Camera traps show foxes are the major predator of flatback turtle nests at the most important mainland western Australian rookery

Study species

Flatback turtles are an Australian endemic species, nesting only on Australian beaches from Exmouth Gulf, WA, to Bundaberg, Queensland (Qld) (Limpus et al. 1988; Limpus 2007). Flatback turtles on the WA Pilbara coast nest during the austral spring and summer, with females laying an average of ~50 eggs per clutch, and have an incubation period of ~39–48 days (Pendoley et al. 2014).

Site description

This study was conducted across three beaches adjacent to Mundabullangana Station (−20.42°S, 118.05°E), a pastoral lease 70 km south of Port Hedland in the Pilbara region, north-western WA (Fig. 1). The Munda rookery consists of the following three beaches: Cowrie Beach ~3 km in length, Victory Beach ~500 m in length, and Munda Beach ~11 km in length, of which nesting occurs on the eastern 6 km. This is one of the largest flatback rookeries in Western Australia, with an estimated 1800 nesting females per year (Pendoley et al. 2014), and is the only large mainland rookery for the north-western shelf of the continent, between Exmouth Gulf and Port Hedland (FitzSimmons et al. 2020).

Fig. 1.

(a) Location map of Mundabullangana Station, Western Australia. (b) Map of beaches showing camera sites and flatback turtle (Natator depressus) nesting density of beaches.

The Munda rookery is flanked by river systems at the north-eastern end of Cowrie Beach and the south-western end of Munda Beach. The coastal zone is characterised by sheltered low wave-energy sandy beaches, with predominately high secondary dunes interspersed with rocky platform shelves and sparse coastal vegetation. The area has a semi-arid climate, with summer maximum temperatures averaging 36°C and often exceeding 40°C (Bureau of Meteorology Climate Statistics, Port Hedland Airport). The region experiences strong winds through the nesting season, as well as cyclonic activity, that can lead to significant beach erosion (causing termination of Year 1 data collection).

The threat of fox depredation on flatback nests was recognised at Munda in the late 1990s, and 1080 toxin baiting by the Department of Biodiversity, Conservation and Attractions was undertaken on an ad hoc basis from 2002 until 2010 on Unallocated Crown Land along the beachfront adjacent to the pastoral property. Continuation of such control was contingent on monitoring data showing the extent and spatial distribution of turtle nest depredation. Mundabullangana Station staff undertook opportunistic control of foxes on the pastoral lease through ground shooting, and 1080 baiting around water points for dingoes (Canis familiaris; Jackson et al. 2017) up to the 2013/14 season (station staff, pers. comm.). Baits for dingoes have a dosage that is lethal for foxes and therefore may result in a reduction of foxes; however, it is undetermined whether any incidental uptake by foxes has occurred during dingo baiting efforts.

Experimental approach

Two camera-trap studies were undertaken in consecutive years at Cowrie, Victory and Munda beaches at the Munda Rookery. In both years, sites were accessed using a quad bike (all terrain vehicle) to move equipment and cover distances effectively; the bike was driven below the high tide mark and not used during the highest tide peaks. The first year was a pilot study (Year 1) to determine predator species, depredation levels, and whether camera monitoring would be feasible. The second year (Year 2) expanded on Year 1 and included modifications to the experimental design and improved camera set-up to reduce the number of false triggers that are likely to be due to capturing moving sand in the distance, wave action, and sun glare (Fig. 2).

Fig. 2.

(a) Year 1 and (b) Year 2 camera set-up to monitor flatback turtle (Natator depressus) nests at Mundabullangana Station rookery, Western Australia. (a) In Year 1, cameras were attached to a vertical metal stake, ~0.4 m above the sand surface and situated 3 m from the centre of the egg chamber. Vegetation from the field of view had to be removed. (b) In Year 2, cameras were attached near the top of an angled metal stake, ~1.3 m above the sand surface. Minimal vegetation removal was therefore required and there was little sky in most frames, which eliminated false triggers owing to sunrise and sunset. Cameras were orientated in a southerly direction ±45° to reduce glare and wave action (which can cause false triggers).

Data handling

Camera-trap photos were individually viewed as jpeg files and data managed using Microsoft Excel 2010. A photo-capture event was defined as a series of images with a time interval of 30 min between the previous and next group of images of the same species (Dinata et al. 2008). Fox behaviour was categorised as ‘walking through’ (no behaviour directed towards the nest location), ‘sniffing/investigating’, ‘scent-marking’ (urinating or defecating), and ‘depredation’ (digging or consuming hatchlings).

Toward objective (1a), to investigate direct effects of depredation on flatback turtle nests, we analysed the egg fate for 61 ‘known-relocated’ nests monitored in Year 2 (total losses could not be calculated for Year 1 because of incomplete monitoring for the season) that were confirmed by exhuming the marker placed at the nest during laying. We compared nest losses (number of eggs/eggshells counted after hatching as a proportion of the recorded number of eggs laid) for 14 depredated nests (depredation recorded either through direct observation or camera trapping) and 47 intact nests (those where no depredation was recorded) using Mann–Whitney–Wilcoxon Test in R (R Core Team 2018). Toward objective (1b), to investigate potential indirect effects of exposure of the nest chamber and potential thermal change on egg development, we compared proportion of undeveloped/unhatched eggs between depredated and intact nests using a Mann–Whitney–Wilcoxon test. Similalry, to test whether there was an impact of the potential inclusion of debris on hatchling survival, we compared the dead hatchlings (as a proportion of the original eggs laid) between depredated and intact nests using a Mann–Whitney–Wilcoxon test.

Toward objective (2), to identify what potential cues nest predators might have used to locate nests, for 69 nests monitored in Year 2, we used separate Pearson’s χ² tests (calculated in Microsoft Excel) to compare the fate of nests (depredated or intact) by (2a) nest density (high or low; comparing n = 12 nests on low nesting-density beaches, and n = 57 nests on high nesting-density beaches), (2b) human handling (yes or no; comparing n = 21 nests covered by turtles, and n = 48 nests exhumed and re-covered by the experimenters), or (2c) marking method, comparing a subset of n = 62 nests that could be monitored via daily observation, with n = 26 nests being also monitored by camera (camera) and n = 36 nests that had only a wooden stake positioned near them (no camera). Note that we did not have confidence in recording the positions of known nests without using some physical marking method. For all three tests, expected values were calculated assuming that an equal proportion of nests was depredated by category tested.

Toward objective (3), to compare the number of depredation events detected by direct observation or camera trap, we compared the total numbers for 26 nests monitored using both methods by χ² test. For this test, expected values were calculated assuming that an equal proportion of nests was identified as depredated by each monitoring method.

Toward objective (4), to identify temporal patterns in nest-depredation events by foxes, we compared the number of fox depredations for the incubation phase when nests would have contained eggs only (0–37 days post-laying; Pendoley et al. 2014), and for older nests likely to contain hatchlings (38–47 days post-laying) by χ² test, with expected values calculated assuming an equal proportion of depredations per day of monitoring. We also compared fox activity (camera-trap capture rate per 24-h period as the dependent variable), with the day of the field season, year of study, and the presence of people monitoring turtle nesting on the beach during the night (yes or no) as independent factors, by using the glm function in the ‘stats’ package in R.

Statistical significance was accepted at α < 0.05. Values are given as means ±1 s.d. throughout.

(1a) Direct and (1b) indirect effects of depredation on flatback turtle nests monitored for the complete nesting period in Year 2

Over a quarter (28%, 19 of 69) of the monitored nests were depredated. All predation events were attributable to a species on the basis of visual assessment. Of the 19 nest depredations, 17 were initially depredated by foxes and two by Gould’s goannas (Varanus gouldii; Table 4). Multiple depredation events were observed for eight depredated nests (42% of the 19), with some nests being re-visited up to five times. There were two re-visits by goannas on nests that had initially been opened by foxes. A fox revisited one nest that was initially opened by a goanna. There was no record of multiple depredation events for the non-camera nests.

(2) Potential cues that nest predators could use to locate nests monitored in Year 2

(2a) Nest density did not increase depredation risk because there was no significant difference in the proportion of depredated nests (χ12=1.45, P = 0.228) between high-density (14 depredation events of 57 nests; 24%) and low-density (five depredation events of 12 nests; 41%) nesting beaches in Year 2 (monitored for the complete nesting period). (2b) Although it did not reach statistical significance (χ12=2.66, P = 0.103), human handling at the time of laying could increase the risk of predators finding nests, with 33% of nests that were human-covered being later depredated (16 of 48 monitored nests depredated) compared with 14% of turtle-covered nests (3 of 21 monitored nests where egg counts were determined from video recordings depredated). (2c) The presence of a camera did not increase depredation risk, because there was no significant difference (χ12=0.15, P = 0.698) in the proportion of depredation events recorded through daily observation for nests that were also monitored with camera (6 of n = 31 camera nests depredated; 19%) or not (6 of n = 38 non-camera nests depredated; 16%; Table 5).

(3) Number of depredation events detected by direct observation or camera trap in Year 2

For the n = 31 nests monitored simultaneously using both methods, camera traps showed twice the depredation events (39%) compared with 19% of nest predation events detected through daily observation (χ12=6.24, P = 0.012). Cameras and observation both detected depredation events on five occasions, and cameras recorded a further seven depredation events that were not detected by observation (Table 3). One nest depredation was detected through observation but was not detected by camera. Camera failures also occurred with cameras missing fox activity on at least 11 instances (across nine cameras), with fox footprints evident in subsequent images (when other animals triggered the cameras).

(4) Temporal patterns in nest depredation events by foxes

The red fox was the most common vertebrate predator caught on camera across both years, with 235 camera-trap capture events in Year 2, 55% of which showed the animal simply walking past the camera trap, 38% showed sniffing/investigation, 1% showed urination/defacation, and 6% of which showed nest depredation. Foxes were more active in the first half of the night, with peak activity times from 20:00 hours to 00:00 hours (Fig. 4a); most of their activity was therefore likely to be hours old by the time that nest observers commenced their daily monitoring during daylight the following morning. However, foxes were still active on the beach for close to 1.5 h after sunrise, with two capture of foxes occurring in the middle of the day.

Fig. 4.

Timing of flatback turtle (Natator depressus) nest depredation at Mundabullangana Station rookery. (a) Circadian pattern of red fox activity captured on camera traps during Year 2 field season and displayed in a 24-h time format. (b) Incubation age for the 19 depredated nests; only initial depredation data by goanna (Varanus spp.) and red fox (Vulpes vulpes) are shown. Egg phase is the time from when the eggs have been laid to hatching, up to and including Day 37. Hatchling phase is from when the first hatchlings have begun to emerge from the eggs, but are still within the sand, from Day 37 onward. (c) Scatterplot of camera-trap rate for red foxes against day of the field season (the field season was cut short in Year 1 because of a cyclone; nest-fate data from Year 1 are included in analyses if the nest incubation period for that nest was complete). Lines and ribbons represent the line of best fit and 95% confidence intervals.

The earliest depredation event of a turtle nest, on Day 12 of incubation, was by a fox. The last depredation event on a nest (excluding re-visits) was on Day 47 after oviposition (likely to be about the time of hatching), and was also by a fox. Four depredation attempts by foxes occurred during the incubation phase (i.e. 0.1 nest depredations per incubation day), whereas 12 were for nests older than 37 days post-laying (i.e. 1.2 nest depredations per incubation day); the difference in depredation rate between these two periods (Fig. 4b) was statistically significant (χ12=27.57, P < 0.001). There was also a significant positive relationship between fox detection rate and the date of the field season (t₉₆ = 3.13, P = 0.002) (Fig. 4c), further suggesting that greater predation by foxes of hatchlings than eggs was likely to occur. For this multiple-regression analysis, there was also a significant effect of year of study on fox detection rate (t₉₆ = −1.98, P = 0.050), with overall less fox activity in Year 1 than in Year 2, but no significant difference for fox activity depending on whether or not there were people present on the beaches on each night (t₉₆ = −0.26, P = 0.795).

Images were captured on two occasions of foxes with mouths full of eggs as they walked through the frame. Cameras also captured images confirming foxes’ propensity for surplus killing on four occasions; foxes were observed to dig up multiple hatchlings in a nest, kill them as they were dug up, eat some, then gather as many as four in their mouths and depart (Fig. 3b). On this occasion, other dead hatchlings were left on the surface around the nest.

All the bird and reptile predator activity was recorded during the day; 59% of records were within 4 h of daylight and therefore would likely be detected when daily nest monitoring commenced. Three of the goanna depredations occurred during the incubation phase, with two of these events being the initial depredation of a nest, and the third occurred after a fox had already opened the nest (Fig. 4b). The longest hatchling emergence period for a single nest was 6 days, measured from the day a fox dug up hatchlings to when hatchlings on the beach surface above the same nest were predated by a whistling kite, activating the camera (Fig. 3).

(1) Direct and indirect effects of depredation on flatback turtle nests

Foxes were responsible for the greatest confirmed nest depredations, supporting previous studies showing that the red fox is a significant predator of marine turtle nests across its native (MacDonald et al. 1994; Brown and Macdonald 1995; Yerli et al. 1997) and introduced (Limpus 2007; Nordberg et al. 2019) geographic range. The red fox is also a known predator of freshwater turtle nests (e.g. Burger 1977; Congdon et al. 1983; Thompson 1983; Marchand et al. 2002; Spencer 2002; Dawson et al. 2016), and has been argued to represent a significant threat to conservation-significant populations (but see Chessman 2021).

Repeated nest exhumation is likely to increase the total impact of depredation because it exposes the nest chamber and is also likely to attract other nest predators (e.g. Lei and Booth 2018). With 42% of depredated nests being re-visited a second time, and monitored nests re-opened up to five times, an already-disturbed nest may have greater visual or scent cues and therefore an increased risk of further depredation. Similarly, Welicky et al. (2012) and Yerli et al. (1997) reported greater depredation rates for loggerhead turtle nests that had already been opened by predators.

There are also indirect impacts of nest exhumation itself that could increase losses, namely, decreased hatchling emergence for partially depredated nests (Nordberg et al. 2019) and loss of eggs damaged and destroyed in the act of depredation by foxes that were not directly consumed or taken by the predator (Limpus 1971). Although egg losses for disturbed nests were twice those for intact nests, our small sample size precluded statistical significance in this study and we found no difference in egg and hatchling mortality for depredated and intact nests. We also note that the majority of nest depredations we recorded were late-term, when hatchlings were about to emerge or already emerging. Furthermore, the semiarid climate and sand composition at the Munda rookery also readily collapses back into diggings, which can also help preserve viability of remaining eggs.

(2) Potential cues nest predators could use to locate nests

We found no effect of nest density on depredation risk. Additionally, despite camera-monitored nests having additional sensory cues (i.e. infrared flash visual cues and olfactory cues owing to weekly camera servicing), there was no significant difference in depredation rate (detected through daily observations) for camera versus stake-marked nests, suggesting that the presence of cameras did not attract (or deter) foxes. This result supports previous studies that have examined nest-marking methods (Tuberville and Burke 1994; Burke et al. 2005; Strickland et al. 2010). In contrast, Mroziak et al. (2000) and Nordberg et al. (2019) found that the presence of wire cages/mesh may result in increased depredation attempts by raccoon (Procyon lotor), grey fox (Urocyon cinereoargenteus) and yellow-spotted goanna (Varanus panoptes). We could not compare our marked nests with completely unmarked nests owing to a failure to be able to relocate them, but future studies could potentially use high-accuracy GPS or buried markers (e.g. VHF transmitters) to re-find unmarked nests.

It is possible that human handling at the time of laying could increase the risk of predators finding nests, although this result did not reach statistical significance owing to our sample size. It has been shown that foxes will investigate and depredate artificial freshwater turtle nests where there has been any sign of soil disturbance (Dawson et al. 2014), which presumably acts as a cue for some predators. The likelihood of olfactory cues left by researchers is also a possibility that needs to be considered, especially when dealing wth predators that use scent to locate nests.

(3) Number of depredation events detected by direct observation or camera trap

While there was no evidence that camera traps made nests more vulnerable to being located (as indicated from visual observation data), they effectively doubled the number of depredation events detected compared with daily in-person observation, and confirmed the identity of nest and hatchling predators. Daily visitation by observer to record nest depredation events requires that the observer accurately interprets the signs of a predator, including tracks, disturbance of the ground, and egg/hatchling remains. Strong winds and the nesting actions of other marine turtles can obscure or cover predator sign when nests are visited hours later during daylight, and visitation of multiple predators within a 24-h period also cannot be documented. Quantifying the degree of nest depredation is further confounded by the removal of eggs or hatchling remains by birds or crabs, as observed in this study. Recording nest success by observation alone can therefore significantly underestimate depredation rates. Another advantage of camera traps was that it showed the identity of the predator, timing and frequency of their activity, and showed predator behaviour around the nests. Camera trap data also confirmed predation on hatchlings, which is not possible with other means of nest monitoring (e.g. daily observations, track counts in the sand, Lei and Booth 2017).

(4) Temporal patterns in nest depredation events by foxes

We found that foxes were more likely to depredate late-term nests that were more likely to contain emerging hatchlings, which may be due to noise the hatchlings make as they pip their eggs (Guinea et al. 2015) or additional olfactory cues on release of fluids from the eggs (Lei and Booth 2018). We also recorded increased fox presence on beaches late in the season (see also Lei and Booth 2017). Nest depredation records based on observer monitoring (i.e. comparing counts of eggshells with eggs laid) do not capture loss of hatchlings and will therefore underestimate total predation losses. Furthermore, we observed foxes leaving nests with eggs and hatchlings in their mouth suggesting caching behaviour (MacDonald et al. 1994). These findings could suggest that synchronous hatching, while having benefits in conserving energy (Rusli et al. 2016), may assist predators in locating the nest as the hatchlings wait to emerge from below the surface. Our findings are inconsistent with those of Limpus (1971), who claimed that depredation of flatback turtle nests by foxes occurs at all stages of incubation. Our data also differ from studies showing 90% of green turtle (Chelonia mydas) nests at Akyatan Beach in Turkey were depredated by foxes in the first 3 days (Brown and Macdonald 1995), and loggerhead turtle nests were depredated within the first 48 h of laying (MacDonald et al. 1994), when soil disturbance was likely to be an important factor in locating nests (Lei and Booth 2018).

Limitations of this study

Although camera traps proved to be a useful method of monitoring nest depredation and detecting predator activity and behaviour on the beach, they still missed some events and have their own limitations (Meek et al. 2015). Camera traps may fail to capture predation events because of the hot environment, where temperature range was between 25°C and 44°C, reducing contrast between the predator species’ body temperature and the environment. Cameras may be particularly problematic in detecting reptile predators that have a body temperature closer to ambient temperature on warm days, and are therefore likely to underestimate predation by goannas (but see Lei and Booth 2017, and Madden Hof et al. 2020 who, nevertheless, recorded substantial numbers of varanid detections on camera trap). In 11 instances, fox footprints were noticed when another source triggered the camera (e.g. sand movement from a female turtle out of field of view), suggesting fox activity that had not been detected by camera trap. The potential for false negatives means that care needs to be taken when estimating nest depredation events from camera-trap data.

Conclusions and recommendations for management

Like any sea turtle rookery, Mundabullangana will be subjected to multiple pressures. We have quantitatively shown that fox predation of eggs is relatively high at 26% of monitored nests and, with further investigation, this might be refined spatially between low- or high-density beaches (Fowler 1979), and temporally within and among seasons. Total predation rates (by all predators, including foxes, crabs, varanids and birds) would be higher if the predation of live hatchlings during transit from the nest to the water could be added. In addition, anecdotal evidence suggests further damage to nests and beaches due to inundation (e.g. Behera and Kaiser 2020) and cyclonic activity, as witnessed in the present study. All pressures combined mean that the likely total loss of eggs and hatchlings would exceed 30%. Being one the largest rookeries for the flatback turtle globally, and its use as a reference site for a major gas development at another rookery (Gorgon Gas Project), management actions to reduce nest depredation by the introduced red fox should be implemented where is it is feasible and cost effective.

We therefore suggest that long-term monitoring of egg and hatchling predation is established (frequency to be determined) and combined with invasive animal population control. The reduction of fox numbers is a highly feasible and cost-effective action if the main goal is to protect turtle eggs and hatchlings along the narrow coast strip that supports the turtle nesting beaches. Fox mitigation presents no adverse risks to turtles and furthermore provides additional ecological value to the wide area where native fauna populations are generally depleted in the presence of introduced grazing and predator species. Monitoring predation impacts (e.g. Madden Hof et al. 2020) and fox population size will inform an adaptive management approach that ensures that effective and efficient predator control continues into the future.