Introduction

Anthropogenic activities have accelerated with human population growth and associated demands in resources. As a result, loss and fragmentation of native habitat has been established as a leading cause of biodiversity decline (Maxwell et al. 2016; Horváth et al. 2019) further compounded by ongoing climate change (Scheffers et al. 2016). As primary habitat becomes rarer, species are being restricted to smaller, degraded parts of their natural range, isolated by lack of connectivity, and increasingly exposed to anthropogenic threats.

Within recent decades conservation efforts have accelerated in an attempt to reverse negative trends in wildlife populations (Rands et al. 2010; Evely 2012). Most efforts are towards protecting contiguous primary habitat (Tulloch et al. 2016; Wintle et al. 2019), considered to have higher resilience and biodiversity (Herse et al. 2018). However, with escalating anthropogenic activities (e.g., infrastructure development), primary habitat is becoming scarce and fragmented in the landscape. Species are therefore increasingly found living outside of primary habitat, i.e., in secondary habitat classified as disturbed or regrowth- the value of which is debated. Studies have suggested secondary habitat becomes increasingly important for species conservation as a means of maintaining habitat connectivity and total habitat area available (Chazdon et al. 2009; Hilje and Mitchell Aide 2012) while other studies have found little benefit for biodiversity (Gardner et al. 2008). The response to habitat disturbance and overall anthropization is likely species specific, dependent on their ability to obtain sufficient resources for survival and reproduction (Jones et al. 2021) in modified landscapes. This is particularly important to consider, for threatened and specialist species which are less adaptable and more vulnerable to habitat disturbances.

There is evidence showing that restoring secondary habitat has benefits for species occurrence and abundance (Lewis-Phillips et al. 2019; Casazza et al. 2021), but can fail without understanding where restoration should occur and how to maximise these efforts (Sutherland et al. 2004; Arlettaz et al. 2010; Segan et al. 2011). It is critical to investigate the value of secondary habitat and its capacity to compensate for the impacts of new developments on biodiversity, and whether these could provide a means to compensate for ongoing primary forest loss (Bennett et al. 2017).

The success of restoring or protecting habitat is dependent on whether species will use and persist in such habitat. At the patch scale, characteristics such as habitat composition, structure and quality are known to influence species occurrence and fitness parameters in a variety of taxa (Street et al. 2016; Gardiner et al. 2018; Sanders and Chalfoun 2018; Rios et al. 2021). At the landscape scale, the availability of resources such as the area of habitat available (e.g., available forest cover) and distance to high value resources (e.g., water) are important to ensure species can recolonise new areas and persist when extreme events such as drought or bushfires occur (Bjørneraas et al. 2012; Lesmerises et al. 2012). Threats from anthropogenic factors (e.g., roads), must also be considered as these can have direct and long-term impacts on species movement, survival and population persistence (Benítez-López et al. 2010). In addition, behavioural processes occurring at various scales, such as foraging within the patch scale (home range), and dispersal at the landscape scale (Frey et al. 2012) will influence the scales at which different factors influence a species occurrence.

Australia is a biodiversity hotspot but one of the leading countries in rates of habitat loss and species extinction (Simmonds et al. 2019; Ward et al. 2019). The major industries of agriculture, infrastructure development and mining are economic titans and large contributors to cumulative land clearing (Reside et al. 2019). Since the arrival of Europeans, biodiversity has encountered a myriad of additional threats including introduced and invasive species, disease, climate change and changing fire regimes (Taylor et al. 2011; Ward et al. 2020). Despite being a politically stable and an economically wealthy country, Australia’s main environmental law ‘Environment Protection and Biodiversity Act’ (EPBC) has been unable to halt negative impacts on biodiversity (Ritchie et al. 2013; Ward et al. 2019), even for iconic charismatic species like the koala (Phascolarctos cinereus). The koala is an arboreal specialist that is now endangered across the northern half of its distribution (Australian Capital Territory, New SouthWales and Queensland), due to multiple threats including habitat loss, predation, vehicle collisions, and disease. Despite the species’ protected status, koala populations have lost up to 50–80% of population densities in part of Queensland in the last 20 years (Rhodes et al. 2015) and experienced an average decline of 24% in population numbers over the last three generations across Australia (Adams-Hosking et al. 2016). With increasing scrutiny of the EPBC Act and its shortfalls (Cristescu et al. 2018) including lack of appropriate implementation of habitat offsets (Maron et al. 2015), there is a need to increase the effectiveness of habitat restoration in increasingly modified landscapes.

Previous research has reported that habitat experiencing little to no disturbance, like old growth or protected primary habitat, influenced the occurrence, movement and persistence of koalas, driven by the total area and density of preferred food trees (Eucalyptus species) (McAlpine et al. 2006). More recent studies have highlighted that koalas do not discriminate between primary and secondary habitat (Barth et al. 2019; Cristescu et al. 2019), raising questions on the potential value of secondary habitat for koalas. Although the species occurs in secondary forests, little is known regarding what influences their occurrence in secondary forests and whether these can be used as a guide for restoration. However, Law et al. (2018) have found resilience of koalas to timber harvesting in NSW (Law et al. 2022a, b). Using a multi-scale approach, we determined the probability of occurrence in primary (undisturbed forest) and secondary eucalyptus forests (previously disturbed forests) at multiple scales, and accounted for environmental and anthropogenic variables that may influence koala occurrence. Furthermore, we investigated the interactions between secondary eucalyptus forests and other variables, in order to determine what may further influence koala occurrence in relation to secondary forests.

Material and methods

Study area

Koala surveys were carried out in seven local government areas of Queensland, in eastern Australia: Sunshine Coast, Noosa shire, Gympie, Toowoomba, Moreton Bay, Redlands City and Fraser Coast. We separated North Stradbroke Island from Redlands City Council due to its distinct geographical and vegetation characteristics as an island population, therefore totalling eight regions. The climate across Queensland is subtropical, with broad habitat categories ranging from open dry sclerophyll forests, Acacia- and Callitris-dominated forests, and patchy rainforests. Native vegetation across the region experienced the highest levels of clearance for the expansion of cattle and agriculture over the last 50 years (McAlpine et al. 2009; Reside et al. 2017), resulting in large-scale fragmentation. Sites were selected based on project purpose to either determine koala occupancy, distribution and genetic population structure.

Koala occurrence

Given koalas are cryptic and often difficult to find, we followed the Cristescu et al. (2015) method of utilising detection dogs to find evidence of koala presence. A team consisting of a handler and detection dog specifically trained for koala scent (scat or fur, n = 7) were deployed across the study area for different purposes (koala habitat mapping, occupancy, genetic sampling or live koala detection). Based on previous surveys, in koala-positive sites, casual surveys with detection dogs are 19 times more efficient than surveys with humans and 153% more accurate in confirming koala presence (Cristescu et al. 2015, 2020). In a casual survey, the detection dog is allowed to search freely and cover a large area without being constrained to a specific area (i.e., as opposed to a method such as the Spot Assessment Technique where the dog is limited to 30 trees (Cristescu et al. 2015)). Therefore we only used the casual surveys. The target area varied from entire regions, townships to specific reserves or private properties, therefore the search area for each survey varied substantially.

All surveys were conducted in the same manner with no prerequisite in terms of habitat type. Each survey began once the detection dogs were motivated to find the target scent and ended when the target scent was found or the team had searched the area, often within 40 min. If a detection dog indicated on a target scent, the handler confirmed koala presence/scat sample by sight (Triggs 2004) and recorded its location. We recorded a site to be used by koalas if koala scats of any age or a koala were found, and absent when the target scent was not detected during the survey. The teams also recorded the total search time to account for survey effort.

We conducted a total of 2230 surveys between 2015 and 2020. We then filtered our data for surveys that were specifically used to ascertain koala presence/absence (n = 704). For surveys that were undertaken using detection dogs specially trained to find more specific scents (i.e., very fresh scats for molecular analyses or koalas), we only included surveys that were positive and removed negative surveys (n = 249). This is because detection dogs trained only on fresh scats (less than one week old), or only on koalas, will not indicate on older koala scats—this however does not mean koalas are absent from the area, only that koalas have not recently (one week) used the area. To ensure there were no biases, we ran analyses excluding specifically trained dogs and there was no change in results (see Supplementary 1). As a result, a total of 959 surveys was used for subsequent analyses, of which 575 were positive for koala presence and 384 were negative (Fig. 1).

Fig. 1
figure 1

Spread of survey sites in Queensland using detection dogs to determine koala occurrence (n = 958) across primary (dark grey) and secondary (light grey) habitat. Black circles represent positive sites, white circles represent negative surveys

Full size image

Environmental variables

If a site was positive, variables were extracted for the first scat found of the survey. If the site was negative, variables were extracted from the survey’s starting point. Variables were extracted at three different buffer scales: (i) 250 m, (ii) 500 m and (iii) 1500 m radii. The buffer sizes were chosen to represent a conservative measure of the varying dispersal distances displayed by koalas across their range, which have been used in previous studies determining local and landscape drivers of occurrence (Dique et al. 2003a; McAlpine et al. 2006; Rhodes et al. 2008).

We extracted all variables of interest for each buffer size using ArcGIS (10.8) from the following spatial layers (also see Table 1):

  1. (i)

    Forest extent: we used the Forests of Australia 2018 dataset created by the Australian Bureau of Agricultural and Resource Economics and Sciences (data.gov.au), mapping the dominant forest types with a one hectare resolution. Given koalas are considered specialist folivores of the Eucalyptus genus, and these are likely the genus used for conservation efforts (i.e., restoration), we considered the extent of eucalyptus forests. We aggregated and clipped forest types and calculated the extent of Eucalyptus and Melaleuca forest in each buffer size. Other species of food trees such as Corymbia and Lophostemon were not explicitly mapped as dominant forest types however may be present within Eucalyptus forests (observations in the field).

  2. (ii)

    Proportion of Secondary forest: we intersected the extent of Eucalyptus forest within buffers with the Remnant regional ecosystem mapping 2019 (Queensland Spatial https://qldspatial.information.qld.gov.au/). We calculated the amount of Eucalyptus forest that was classified as either primary or secondary eucalyptus forests within each buffer. Primary eucalyptus forests are classified as original native remnant forest where the dominant canopy species is greater than 70% of the height and compromises of more than 50% of canopy cover that shows no evidence of disturbance (Fig. 2). These are often national parks, protected properties or indigenous protected areas (i.e., North Stradbroke Island). Secondary forest is classified as non-remnant-original vegetation which has experienced some disturbance (i.e., logging, thinned, or poisoning of invasive plants) and has less than 50% cover remaining of the dominant canopy, and or shows evidence of present regrowth. (Neldner 2022). This includes state forests and reserves where extractive activities are permitted. The proportion of secondary and primary Eucalyptus forest was inversely proportional to each other, and therefore we only kept the proportion of secondary forest for analyses.

  3. (iii)

    Water: the total surface water area in each buffer was calculated as previous work has found koala survival can be linked to distance to water (Gordon et al. 1988; Wu et al. 2012; Smith et al. 2013) where koalas have experienced drought, heatwaves or live in semi-arid regions. Water also plays an important role in habitat resilience (Mella et al. 2019). We extracted the area of water from the National Surface Hydrology spatial layer (ga.gov.au). We excluded water bodies classified as ‘foreshore flats’, associated to beaches and likely to include saltwater bodies which are not specifically linked with koalas.

Table 1 Name and definition of variables used to determine koala occurrence across our study site. Region was used as a random effect in our models
Full size table
Fig. 2
figure 2

Example of sites surveyed for koala presence. The top row are sites classified as primary forest with more than 50% of the dominant tree canopy as undisturbed. The bottom row are examples of sites classified as secondary forest with more than 50% of the dominant canopy disturbed and may contain regrowth

Full size image

In addition, from each point, we extracted the following variables:

  1. (iv)

    Road: we calculated the distance to the closest major primary road using the Near tool. Data were extracted from spatial layer Global Roads Inventory Project dataset (globio.org).

  2. (v)

    Phosphorous: Soil phosphorous is considered important for the health of eucalyptus forests (Thomas et al. 2006) and was previously highlighted as a contributor to koala presence (DES. 2020). We extracted mean mass fraction of total phosphorous in soil by weight (%) within topsoil layers of 0–5 cm from the Soil and Landscape Grid National Soil Attribute (data.csiro.au).

  3. (vi)

    Elevation: Elevation has previously been a variable included in determining occurrence of koalas (Sequeira et al. 2014). We extracted elevation from Digital Elevation Model based on 30 m grid cells (pid.geoscience.gov.au).

  4. (vii)

    Rain: As an indicator of available water content in leaves, we collected rainfall statistics from weather stations in the closest towns to our survey points—It is reported that there is a time lag between weather events and vegetation response (Wu et al. 2015; Ding et al. 2020); some of our surveys coincided with El Nino events and cyclone season occurring over many months, so as a conservative time frame we extracted average rainfall recorded from the year prior to conducting surveys. This was obtained from the Bureau of Meteorology (bom.gov.au).

Statistical analyses

We used Generalized Linear Mixed Models (GLMMs) with a binomial distribution to evaluate the relative influence of environmental variables at the three buffer scales on the occurrence of koalas. We considered the following ten variables for our models: proportion of secondary eucalyptus forest, total forest extent (m2), distance to closest major road (m), total surface water area (m2), soil phosphorous (%), average rainfall (mm) and elevation (See supplementary 2)—these variables were scaled before being fit to models. We also included year to account for variations on the probability of detection for koalas as our study period encompassed different phases of the El Nino-Southern Oscillation which can result in drought (El Nino) to heavy rainfalls (La Nina), and influence koala numbers (Seabrook et al. 2011; Speight et al. 2018). We also included total search time (seconds) for each survey to control for the potential effect of difference in survey length. Regions was included as a random effect in these models to account for repeated observations in each region over the years as well as differences between regions.

We used Pearson’s coefficients to test for correlations between variables, particularly for forest extent and proportion of secondary forest to ensure these were not inversely correlated (Supplementary 3). We also investigated multicollinearity among explanatory variables using the variance inflation factor (VIF), where we considered variables with VIF values > 3 to be highly correlated. All covariates had VIF < 2, however soil phosphorous and elevation showed some correlation and therefore we only kept soil phosphorous for our models. Goodness-of-fit and spatial autocorrelation (Global Moran I) of the models were assessed using residual diagnostics following the procedures described in the DHARMa package (Hartig 2020). We included the location of surveys as coordinates (eastings and northings) in our models to correct for minor distance-based autocorrelation.

We built seven models for each buffer size (250, 500, 1500 m radii). All models included all 10 covariates, and six of these models included a combination of one or two interactive terms between the proportion of secondary forest, and Road, Water and or Rain (Supplementary 4). These interactions were chosen to test the hypothesis that covariates associated to anthropogenic threats (distance to roads) and habitat productivity or resilience (water area and rain) could interact with the proportion of secondary forest to influence koala occurrence (by either enhancing or buffering its effect). All candidate models fit were ranked using the Akaike Information Criterion, the top model was chosen as having the lowest AIC, if models fell within two AIC of each other the most parsimonious model was chosen. We calculated the marginal and conditional R-squared for each of the three best models. All statistical analyses were performed using the software R 4.0.0, and using package lme4 for all GLMMs.

Results

Our top ranked models included all variables and one interaction term (Table 2). Across all buffer sizes, we did not detect any effect of forest extent on koala occurrence. We found a strong positive effect of longitude, suggesting occurrence was greater closer to the coast than inland as expected for our study area while there was a negative effect of prior average rainfall suggesting higher rainfall patterns were less preferable for koalas within our study area. There was strong positive effect of phosphorous in soil suggesting that eucalyptus health influenced koala occurrence. Our results also suggest occurrence likelihood was greater with increasing time spent searching. The distance to major roads had a consistent positive effect on koala occurrence at 500 and 1500 m radii, where surveys located in habitat patches further from major roads had a higher probability of koala occurrence while the total surface water area did not directly influence koala occurrence. Koala occurrence in secondary forest was estimated to be between 40 and 60%, but this varied at each buffer size depending on interactive factors: distance to road or amount of water. Below we detail results that are specific to each buffer size:

Table 2 Summary of the top candidate models at each buffer size. Including odd ratios (OR) for each variable. Values more than 1 represent an increased likelihood of occurrence as the variable increases, while odds ratios less than 1 indicate decreased likelihood of occurrence
Full size table

Small scale: 250 m radii buffer

The top-ranked model indicated there was a strong interaction term between secondary forest and distance to major road at this scale (df: 13, AIC: 853.2, AICw: 0.393, R2m:0.25, R2c: 0.36). On average, koala occurrence in secondary forest was predicted to increase by 25% with increasing distance from major roads. While inversely, koala occurrence was predicted to be higher closer to major roads in buffers containing a small proportion of secondary eucalyptus forests (Fig. 3).

Fig. 3
figure 3

Predicted values plotted for each top model’s interaction term with proportion of secondary forest (represented on the logit scale). Each line represents the lower than average (mean-standard deviation), average, and higher than average (mean + standard deviation) amount of A Distance to major road in the 250 m radii buffer, B Water Area in the 500 m radii buffer and C Water Area in the 1500 m radii buffer. Shaded area represents the 95% confidence intervals

Full size image

Large scale: 500 and 1500 m radii buffer size

The top-ranked model for the 500 m radii buffer (df: 13, AIC: 973.5, AICw: 0.443, R2m:0.26, R2c: 0.33) and 1500 m radii buffer (df: 13, AIC: 1025.5, AICw: 0.431, R2m:0.21, R2c: 0.38) indicated that the probability of koala occurrence in secondary habitat interacted with the extent of water in the area. On average, koala occurrence in secondary forest increased with greater total surface water area within the buffer. Inversely, when water area was lower, koala occurrence was predicted to be higher in buffers with a low proportion of secondary eucalyptus forests (and therefore where the proportion of primary forests is high) than in buffers composed mainly of secondary forests (Fig. 3).

Discussion

Our results show that koala occurrence was influenced by the proportion of secondary forest, phosphorous soil content, water availability and presence of a major anthropogenic threat (distance to major roads). The negative association with rain was likely driven by higher than average rainfall recorded the year prior in Noosa and Sunshine Coast regions where koala presence was low. Previous research focusing on koala occurrence in primary eucalyptus forests have reported quality and configuration to be important for koala occurrence, but that these operated at varying scales. At the patch scale, the proportion of food trees in primary eucalyptus forests is important (Dargan et al. 2019), while at the landscape scale koala occurrence was driven by the proportion of forest area, regardless of primary Eucalyptus food trees (McAlpine et al. 2006; Smith et al. 2013). In our study, the total extent of eucalyptus forests showed no influence on the occurrence of koalas. There was evidence of presence in both primary eucalyptus and secondary eucalyptus forest, but overall occupancy was predicted to be lower in secondary eucalyptus forests depending on distance to roads and water area. Similar findings are reported by Law et al. (2022b)—where the landscape context was more important in determining the occurrence of koala on private land in a modified landscape than extent of forests. Primary forests harbour higher biodiversity and aid in supporting habitat restoration of adjacent areas (Cadavid-Florez et al. 2019; Zahawi et al. 2021), therefore conserving primary Eucalyptus forests should always be a priority to ensure koala persistence. However, there was still evidence of koala occurrence in secondary forests which could further be maximised if the landscape context is considered.

Soil phosphorus as an important determinant of koala occurrence

Ecosystem health is an important determinant of habitat resilience and wildlife persistence. In previous studies soil characteristics were sometimes found to be important attributes for koala habitat use (Dargan et al. 2019). In our study, we found that the level of phosphorus in the top-soil layer was a significant driver of koala occurrence at all spatial scales. Phosphorous is considered an important element promoting the health and growth of eucalyptus forests (Silva et al. 2018), and thus may affect food availability for koalas. Furthermore, phosphorous fertilisation is reported to promote nitrogen absorption (Graciano et al. 2006)—an important component of protein in koala food (Moore and Foley 2005; Stalenberg et al. 2014). Habitat management including plantings may benefit from targeting areas, when all other factors influencing koala occurrence are equal, where soil has greater phosphorous amount, or fertilising tree saplings to promote growth and nutritional value. This could become increasingly important as the nitrogen content of trees is expected to decrease with increasing atmospheric carbon dioxide (Hovenden and Williams 2010), impacting herbivores, particularly specialist species like the koala. We acknowledge that there are other variables that contribute to soil health such as presence of mycorrhizal fungi (Ishaq et al. 2018), distance from agricultural soil, Ph levels and water absorption (Wardle et al. 2004), however phosphorous may serve as an initial indicator of habitat health to target habitat placement for restoration. We note that we did not average phosphorous levels across the whole buffer but focused on one point (where the surveys started or scats were found) as a proxy for the quality of the closest habitat used by koalas.

Roads as a major anthropogenic threat

Road mortality is a significant threat to wildlife populations (Fahrig and Rytwinski 2009) including koalas across their range (Lunney et al. 2022a, 2022b; Melzer and Black 2022), with a reported 300 vehicle related deaths per year in South East Queensland alone (Dique et al. 2003b; de Oliveira et al. 2014). Major roads, with higher volumes and speeds of traffic can act as movement barriers, decreasing habitat connectivity and gene flow (Dudaniec et al. 2013). Locally, road mortality can have a major impact on individual survival and population dynamics regardless of traffic volume (Rhodes et al. 2011, 2014). Dispersing juveniles have the highest mortality rates (Dique et al. 2003a, b), with vehicle collision being the second single contributor to mortality in New South Wales and South East Queensland (Rhodes et al. 2011, 2014). Koalas living in suburban areas readily cross roads to move between patches (de Oliveira et al. 2014; Dexter et al. 2018) and do not appear to avoid them. Moreover, larger highways are likely to limit movement due to the presence of fauna barriers, and high volumes of traffic decreasing chances of surviving crossing attempts. Our results add weight to the importance of managing road threats, showing the presence of major roads negatively impact koala occurrence at all spatial scales.

Koala occurrence in secondary forests: importance of the landscape context

We also found an interaction between secondary forests and roads in which the occurrence of koalas near major roads is predicted to be lower in secondary habitats than in primary eucalyptus habitats. Primary eucalyptus forests are often larger protected areas with less edge effects, and likely to be more resilient to threats like roads. Our results support previous findings reporting little effect of roads on koala occupancy when koalas are found in contiguous forests in New South Wales (Law et al. 2017, 2018). Interestingly, in our results this effect was only detectable in the models built for the smallest buffer size, (i.e., 250 m radii) suggesting that this interaction between secondary forest and distance to major roads has the strongest effect at smaller scales. High traffic and wider crossing area associated with major roads increases mortality rates where animals attempt to cross, in turn decreasing colonisation and abundance in fragmented landscapes (Hamer et al. 2021), which is likely represented in our results. This also suggests that when targeting habitat patches of secondary forests for koala conservation, or adding habitat through restoration, success could be increased if patches are not contiguous with major roads.

Secondly, we also found koala occurrence to be influenced by the interaction between total surface water area and the proportion of secondary forest at the larger buffer scales. Our models showed that the probability of koala occurrence in secondary forest was higher where there was greater water area, inversely the probability of occurrence was higher in primary eucalyptus forests when water area was lower. Primary eucalyptus forests are usually older and likely have deeper roots to access ground water, therefore less sensitive to fluctuations in water availability. Previous studies have reported the importance of free-standing water bodies and rainfall at the edge of the koala’s range, where resource abundance was lower (Gordon et al. 1988; Davies et al. 2013a, 2013b). Furthermore, water availability in the landscape is positively correlated with plant and animal diversity, promoting vegetation cover and persistence (Chamberlain et al. 2020), but also act as climatic refugia (Gordon et al. 1988; McAlpine et al. 2015; Selwood et al. 2015; McLaughlin et al. 2017). This is particularly important to reduce the impacts of droughts and bushfires, which are increasing in severity and extent with climate changes (Reckless et al. 2018) and longer phases of El Nino oscillations.

Koala habitat mapping in Queensland has recently acknowledged the value of secondary forests (DES. 2020), and our study further supports this and adds the need to carefully consider the surrounding landscape beyond habitat extent. We acknowledge the landscape layers used in this study are coarser than preferred. The resolution and therefore classification and delineation of habitat may influence model accuracy (Deblauwe et al. 2016; McKerrow et al. 2018). For example, in areas of high urban footprint, single or small stands of trees important for local koalas may not have been considered within either classification. This may oversimplify the landscape composition available for koalas and not capture finer scale measures of habitat composition or quality. Improvements in satellite imagery, updating of secondary vegetation, and habitat mapping are in progress, which will allow for more robust and fine scale landscape assessments to include single stands of trees. Future studies should take advantage of this to model refugia and target areas for restoration within the koala’s distribution.

Conclusions

Maintaining habitat is the first line of defence for biodiversity, however it is often complex as a myriad of factors influence whether species will use, and persist in, the areas we allocate to them, whether through conservation or restoration. The combination of human pressure on habitats and ongoing climate change are likely to augment the impacts of roads and water availability on wildlife persistence. Overlooking the landscape context can negate the benefits of habitat protection or restoration if not accounted for in management. Overall, our study adds to the body of knowledge supporting that, although primary forests should remain a priority for conservation, secondary forests have a great potential for koala conservation, with the reservation that mitigating anthropogenic threats and promoting resilience might need further consideration. This is crucial to secure species persistence despite an increasingly degraded landscape matrix.