historical divergence (Cope et al. 2014). Nevertheless, this large-scale structure was relatively weak, compared to the pronounced differentiation observed among small, recently isolated populations. Better quality genomic data are required to test whether some gene flow still occurs at the intersection of lineages. Regardless of whether or not gene flow is occurring currently, the lack of strong large-scale structure suggests that the risk of outbreeding depression while cross-translocating individuals from eastern to western populations is low, and negligeable for mixing any populations within eastern or western groups. While carrying a low risk of outbreeding depression, reconnecting distant populations might bring stronger benefits than reconnecting nearby populations (Frankham et al. 2011).

Whole genome sequencing data from a few high-quality samples might allow testing for contemporary isolation between eastern and western populations, by providing mitogenome sequencing to confirm the degree of restriction of maternal gene flow (as mitogenomes are inherited almost exclusively from mothers in most animals), and compare it with nuclear gene flow (Morales et al. 2017). Nevertheless, in some species of bird with female-biased dispersal, nuclear gene flow may persist via male-mediated gene flow even when female gene flow is restricted by selection (Pavlova et al. 2013; Low et al. 2024).

The strong influence of sample type on genotyping success highlights the importance of carefully designed sampling strategies in wildlife genetics studies. While fresh tissue and egg membranes preserved in ethanol yielded higher-quality data, only ~34% of opportunistically-collected samples (e.g. feathers) produced usable genotypes. Importantly, we conclusively show that genetic diversity (and by inference, inbreeding) cannot be reliably estimated from standard genomewide sequencing of the degraded samples, even when applying sophisticated workflows designed to reduced biases in estimates (Pavlova et al. 2025). This limitation underscores the need for targeted sampling of high-quality biological material when establishing genetic baselines for long-term monitoring. Such material can include blood samples from chicks used for genetic augmentation or fresh egg membrane samples rinsed of debris, and stored in ethanol. Creating a targeted SNP-array can further cut sequencing costs (Helyar et al. 2011). Notably, kinship-based metrics proved useful even in the context of reduced data quality, suggesting that relatedness analyses may provide a particularly valuable tool for detecting early genetic risk.

Male bias was previously reported in the samples from SA malleefowl population (11 females to 25 males) (Stenhouse et al. 2022). Malleefowl feathers collected from active mounds (the main source of adult samples) were expected to be mostly from males. This is because during the egg-laying season, males spend most of their time near their mound, maintaining the incubation temperature and protecting the mound from weather and intruders. In contrast, females actively forage to produce large eggs, and may come to the mound only to lay, every 4-5 days.

We did not find statistical support for differences in sex ratio among fragmentation levels. However, the observation that more fragmented landscapes tended to have fewer females per male warrants further investigation. Feather samples were collected at active mounds attended by a single male and generally laid in by one female. A reduction of female presence at mounds in isolated habitats would be expected to result in fewer active mounds; however, excavations indicate that active mounds—even in isolated populations in VIC and NSW—consistently contain eggs. One possible explanation is that females spend less time at individual mounds in isolated areas, resulting in fewer detected female feathers. Alternatively, males in larger and more connected populations may have their mounds used by multiple females, whereas single females may predominate in more isolated patches. Female scarcity in isolated habitats could potentially facilitate polyandry if males are unable to effectively monopolise females during the egg-laying period (Armansin et al. 2020), although this has not been documented in malleefowl (but two active mounds with different females (although males may maintain two mounds with different females; Weathers, Weathers, and Seymour 1990). If these mechanisms operate, fewer eggs per mound would be expected in isolated populations relative to connected ones. Additionally, different patterns of relatedness might emerge, such as more paternal half-siblings within mounds in connected populations and more maternal half-siblings among neighbouring mounds in isolated populations. Further data are required to evaluate these hypotheses.

Stronger male-bias in isolated landscapes, if confirmed with additional sampling, would be consistent with elevated female-based mortality during dispersal into fragmented landscapes. Habitat isolation was inferred to disrupt dispersal of female brown treecreepers, resulting in many isolated habitat patches lacking females (Cooper, Walters, and Ford 2002). Under inbreeding, we could also expect higher mortality of the heterogametic sex (female birds) due to higher susceptibility to genetic problems (i.e., Haldane’s rule; Laurie 1997; Orr 1997). Stronger male-bias in isolated sites, consistent with female-biased dispersal being reduced by isolation, was observed in yellow-tufted honeyeater and eastern yellow robin (Harrisson et al. 2014; Radford et al. 2021). Higher female mortality due to inbreeding was suggested for the helmeted honeyeater (Pavlova et al. 2023).

Management Recommendations

Assisted gene flow through targeted translocations could help restore connectivity of malleefowl populations and reduce genetic drift, particularly where natural dispersal is constrained. We recommend that assisted gene flow be conducted into all populations smaller than 500 breeding pairs, with priority given to the smallest populations that already show signatures of genetic problems. At the same time, improving sampling strategies to prioritise high-quality genetic material will be essential for establishing baseline estimates of genetic diversity and monitoring future change. Genetic monitoring conducted annually after the first translocated individuals reach maturity, to a) assess the success of translocations and adjust methods if needed, b) modify the intensity and frequency of translocations, and c) assess which of the multiple gene flow options are more beneficial. Genetic parameters for monitoring success are described in Box 1 of Pavlova et al. (2025).

The approach for genetic management we applied here to the malleefowl should be useful for comparable systems with little genetic data available.