The management of populations of threatened species requires the capacity to identify areas of high habitat value. We developed a high resolution species distribution model (SDM) for the endangered Pilbara northern quoll Dasyurus hallucatus, population using MaxEnt software and a combined suite of bioclimatic and landscape variables. Once common throughout much of northern Australia, this marsupial carnivore has recently declined throughout much of its former range and is listed as endangered by the IUCN. Other than the potential threats presented by climate change, and the invasive cane toad Rhinella marina (which has not yet arrived in the Pilbara). The Pilbara population is also impacted by introduced predators, pastoral and mining activities. To account for sample bias resulting from targeted surveys unevenly spread through the region, a pseudo-absence bias layer was developed from presence records of other critical weight-range non-volant mammals. The resulting model was then tested using the biomod2 package which produces ensemble models from individual models created with different algorithms. This ensemble model supported the distribution determined by the bias compensated MaxEnt model with a covariance of of 86% between models with both models largely identifying the same areas as high priority habitat. The primary product of this exercise is a high resolution SDM which corroborates and elaborates on our understanding of the ecology and habitat preferences of the Pilbara Northern Quoll population thereby improving our capacity to manage this population in the face of future threats.

Northern Quoll, Pilbara, MaxEnt, biomod2, Sample Bias, Cane Toad, Threatened Species

Species distribution models (SDMs) use environmental data from known locations of a species to predict places where that species could potentially occur within landscapes or regions (Booth et al. 2014). SDMs have been used to identify critical habitats for species with greatly reduced distributions (Hamilton et al. 2015; Jetz and Freckleton 2015; Manthey et al. 2015), provide potential translocation sites for species based on known habitat requirements (Adhikari et al. 2012) and predict the movement of invasive species across landscapes under different scenarios (Kearney et al. 2008; Elith et al. 2010). Spatially explicit probability of presence, or prediction of occurrence maps, generated using SDM algorithms, have been used to inform conservation planning and habitat management at both coarse and fine scales. Consequently, they can guide or prioritise future survey efforts and aid in assessing the conservation status of target species. SDMs relate known occurrences of a species with various environmental variables and predict a probability that a species will occur in areas where no data on its occurrence is available. Thus, they help to identify potentially suitable habitat (Elith and Leathwick 2009; Guisan and Thuiller 2005).

The accuracy of a SDM depends on such factors as the quality and appropriateness (in regard to sample size and representativeness) of the presence and/or absence data for the target species or community, the expertise of the modeller, the selection of an appropriate modelling tool (or software package), the selection of an appropriate suite of predictive/independent variables, the quality of the variable data used, and an acknowledgement of the strengths and limitations of the SDM (Elith et al. 2010; Guillera-Arroita et al. 2015). However, where there are shortfalls in appropriate data, modellers often compensate by focussing their efforts on the development and evaluation of novel methods to improve the performance of their models, and thus, to better predict the environmental suitability for species in applied studies (Barbosa and Schneck 2015; Guillera-Arroita et al. 2015).

In this study we set up a SDM to determine the potential distribution (PD) of the northern quoll Dasyurus hallucatus population found in the Pilbara biogeographic region of Western Australia (Thackway and Cresswell 1997). This is a unique population of a threatened species for which conservation management to date has been limited by significant knowledge gaps (Cramer et al. 2015). Limited and largely opportunistic sampling has resulted in constricted and potentially biased presence data and this in turn has resulted in problems in determining appropriate predictive or independent variables. Furthermore, the literature associated with these SDMs shows no evidence that the models have been tested for covariance and sample bias as described by Hijmans (2012) and Fithian et al. (2015), nor have differences between these SDMs been evaluated or explained.

Northern quolls are a suitable subject for distribution modelling as they have a strong habitat affiliation with complex rocky areas, often in close association with permanent water (Begg 1981; Braithwaite and Griffiths 1994; Oakwood 1997; Pollock 1999; Schmitt et al. 1989). In the Pilbara, quoll habitat affiliation aligns with mesas or ranges which are often the focus of iron-ore extraction (channel-iron deposits and banded iron stone formations) and granite outcrops which are often quarried for road and rail bed materials (Ramanaidou and Morris 2010).

Once widely distributed from the Pilbara region of Western Australian (WA) across northern Australia to southern Queensland (Figure 1), the mainland distribution of the northern quoll has now contracted to several disjunct populations (Burbidge et al. 2009; Oakwood 2008). This collapse has largely been linked to invasion by the toxic cane toad Rhinella marina (Braithwaite and Griffiths 1994; Doody et al. 2015; How et al. 2009; Oakwood 2004; Woinarski 2010). Other impacts currently causing rapid and severe declines in northern Australia’s critical weight range mammal fauna (i.e. terrestrial species within the weight range 35 – 4200 g are considered to be particularly vulnerable to introduced predators) are also likely to also be impacting on the carnivorous northern quoll (Burbidge and McKenzie 1989; Cramer et al. 2016). These include altered fire regimes, the grazing impacts of introduced herbivores, climate change and both enhanced mortality and competition for resources (including prey animals) from introduced predators, in particular the feral cat Felis catus (Burbidge et al. 2009; Cook 2010; Woinarski et al. 2015). As a consequence of all these recent declines, the northern quoll is listed as endangered under both the Commonwealth’s Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act 1999) and the Western Australian Biodiversity Conservation Act 2016.

The main WA populations of northern quoll occur in two discrete mainland regions, the Kimberley and Pilbara, separated by the arid Great Sandy Desert. Both mitochondrial DNA sequences and nuclear microsatellite loci reveal clear differentiation between Kimberley and Pilbara populations and a greater distinction between these populations than those in the Northern Territory and Queensland (Spencer et al. 2013; Westerman and Woolley 2016). These WA populations also differ from those remaining in Queensland and the Northern Territory in regard to both genetic structure and demographic parameters and represent the last intact populations in Australia that have not experienced major declines subsequent to the introduction of the cane toad and consequently display the highest levels of genetic integrity (How et al. 2009; Spencer et al. 2013; Spencer 2010).

Given that the Pilbara population of the northern quoll is genetically and demographically distinct from all other populations, retains its pre-European genetic diversity, is currently outside of the cane toad’s distribution, and has much of its habitat still intact, this population has been assigned a high conservation, research and management priority (Cramer et al. 2015).

The available presence data for this population is clustered around areas of development interest to the mining industry, or where targeted surveys have occurred. This begged the question: was this an example of sample bias or a true representation of northern quoll distribution? Sample bias, where sampling has not been uniform over the project area, e.g. where only easily accessed areas, or known populations have been sampled, has the potential to distort a SDM (Phillips et al. 2009). Lacking true absence data for this exercise and being aware of the limited capacity of pseudo-absences to compensate for high levels of sample bias (Phillips et al. 2009), we sought to find a method to eliminate or minimise sample bias in our SDM.

Our objective was to construct a high resolution SDM for the Pilbara northern quoll by applying an innovative form of bias compensation to a well proven modelling method, MaxEnt, and testing this SDM with an ensemble model.

The independent variables that we used were suitably diverse in nature with an acceptable level of covariance within the variables used in the final selection. The use of the jack-knife test to determine the final suite showed that the removal of any of the variables selected in the final suite would have compromised what was a strong model. The literature informs us that the number of variables used was appropriate for a modelling exercise of this nature (Phillips and Dudík 2008; Stockwell and Peterson 2002). Furthermore, the fact that all runs of all models applied these variables differently and, in most instances, allotted acceptable levels of significance to all or most of these variables (while consistently producing high test values), supports this finding. It should be acknowledged that the bias grid was derived from pseudo-absence data for which suitability was determined through a series of assumptions. Consideration of sampling bias was necessary because of a potential shortfall in suitable presence/absence data and the fact that most of the targeted sampling for the northern quoll has been undertaken over a relatively limited area and timeframe. Specifically, many surveys have been conducted in mining areas and associated infrastructure (either historical, current or proposed) which tend to confine the survey effort to particular types of geomorphology. Surveys of non-volant mammals conducted in the region tend to be for environmental impact assessments associated with mining or part of comprehensive regional surveys (e.g. Bamford Consulting Ecologists (2013); Biota Environmental Sciences (2012); Eco Logical (2012)) or part of comprehensive regional surveys (McKenzie et al. 2009) and hence absence of northern quoll in these surveys could be reasonably assumed to be true absences.

As demonstrated (Table 1, Figure 2, Suppl. materials 2 and 3) northern quolls were found to conform strongly to ecological habitat associated with the vegetation, and slope, topography of the rocky areas of the Pilbara bioregion. Primary areas of northern quoll occupation in the Pilbara included the western edge of the Hamersley Ranges, in the granite outcrops of the Abydos Plain, and in the more rugged areas of the Chichester Ranges. Our models identified a low likelihood of occurrence in the Fortescue River floodplain and its upper catchment, the sandy coastal regions of the Pilbara and in the central and southern parts of the Hamersley Ranges. We also identified several areas with a high probablity of presence which have not, as yet, been adequately sampled and which have not been identified as habitat in previous modelling efforts. This is particuarly so in the eastern Pilbara IBRA region. Here, areas are given a high probability of northern quoll although few to no records of the species exist. This area and others identified as of high probability would be logical areas for further survey effort (as well as providing a ready field-based validation of the model), and demonstrate the utility value of SDMs.

The conservation of a threatened species requires good information about population locations and ecological requirements within its geographic range, particularly when threatening processes are ongoing. Applying appropriately selected surrogate species data to both the MaxEnt and biomod2 software packages has enabled us to develop SDMs which identify remarkably similar core areas of likely quoll habitat, as well as less optimal habitat that may only be occupied in favourable conditions. These apparently less favourable habitats may however be of high conservation value as current information suggests that all Pilbara northern quoll populations are genetically linked, and high level of dispersal occurs between geographically distant populations (Spencer et al 2013, Woolley 2015). Consequently, smaller populations of northern quolls in less-preferred habitat may be important in maintaining gene-flow throughout the region.

The SDM pairs analysis (Figure 6) shows a very strong overall correlation between the preferred MaxEnt SDM and the ensemble model with minor differences in predicted habitat being more about conflicts of scale rather than the actual areas nominated. The correlation tests also indicate a strong similarity between the preferred MaxEnt SDM and the ensemble model overall, and a very strong similarity between these models in identifying areas of high habitat value, i.e. when comparing the highest quartiles. This supports our earlier observation that the difference between the two models is largely one of resolution rather than different predictions.

Figure 6.

Pairs analysis between MaxEnt and biomod2 SDMs using 10,000 random points.

In comparison with previous SDMs on this northern quoll population (Biologic 2012; CliMAS 2014; Eco Logical 2012) our SDM (particularly the MaxEnt model) picks up many of the same areas identified as having a high habitat value, but with an apparently greater level of definition. This is particuarly so in those areas which have not been heavily and specifically sampled for northern quoll. In short, there is a tendency for these models not to project far beyond those areas where northern quoll are known to exist. As such these models tend to tell us little more than what we already know, thereby limiting their value for conservation planning. This could also be said for the trial MaxEnt and biomod2 SDMs constructed without the use of surrogate species, either in the form of a bias file or as pseudo-absences. In both cases, areas of high quality habtat appeared to be much more limited to areas already known to be habitat and less likely to projet into areas where northern quoll records are either absent or less frequent.

In this study we have demonstrated a methodology capable of addressing three of the more common problems associated with SDMs, specifically how to: 1) address bias in a high resolution SDM over a large and diverse landscape with a limited, and potentially biased, presence only data set; 2) selecting an appropriate suite of predictive variables for the construction of such a model; and 3) establish a means by which the suitability and outputs of a modelling tool can be verified. We have developed an innovative approach to constructing an SDM by pre-emptively identifying problems likely to arise due to data limitations and addressing these issues by reviewing the options available and selecting a combination of responses which minimise bias effects and meet the needs and constraints of the modeller.

We note that a true comparison between a model with randomly selected psuedo-absences and a bias-corrected model using a bias layer created from surrogate presences, remains the preferred way to demonstrate the usefulness of this form of bias compensation. However, in the absence of broad-scale sampling and ground truthing to test truth versus prediction (as is proposed) the demonstrated approach remains the most feasible form of bias compensation under the given circumstances.

Being aware of the resource and skill limitations preventing many conservation managers from constructing SDMs, this methodology was deliberately selected to meet these limitations. All data used was freely available and all software used was freeware, other than the use of a commonly used GIS package that could also be substituted with freeware. Skill levels were limited to what the authors considered average for an ecological project team, i.e. no modelling, statistical, GIS or programming specialists were required for this study.

In this exercise we produced a SDM that predicted areas where new populations and sub-populations of the northern quoll might be found outside of areas currently known to be habitat. By identifying areas of high habitat value, this SDM facilitates the identification and conservation of high priority habitat areas, potential translocation sites and potential movement corridors. As a consequence of having identified a sound suite of predictive variables, our understanding of the habitat requirements of the Pilbara population of the northern quoll has been increased. Finally, by comparing the distributions identified through this exercise with proposed mining and infrastructure projects in the environmental impact assessment process, this SDM can be used to minimise impacts on this unique and important northern quoll population.

The ensemble modelling process validates our choice of the MaxEnt model with bias file as the preferred SDM. However, this is a desktop exercise derived from a relatively small and uniform sample. This model should be validated and refined through on ground sampling and research. Our study exemplifies a preferred practice in the use of SDMs by corroborating our findings with a control, in this case one derived through an ensemble modelling approach. We commend this practice to modellers and caution against outcomes derived from only a single modelling approach. Our study has revealed the most comprehensive known refined distribution map for the endangered Northern Quoll. It has confirmed their reliance on rocky upland habitats and their limited distribution within the Pilbara region. These outcomes will be of great importance to land managers when considering the impacts of planned developments within the region.