Discussion

Estimating population density is fundamental to ecology and conservation but arboreal nocturnal animals pose unique challenges for achieving this. Cryptic behaviour and efficacy of capture/detection can be a hindrance for accurate density estimation in these animals. We address this challenge for the sugar glider in the context of its unusual role as a threat to a critically endangered species (Owens et al. , 2022, Stojanovic et al. , 2014). Our literature review revealed that four decades worth of effort have not changed capture or trapping rates, with little innovation in baiting approaches. In our field study, we showed that the traditional bait used by practitioners are less effective than a new bait with fish oil.
Using fish-oil baits, we tripled the usual detection likelihood of past studies and showed that it is possible to estimate density in environments where typically detections are too sparse to support rigorous analysis. Our modification of the lure incorporates contemporary knowledge of the preferred diet of sugar gliders. Within our field study, fish oil improved glider detection probability 33-fold (0.33) compared to honey (0.01). Moreover, we detected nearly three times as many individuals and recorded 23.6 times more detections with fish oil than honey, which improved confidence in density estimates. Gliders may be protein-limited by their typical diet (Smith & Green, 1987), necessitating the consumption of invertebrates (Smith, 1982) or vertebrates (Stojanovic et al. , 2014). Fish oil is known to appeal to carnivores worldwide (Austin et al. , 2017, Heinleinet al. , 2020), and the oil-based emulsion likely persisted longer on trees than soluble honey. Our study is a promising new approach for implementing field-surveys for this species.
Our literature meta-analysis of previous studies explored the range of approaches used in the field and their effectiveness. Methods developed long ago were subsequently applied to most studies with relatively little innovation in trapping techniques. Although this improved comparability among studies, generic baits yielded low capture and trapping rates. Our estimated glider density (0.12 ha-1) is within the range reported in other studies in continuous habitat (0.09-0.54 ha-1) but amongst the lowest reported (Gracanin et al. , 2022, Jackson, 2000b, Quin, 1995). Comparison of densities is complicated, however, by the different analytical approaches between studies (e.g. smaller buffers and sampling area calculations) (Jackson, 2000b, Quin, 1995) compared to SECR (Effordet al. , 2009). Gracanin et al. (2022) provides the only SECR-derived estimate (range: 0.09 ha-1 in continuous forest to 0.7 ha-1 in fragments) and had too few detections in one continuous forest patch to calculate density. Fewer sugar gliders have been estimated in continuous forests than in fragmented forests across multiple studies (Jackson, 2000a, Lindenmayeret al. , 2021).
The accuracy and precision of density and home range estimates using SECR increases with the number of recaptures. Our data suggest that fish-bait improved recaptures, but not universally, which hindered home range estimation for rarely-observed individuals (i.e. those with home range centres outside the grids). Increased confidence in our estimates may have been attained with a longer study (which may violate closure) or a larger trapping area. For individuals that had sufficient detections, estimated core home range was 1.3ha-1which is narrowly within the range of 1.2-2.4ha-1 in fragmented forests reported by Gracanin & Mikac (2022b). Average movements indicate that sugar gliders in our study travelled approximately 170m per night. The maximum linear movement of 552m we observed is both greater (Suckling, 1984) and less (Gracanin & Mikac, 2022b) than the maximum displacement elsewhere in fragmented forests.
Our captures of new individuals in the fish survey increased substantially, and our capture rates were comparable to the literature. It is feasible that there is a maximum possible capture/detection rate, which may be an artefact of underlying density and behavioural variation of gliders. Other studies have also found 65-80% of sugar gliders known to be alive in a given area are regularly recaptured, but the remainder are rarely detected (Quin, 1995, Suckling, 1984). Indeed, in our study, two gliders that were marked in the week before camera deployment were not sighted again. Our meta-analysis found that trapping outcomes increase with decreased forest connectivity. Across the literature, models of trapping success, defined as repeat captures over time, improved with the inclusion of trap nights and year, in addition to habitat. Our meta-analyses found that captures of unique gliders likewise increased in fragmented habitat, suggesting a density effect, and that unique captures also increased with minimum trap height. This was supported by our findings that detectability improved with greater trap height, even though this was not an important model of density estimation (Table 1).
Our field study makes three key assumptions. The first is that the order of surveys did not impact on the observed differences in detectability. Applying experimental treatments in separate surveys, rather than within surveys has been used effectively by similar studies contrasting control and treatment baits (du Preez, Loveridge & Macdonald, 2014). Nonetheless, there is likely to be some effect of habituation (Gracanin & Mikac, 2022a), and we accounted for this by including temporal and behavioural effects in our models and restraining our sampling duration. Second, we assume the “selfie-traps” did not impact on detection probability, which we cannot rule out. Third, we assume that unmarked individuals were always correctly identified but cannot exclude the possibility of misidentification. We controlled for this by excluding any recordings that were difficult to identify (19.3% of videos) which negatively biased our estimates of density. However, marking gliders prior to the SECR trial increased the efficiency and confidence of resighting gliders on videos.
We have several recommendations for practitioners. First, we recommend fish-oil in sugar glider lures and baits. Furthermore, we encourage the trial of protein-based or novel olfactory lures for other Petaurus spp ., which are also traditionally lured by honey baits (Goldingay & Sharpe, 2004, McBurney, 2019, Stobo-Wilson et al. , 2021). Second, we recommend further study for potential novel attractants, as have been done for other mammals (Heinlein et al. , 2020, Jackson, Hartley & Linklater, 2016, Morgan, 1990). Third, we advise limiting the duration of studies to minimize conditioning behavioural responses (especially if baits are accessible for consumption) and violating assumptions of population closure. Fourth, for studies that do not explicitly identify individuals, we suggest spacing cameras beyond the mean maximum distance moved. Lastly, traps in the mid-strata and above are more likely to capture a higher number of individuals than traps close to the ground, but ultimately, bait type is the most important factor for detectability.
Accurate estimation of density is fundamental in biology and our study shows the benefits of carefully scrutinizing the methods used to achieve this goal by increasing the underlying detection rates. We show that substantial improvements in the accuracy of estimation can be achieved when the methods are tailored to suit the specific biology of the target species. This sometimes may not be obvious from long-established practices in the literature and we encourage other practitioners to re-evaluate the techniques used for detecting species based on contemporary knowledge of their biology. Unattractive baits may mean the difference between success/failure of a study because low detections render density and home range estimates inaccurate or incalculable. This can be a major problem given the importance of good detections to understanding animal-habitat requirements, evaluate threats, prioritize actions and allocate limited conservation resources. Failure to adopt survey approaches suited to the ecology of the target species may inadvertently sabotage projects seeking to gain new ecological insights.