Response of a koala (Phascolarctos cinereus) source population following a large-scale progressive translocation harvest

Study area

The East Coomera study area is in the north-eastern section of the City of Gold Coast Council in Queensland, Australia. The area is bounded in the south by the Coomera River, in the west by the Pacific Motorway, rural farming lands in the north, and the southern section of Moreton Bay to the east (Fig. 1). The study area covered approximately 3640 ha and was estimated to contain 1716 ha of suitable koala habitat (Biolink 2007b). Most of the suitable habitat (1035 ha) was located within the planned urban area that was estimated to be supporting around 70% of the local koala population (Biolink 2007a). Development of the new Town Centre would involve clearing or modifying the majority of koala habitat in the urban area, leading to displacement of resident koalas to other locations, some of which were assessed as high-risk (e.g. Pacific Motorway, other major roads, railway line, existing residential areas).

Fig. 1.

East Coomera koala translocation study area.

Translocation background

To avoid forced dispersal of koalas from development sites and exposing them to increased risks of mortality, 166 koalas were progressively translocated away from the study site over a 5-year period from 2009 to 2013. The translocations away from the East Coomera source population totalled 18 koalas (two adult males (AM), eight adult females (AF), two subadults (SA), six juveniles (J)) in 2009, 45 koalas (7AM; 19AF, 6SA, 13J) in 2010, 17 (4AM; 6AF, 6SA, 1J) in 2011, 31 (8AM; 15AF, 5SA, 3J) in 2012, and 55 (15AM; 22AF, 14SA, 4J) in 2013. Because the study involved a mitigation-based translocation, it was necessary to recruit all koalas for translocation ahead of clearing on approved development sites. Hence, a diversity of age-sex classes and health profiles were included in the translocated group. As well as removing individuals from risk of immediate harm, we hoped that this would also reduce risks of overcrowding, which could theoretically increase susceptibility to stress-related illness and disease transmission (see Brearley et al. 2012; Mundey 2012; Carreira et al. 2020). Any koalas with clinical signs of disease or that tested positive for active chlamydial infection were admitted for hospital treatment prior to translocation. Because the source population is the focus of this component of the translocation study, the criteria for recipient site selection is not presented here.

Koala surveys

The field surveys were conducted in two periods as consultancies for the City of Gold Coast: (1) baseline surveys in 2006; and (2) resurveys in areas of remaining habitat in 2017 (Biolink 2007a, 2017). Over the period between survey events, approximately 750 ha of the available habitat had been cleared. Survey sites were located based on a grid positioned over the study area. Grid cells measured 350 × 350 m except in the narrow area located between the Pacific Motorway and the railway corridor. Grid cells there were 250 × 250 m to allow for adequate sampling. Survey sites were located where grid intersections coincided with areas of native vegetation containing eucalypts. Garmin GPS units were used to locate survey sites. Sites were sampled using the Spot Assessment Technique (SAT) of Phillips and Callaghan (2011). This involves searching a 1-m catchment around the base of 30 nearest trees (≥10 cm diameter at breast height (dbh)) at each site for koala scats. A direct count of koalas was also conducted by at least two experienced observers searching every tree within a 25-m radius of the SAT centre tree (i.e. covering a search area of 0.196 ha). Only independent-aged individuals were entered for estimation of koala density values. Details including sex, assumed reproductive status (females only), and any overt signs of illness (e.g. conjunctivitis, wet-stained rump, clumped fur, emaciated body condition) were recorded for each koala observation based on inspection through binoculars.

Analysis of tree species usage data

The scat surveys provide evidence of tree species use (i.e. feeding and shelter) by koalas in both survey periods (2006 and 2017). Results from other studies verify that Eucalyptus species comprise the majority of browse consumed by koalas (e.g. Moore and Foley 2000; Sullivan et al. 2003; Smith 2004; Tucker et al. 2007). Hence, we divided the eucalypt data into two groups: (1) the preferred suite of species; and (2) other eucalypts. City-wide koala habitat assessment surveys undertaken in 2006 involved sampling 11,190 trees (including 26 species of Eucalyptus) across 32 different mapped Regional Ecosystem types (Biolink 2007b). The statistically homogeneous suite of most preferred eucalypt species for koalas throughout the City of Gold Coast was identified as forest red gum (Eucalyptus tereticornis), tallowwood (Eucalyptus microcorys), grey gum (Eucalyptus propinqua/biturbinata), and swamp mahogany (Eucalyptus robusta) (Biolink 2007b). For this study, we were particularly interested in discovering if there was any evidence of shifts in tree species preferences or levels of use that might be due to changes in koala density or possible over-crowding. A relevant example was reported by Martin (1985) from an investigation of koala food trees and defoliation impacts in reasonably isolated habitat at South Gippsland, Victoria.

Analysis of koala activity and koala density

The distribution of koala activity was established to help identify focal areas for the local population and for comparison between the two study periods. In accord with the SAT sampling protocols of Phillips and Callaghan (2011) koala activity levels for each site were estimated by dividing the total number of trees that had scats recorded within the search area around their base, by the total number of trees sampled (typically 30). For both survey events, koala density was estimated by dividing the total number of adult koalas sighted in all 25-m (0.196 ha) radial searches by the total area sampled using this approach. The density estimates for the two survey events (i.e. 2006 and 2017) were compared by selecting with replacement 20 subsets of 50 sites from the 101 sites surveyed in both 2006 and 2017, with 10 subsets allocated to each survey event. Mean density estimates were then calculated for each of the survey events prior to testing for homogeneity of variance (F-test) and comparison using the appropriate t-test.

Estimation of population size

The mean density estimates derived from the radial searches were multiplied by the total area of potential koala habitat derived for 2006 and 2017, respectively, by digitising remaining patches of mixed forest containing eucalypt species from recent air photo imagery. The available habitat at the study site consists of a mosaic of coastal forest communities (i.e. Regional Ecosystems as identified by the Queensland Herbarium) dominated by eucalypt and non-eucalypt canopy tree species, including the local preferred koala food and shelter resources. This approach is comparable to that reported by Dique et al. (2004) for an adjoining area in south-east Queensland. The koala density estimates derived by Dique et al. (2004) incorporated a koala sightability correction factor of 1.2. This was not applied in our study because we expect location-specific differences in sightability. Dique et al. (2001) reported that missing 20% of the koalas present produced only a small increase in the standard deviation of density and was little influenced by koala distribution. We have assumed that the percentage of koalas detected was equivalent in our two sample periods. The large numbers of sites that we sampled enabled variation in density to be estimated.

Activity modelling

Koala activity levels for each SAT site were used as a basis for ArcGIS spatial analysis involving a combination of regularised splining using the default parameters to interpolate koala activity across the study area (Biolink 2007a; 2017), followed by contouring. This produces a koala activity contour map that theoretically delineates areas supporting established resident koalas by identifying significant activity values of 22% and 32% based on the east coast medium-high population density activity thresholds detailed in table 2 of Phillips and Callaghan (2011), with the addition of a 10% activity contour. The 10% contour buffers the extent of significant activity and delineates areas of potential important use in elevated sections of the study area where koala carrying capacity is likely to be lower, while also providing insights into aspects of connectivity between areas of higher use.

Analysis of occupancy using scat surveys

In addition to naïve estimates of the proportion of area occupied (naïve occupancy) arising from field surveys, we used an occupancy framework (see Mackenzie et al. 2018) to investigate changes in occupancy between the two time periods. We used the correlated detection model of Hines et al. (2010) as implemented in program PRESENCE ver. 12.24 (USGS Patuxent Wildlife Research Centre, Laurel, MD, 20708, USA). This approach makes use of spatial repeat samples at sites where there is only a single sample occasion. Because there are spatial replicates at the sample units it is likely that those samples are highly correlated. This requires estimation of availability of the target species to be detected. Furthermore, if the correlated nature of the samples is not accounted for the estimates of occupancy will be biased.

We had two periods of sampling that we analysed separately. We were not interested in parameters that related to colonisation and extinction at sample sites given that the spatial coverage of our sites changed over time due to removal of some habitat areas. We had up to 30 spatial replicate surveys at each site consisting of koala scat searches under individual trees. We constructed detection histories (H) for each site reflecting whether a scat was detected (1) or not (0) at each tree (i.e. H = 0,0,0,1,1,0,1,1 and so on). The modelling approach of Hines et al. (2010) includes different parameters than other forms of occupancy modelling. The key addition is the need to model the availability of the target species given the correlated sampling. This model includes occupancy (psi), and detection (P) but in this case, detection is conditioned on availability. The other key difference is the inclusion of three availability theta (θ) parameters: (1) θ₁ is the probability of availability (i.e. koala scats) in the first survey or in this case at the first tree sampled; (2) θ⁰ is the probability of availability where not detected at the previous tree survey; and (3) θ¹ is the probability of availability when detected at the previous tree.

We conducted our analysis in three steps. We first ran different detection models and then used the best outcomes to model the θ parameters. Once the top θ models were identified we refitted the detection covariates to ensure the top model had been identified. We investigated the influence of five tree preference groups as the covariates that might influence detection and the θ parameters. A large body of literature documents that koalas exhibit preference for different tree species or groups of species (e.g. Moore and Foley 2000; Phillips and Callaghan 2000; Phillips et al. 2000; Sullivan et al. 2003; Tucker et al. 2007; McAlpine et al. 2008). Hence, we expected tree preference to influence whether a scat was detected. The tree groups were those identified in the analysis of tree use: group 1, which is the statistically homogeneous suite of most-preferred eucalypt species); group 2, which is the suite of less-preferred eucalypt species in Table 1; and groups 3, 4, and 5, which are three ranked sets of non-eucalypt species. These covariates were scored as present/absent and so were dummy coded for each sampled tree. We included these groups as covariates one at a time with a parameter and then constructed models incorporating several of these covariates.

Models were compared using Akaike’s Information Criterion for small sample size (AICc), to indicate how well they explained the data (Burnham and Anderson 2004). Models were ranked from lowest to highest AICc. Differences in AICc between the top model and another model (ΔAICc) suggest the strength of support for that model (Burnham and Anderson 2004). Models with ΔAICc <2 were considered equally supported. Increasing values of ΔAICc indicate less support for a model. If a covariate added to a top model did not improve model fit by >2ΔAICc it was deemed an uninformative parameter and omitted (see Arnold 2010).

Vegetation condition

Vegetation health assessments were conducted at each SAT site in 2017. Each tree surveyed within the SAT was given a health rating of (1) <30% healthy foliage present (i.e. foliage not missing due to tree death or decline, or mistletoe infestation), (2) 30–70% healthy foliage present, or (3) >70% healthy foliage present (Department of Sustainability and Environment (DSE) 2004). A Geographic Resource Solutions densitometer was used to score canopy cover at 25 points within each 25-m radial survey area, five around the centre point and five at each of the four cardinal points at the outer margin of each assessment site. An average of the 25 values was calculated. The species and number of saplings (i.e. trees <10 cm dbh) were also recorded within each of the 25-m radial search areas at each survey site. Any evidence of declines in tree health or defoliation that may be associated with over-browsing pressures by koalas is likely to be focused on the identified preferred tree species (e.g. Martin 1985).

Tree species use

The baseline data for the East Coomera surveys in 2006 involved 5142 trees from 200 SAT sites. The surveys for 2017 sampled 2615 trees across 106 SAT sites that remained in vegetated areas, including five sites that could not be accessed in 2006. This included 859 trees from 14 eucalypt species and 1756 trees from 27 non-eucalypt species (Table 1). Tree species use in 2017 from sites with scats ranged from 0.88 for swamp mahogany to 0.15 for broad-leaved red ironbark. Tree use data from both sampling periods were determined to be normally distributed. There was a significant increase in the level of proportional utilisation for the pooled suite of seven preferred eucalypt species from 2006 to 2017 (mean 2006 = 0.448, mean 2017 = 0.622; χ² = 14.832; 1_d.f., P = 0.0001). Use of the pooled group of 10 non-preferred eucalypt species was also significantly elevated in 2017 compared to 2006 (mean 2006 = 0.238, mean 2017 = 0.32; χ² = 5.611; 1_d.f., P = 0.0179).

Koala activity levels

In the 2006 baseline surveys, 159 (79.5%) of the 200 SAT sites had evidence of koala activity (i.e. koala scats recorded). The mean koala activity level in 2006 was 28.4% (±2.8% s.e.). A total of 52% of sites (n = 105) had an activity level above the 22% activity threshold for medium use. Koala activity levels in 2017 averaged 41.4% (±2.9% s.e.) with koala scats recorded at 89.6% of the SAT sites. The activity levels for 67% of sites (n = 71) were above the 22% activity threshold for medium use. The mean activity level recorded in 2017 was significantly higher than that recorded in 2006 (F₍₁₀₅₎ = 1.1, P = 0.315, t₍₂₀₅₎ = 3.227; P < 0.001).

Koala activity models

The koala activity models produced from the 2006 and 2017 SAT surveys delineated locations of significant koala activity across the East Coomera study area (Figs 2 and 3). Comparison of the two models indicated that koalas were still resident in 2017 in previously occupied locations where suitable habitat remained. The comparison revealed:

Fig. 2.

Koala activity boundaries derived for the East Coomera koala population study 2006. Field survey sites are represented by the white dots.

Fig. 3.

Koala activity boundaries derived for the East Coomera koala population study 2017. Field survey sites are represented by the white dots.

Occupancy analysis

Correlated detection, 2006

A detection model where tree group 1 was added as a covariate differed to the null model by >100 AIC. The addition of other tree group covariates to the group 1 covariate including less preferred eucalypts (group 2 (g2)) and ranked non-eucalypts (group 3 (g3), group 4 (g4) and group 5 (g5)), substantially improved model fit (Table 2). The addition of group 3 substantially improved model fit. This model had a model weight of 1.0 so was substantially more plausible than any other model. Estimates for detection at a tree conditioned on when a koala was present was much higher for tree groups 1 and 3 than for the other three groups, though it was above 0.5 for all groups (Fig. 4).

Table 2.Model selection results of the top five models for correlated detection for the detection and theta parameters for the 2006 survey.

Model	AICc	ΔAICc	w	ML	k
Detection
p(g1 + g3)	4552.27	0.00	1.00	1.00	7
p(g1 + g4)	4566.19	13.92	0.00	0.00	7
p(g1 + g2)	4567.26	14.99	0.00	0.00	7
p(g1 + g5)	4567.26	14.99	0.00	0.00	7
p(g1)	4570.71	18.44	0.00	0.00	6
Theta (θ) parameters
θ0(g4 + g5), θ1(g4 + g5)	4525.03	0.00	1.00	1.00	11
θ0(g4 + g2), θ1(g4 + g2)	4543.24	18.21	0.00	0.01	11
θ0(g4), θ1(g4)	4544.32	19.29	0.00	0.00	9
θ0(g5), θ1(g5)	4544.38	19.35	0.00	0.00	9
θ0(g2), θ1(g2)	4544.42	19.39	0.00	0.00	9

The covariates represent different ranked tree groups (g) including eucalypts (g1 and g2) and non-eucalypts (g3, g4, and g5). For detection only that parameter is shown. The top detection model was used for all theta models but is not shown. The occupancy parameter (psi) is held constant and is not shown.

Fig. 4.

Parameter estimates (see text for definition) from the top correlated detection model for 2006 koala scat surveys. g, different numbered tree groups (see Table 1); th,  θ (theta). The symbol (/) indicates those groups were estimated as equivalent.

The addition of one of the tree group covariates to the θ parameters improved model fit relative to a null θ model except for the inclusion of group 3. The addition of either group 4 or group 5 differed to the null θ model by >7 AIC. The top model included groups 4 and 5, and consequently groups 1, 2 and 3 were estimated equal. This showed that if a scat was not detected at the previous tree the probability of being present at a tree (θ⁰) was low but almost twice as high for group 1, 2 and 3 compared to group 4 and three times higher than for group 5. If a scat was detected at the previous tree the probability of being present at a tree (θ¹) was very high (>0.7) for all tree groups (Fig. 4). The substantially different values for θ⁰ and θ¹ confirm the correlated detection process of these surveys. The estimate of psi from the top model was 0.81 ± 0.03. The estimate of θ₁ was 0.75 ± 0.07.

Correlated detection, 2017

A detection model in which tree group 1 was added as a covariate differed to the null model by >40 AIC. This model differed to the next best model by >30 AIC and was clearly the most plausible model (Table 3). The estimate for detection at a tree was much higher for tree group 1 compared to all other groups (Fig. 5). The addition of tree group 1 and 5 covariates to the θ parameters improved model fit relative to a null θ model (Table 3). A model with group 3 did not converge so was excluded. The top model that included group 1 was the most plausible, differing to the next model by >10AIC. Under this model if a scat was not detected at the previous tree the probability of being present at a tree (θ⁰) was 0.38 for group 1 but 0.17 for other groups. If a scat was detected at the previous tree the probability of being present at a tree (θ¹) was very high (>0.7) for all tree groups (Fig. 5). The estimate of psi from the top model was 0.90 ± 0.03. The estimate of θ₁ was 0.66 ± 0.11.

Table 3.Model selection results of the top five models for correlated detection for the detection and theta parameters for the 2017 survey.

Model	AICc	ΔAICc	w	ML	k
Detection
p(g1)	2751.77	0.00	1.00	1.00	6
p(g4)	2786.53	34.76	0.00	0.00	6
p(g5)	2789.11	37.34	0.00	0.00	6
p(g3)	2794.86	43.09	0.00	0.00	6
p(g2)	2798.55	46.78	0.00	0.00	6
Theta (θ) parameters
θ0(g1), θ1(g1)	2733.69	0.00	1.00	1.00	8
θ0(g5), θ1(g5)	2747.47	13.78	0.00	0.01	8
θ0(.), θ1(.)	2751.77	18.08	0.00	0.00	6
θ0(g4), θ1(g4)	2752.65	18.96	0.00	0.00	8
θ0(g2), θ1(g2)	2753.79	20.10	0.00	0.00	8

The covariates represent different ranked tree groups (g) including eucalypts (g1 and g2) and non-eucalypts (g3, g4 and g5). For detection only that parameter is shown. The top detection model was used for all theta models but is not shown. The occupancy parameter (psi) is held constant and is not shown.

Fig. 5.

Parameter estimates (see text for definition) from the top correlated detection model for 2017 koala scat surveys. g, different numbered tree groups (see Table 1); th, θ (theta). The symbol (/) indicates those groups were estimated as equivalent.

Koala sightings

Nine koalas were located during the 2006 baseline surveys within the 25-m radial search area at seven of the 200 SAT sites covering 39.2 ha (Biolink 2007a). The repeat SAT surveys in 2017 resulted in six koala sightings from 101 SAT sites covering 19.8 ha. Health profiles were established for each member of the source population when captured for translocation or in situ monitoring. The health rating data was used for survival analysis modelling for the translocation study group compared to members of the source population, and existing residents at recipient sites. Outcomes from that component of the study are reported separately.

Population density and size estimates

A density estimate of 0.30 ± 0.05 (s.e.) koalas per hectare (ha⁻¹) was derived from 101 SAT-based radial searches in 2017. The estimated density from the 2006 surveys was 0.23 ± 0.03 (s.e.) koalas ha⁻¹. Extrapolation of the 2006 density estimate across 2217 ha of koala habitat produced an overall population estimate of 510 ± 67 (s.e.) koalas, compared to 440 ± 66 (s.e.) koalas for 2017 across 1467 ha of remaining habitat (Table 4). The comparison of mean (M) density estimates derived for the two survey events (n = 10 for each event) using the set of 101 repeat sites (M 2006 = 0.234; M 2017 = 0.295) indicated that they were not significantly different (F₍₉₎ = 2.029, P = 0.153, t₍₁₈₎ = 1.116, P = 0.279 two-tailed test).

Tree health

The assessment of tree health in 2017 involved 2615 trees across the SAT survey sites. Overall, 243 trees (9.3%) were recorded in the highest health category (3) >70% healthy foliage, 1909 trees (73%) in the medium category (2) 30–70% healthy foliage, and 463 trees (17.7%) in the lowest category (1) <30% healthy foliage. All sites except for four had ≥50% of trees in the medium health category, three of these recorded ≥40% of trees in the lowest category. Eleven sites in the north-east, south-east and north-west sections of the study area recorded ≥20% of trees in the lowest health category.

Health of preferred koala food trees (PKFTs)

PKFTs were present at 70.8% of the sites and represented ≥20% of surveyed trees at 31.1% of sites. A total of 370 PKFTs were assessed with 74.3% of these assigned to the medium health category (2), displaying 30–70% expected healthy foliage cover (Table 5). The proportion of each PKFT species assigned to the medium health score included tallowwood (80.95%), forest red gum (74.92%), small-fruited grey gum (68.42%), and swamp mahogany (62.50%). Of the forest red gums, 15.2% recorded <30% healthy foliage present. Of sites where PKFTs were present, only four were of concern regarding health ratings for PKFTs compared to non-preferred trees. Three of these sites occur in areas of high browsing pressure due to relatively small patch size and high koala density.

A total of 11,190 saplings were recorded from the SAT sites. Of these, 2.2% (n = 237) were PKFT saplings; predominantly forest red gum, as well as tallowwood, grey gum, and occasional swamp mahogany. Of the sites where PKFT saplings were absent (but other saplings were present), 75.7% (n = 78) contained PKFTs >10 cm dbh. Where PKFT saplings were present, 86.3% of sites (n = 44) contained PKFTs >10 cm dbh (Table 6). Within the 25-m radial assessment areas at SAT sites, PKFT saplings were present at 48.1% of sites (n = 51). Regression analysis did not identify any significant correlation between koala activity levels and the numbers of saplings present at survey sites (y = −0.034x + 44.965, r² = 0.012), which was explored to investigate whether sapling abundance might have a positive influence on koala activity.

Eighty sites (75.5%) recorded mean densitometer readings between 40% and 80% foliage projective cover (FPC). Eleven sites recorded mean FPC readings above 80%, predominantly in north-eastern sections of the study area beyond the urban landscape. Remaining sites with mean FPC readings below 40% were mainly located in or adjoining urban areas or in partially cleared rural areas.

Potential effects of increasing density on reproduction

Evidence from studies on other species has indicated that crowding following sudden habitat loss can have negative impacts on demographic and social structures, as well as species diversity and assemblage patterns (e.g. Vallejos et al. 2020; Amorim et al. 2022; Cudney-Valenzuela et al. 2023). However, despite possible crowding impacts on demographics and survival, overall fecundity may not be reduced, at least over the short term (Amorim et al. 2022). This scenario may be relevant to outcomes for the East Coomera source population, although the reproductive rate of koalas is reasonably low despite some populations having the potential to increase comparatively rapidly (Phillips 2000). While healthy adult female koalas may produce a maximum of one offspring per year following sexual maturity at around 2 years, an average of one young each second year would be most typical over their reproductive life, plus fecundity declines with age and with chlamydial infection (see Handasyde et al. 1990; Martin and Handasyde 1990). The consequence of a shift in population density and demographics may be that key social factors and age-structures involved in regulating population size and carrying capacity could take several years to return to equilibrium. There are several documented cases in Victoria where koala populations have experienced sharp declines in response to unchecked population expansion and over-browsing pressures (Martin and Handasyde 1990; Phillips 2000; Menkhorst 2008). However, there was no evidence of this occurring at East Coomera during the study period.

Capacity of the source population to respond to translocation harvest

The evidence from the East Coomera study demonstrates that relatively large koala populations have the capacity to respond successfully to a progressive harvest associated with a translocation program. This is consistent with broad evidence from koala translocations in Victoria for koala management purposes in the past 90 years (Martin and Handasyde 1990; Menkhorst 2008; Menkhorst 2017; Menkhorst et al. 2019). However, the capacity for any given source population to respond successfully is likely to depend upon critical factors such as the initial population size and demographic parameters, the scale and pace of translocation, population health, habitat quality and functional connectivity, and threatening processes. Several factors may have contributed towards the successful response of the East Coomera source population to the translocations including:

Destabilising effects on the source population may have been far more severe had the translocations involved greater numbers or been undertaken over a condensed timeframe. This knowledge is important for the design of future conservation-based koala translocation programs, where it is crucial to avoid major disruptions to source populations. This is likely to be particularly salient when considering Chlamydia-positive populations with lowered reproductive potential and corresponding increased sensitivity to habitat loss and other disturbance (Phillips 2000). There is some evidence to link koala population declines at source locations with harvesting for translocations where habitat loss was not a potential contributing factor to a downward trajectory (Gall 1978, 1980; Phillips 2000). This suggests that individual source populations may respond differently to translocations based on a potentially complex array of factors, regardless of additional disturbance effects. For example, the work of McAlpine et al. (2006) promoted the importance of size of forest habitat patches, their spatial configuration and quality, and connectivity for the occurrence and persistence of koala populations. Rhodes et al. (2008) found that koala habitat-occupancy values were best described by threshold models, whereby minimum requirements for native forest cover of around 60% were identified for the Noosa area, in the same bioregion as the East Coomera study site. Rhodes et al. (2006, 2014) reported that the probability of koala presence increased with the amount of primary and secondary quality koala habitat and decreased in areas supporting high proportions of marginal habitat, and when the density of roads increased. These factors highlight the importance of context when interpreting outcomes from translocation harvests on source populations.

Management implications

Predicting precise impacts of translocation programs on a source population is a complex consideration irrespective of the target species (e.g. Dimond and Armstrong 2007; Mitchell et al. 2022). The primary aim for this study was to measure changes in the size and distribution of a koala source population following a progressive translocation harvest and the outcomes may help to inform the design and management of future translocation programs. The following points are recommended for consideration when assessing potential impacts of translocation harvest on koala source populations: