Sea turtle conservation programs often focus on protecting their nests and hatchlings, as those are the most vulnerable life stages to terrestrial predators, including humans (Marcovaldi and Chaloupka 2007; Lovemore et al. 2020). A relatively common approach is controlling predators near nesting areas, which can be a very effective way of reducing predation rates (Engeman et al. 2002, 2006, 2010; Lei and Booth, 2017b). However, this can be very expensive (Engeman et al. 2002) and requires a sustained effort for many years (Engeman et al. 2006, 2010). Additionally, predator removal raises ethical concerns and can disturb the local ecosystems (Barton and Roth 2008).

To avoid the downsides of predator control, some nest protection techniques have been used; for instance, covering the nests with metal nets has been successfully applied to deter canid predators (Yerli et al. 1997; O'Connor et al. 2017), and placing aluminium mesh cages around recently dug nests had some success in reducing predation by monitor lizards (Varanus spp.) (Lei and Booth 2017a; Hof et al. 2020). Other protection techniques involve disguising the cues predators use to identify the nests, either olfactory, visual, tactile, or a combination of these (Blamires et al. 2003; Oddie et al. 2015; Lei and Booth 2017a).

Several of the monitor lizards that live along coastlines are large reptiles, starting their activity mid-morning and remaining active during the day (Shine 1986). They are generalist predators, adapting their feeding habits to the available food sources (Losos and Greene 1988). Monitor lizards are known to have an acute sense of smell, relying on chemoreception to detect possible prey (Cooper 1989). At João Vieira and Poilão Marine National Park, Bijagos Archipelago, Guinea-Bissau, West Africa, the Nile monitor (Varanus niloticus, Linnaeus 1758) is the main predator of sea turtle nests (Catry et al. 2002; Barbosa et al. 2018). This protected area comprises a small archipelago of 4 islands with contrasting predatorial pressure over the nests, depending on the size of the rookery. On the most remote of the islands, Poilão, which is uninhabited, up to 40,000 green turtle (Chelonia mydas, Linnaeus 1758) nests are laid per year (Catry et al. 2002, 2009; Barbosa et al. 2018). On this island, predation rates (percentage of nests where at least 1 egg was consumed by predators) by Nile monitors are as low as 3% (Catry et al. 2002). There is a second uninhabited island—Cavalos—where more than 2000 green turtle nests were laid in 2016 (Barbosa et al. 2018). On this island, nest predation is much higher, with 1 study reporting 30% of nests predated by Nile monitors in 2016 (Barbosa et al. 2018). Finally, on João Vieira, the only island with permanent human occupation, predation rates are the highest, ranging as high as 76% in 2011 (Barbosa et al. 2018).

In this study, we tested the effectiveness of 3 different sea turtle nest protection techniques to prevent or reduce nest predation by Nile monitors and compared these results with control unmodified nests. We chose techniques that would interfere with the ability of the lizards to either detect the presence of the nests or their ability to reach the egg chamber. We decided to test 3 different methods: a net covering technique, already employed in many nesting grounds around the world (Yerli et al. 1997; Lovemore et al. 2020; Lei and Booth 2017a), and 2 techniques to try and prevent the lizards from detecting the nests—the first hampering chemoreception and the second hampering their ability to visually detect the nests. We also compared their effectiveness with that of similar methods applied in other nesting grounds.

METHODS

Study Site. — This research was conducted on Cavalos (11°0′57.37″N, 15°42′24.17″W). This island has an area of 210 ha and a coastline of 7 km, almost all of it accessible to sea turtles (Barbosa et al. 2018). The main turtle nest predator on the island is the Nile monitor (30% of the nests predated in 2016); the second is the ghost crab (Ocypode cursor, Linnaeus 1758), which predated 6% of the nests in the same year (Sampaio 2018). To date, there is no estimate of the Nile monitor population size for the island; however, the empiric observations of the research team during this work suggest that it is quite large, and monitors were present every day along the entire coast of the island.

Fieldwork started in early August and lasted until mid-September 2016, which coincides with the bulk of sea turtle nesting activity for the area (Catry et al. 2002). Freshly laid nests were marked/tested from 8 August until 4 September. We marked the nests using wood stakes placed about half a meter from the egg chamber and marked with plastic tags. To guarantee randomness in the selection of the nests and to prevent any unconscious bias from the research team, the island was divided in 4 sections according to the north/south and east/west axes. Each morning, work would randomly begin in 1 of those sections and we would sequentially select the nests for the experiment (control, net treatment; scent covering, and track covering) (Fig. 1). Only freshly laid nests were selected (i.e., nests laid on the previous night). Given that monitor lizards become active mid-morning, it was possible to apply the protection techniques on just a few nests each day, in a few hours after dawn, before any signs of monitor activity and before the nests laid that night were attacked.

View Figure

• Download as Powerpoint
• Download Figure

The nests were monitored daily for at least 10 d. The bulk of predatorial activity usually occurs during the first week after a nest is laid (Gonçalves et al. 2007; Leighton et al. 2009, 2011; Barbosa et al. 2018). The 3 additional days after the period when the main predatorial activity was expected to occur were used to encompass potential delaying effects that the experimental treatments might have on predation. Whenever a nest showed signs of predation, the predator and the date of predation were recorded. Predation was judged visually by inspecting nests for the presence of burrows, eggshells, or tracks, but it was common to actually see the monitors attacking the nests during the morning. There are clear differences between the signs left after predation by lizards when compared with the only other known nest predator of the island, the ghost crab (Barbosa et al. 2018), so a misidentification was extremely unlikely. We did not differentiate full predation and partial predation because after a nest is attacked by the first time, the likelihood of it being attacked again is much higher, as there will be more visual and olfactory cues around the nest (Blamires et al. 2003; Lei and Booth 2017b).

Protection Techniques Tested: Scent Covering. — Thirty milliliters of a 5% (v/v) clove essence aqueous solution were sprinkled in the sand on top of the nest and in the area around it. Clove essence can be a deterrent of predation, as the strong scent may superimpose the eggs/ turtle scents (Oddie et al. 2015). This treatment was applied to 26 nests.

Protection Techniques Tested: Track Covering. — The sand over and around the nest, and the tracks leading to and from it, were moved using a spade and a rake to visually disguise the turtle activity. This treatment was performed in 31 nests.

Protection Techniques Tested: Nest Protection. — The nest was covered with a square metal net measuring 1 m² and with a 1-cm² grid centered above the egg chamber. The net was placed 10–15 cm deep and then covered with sand so that the monitors could not use it as a visual cue to detect the nests. This treatment was performed in 26 nests. The metal net was kept over the nests that were not predated until they were at least 10 d old and then removed to allow hatchlings to emerge.

Statistical Analyses. — The spatial arrangement of the nests, subject to the different protection techniques and of the control (n = 64) nests, was tested for spatial autocorrelation for both the initial selection of nests and the predated nests (Table 1).

Table 1. Moran's I test for spatial autocorrelation between the selection of the green turtle (Chelonia mydas) nests to be used in the experiments (row 1) and the green turtle nests predated by Nile monitors (Varanus niloticus) (row 2). In all the tests, H₀ (no spatial autocorrelation) was not rejected.

View Fullscreen

Treatment effects on the number of days until predation (only if predation occurred) were tested with a generalized linear model (GLM; Poisson distribution). We used a 1-way Fisher exact test to compare the observed and expected number of predated nests for each treatment at the end of the 10-d period. Our null hypotheses were that there would be no differences in either the number of days until predation, or in predation at the end of the 10-d period between any of the treatments applied and the control nests.

We used the Relative Risk (RR %) metric to compare the effectiveness of our treatments with other studies. This metric was proposed so that there would be an objective way to compare results between experiments with very different sample sizes, as long as contingency tables were used to measure the effectiveness of the treatments (Khorozyan 2020).

RESULTS AND DISCUSSION

All 3 treatments resulted in the Nile monitors taking more time to find and predate the protected nests than the controls (Table 2; Fig. 2). The difference in the average number of days until predation from the control nests was significant for the metal net treatment (an almost 6-d increase; Z score = 5.91; p < 0.001) and for the scent-covering treatment (a 2-d increase; Z score = 2.22; p = 0.03) (Table 2). This means that even when the treatments do not prevent predation, they delay it, even if just for a few days.

Table 2. Estimated linear model coefficients for the effect of each treatment on the mean number of days it took for a nest to be predated by Nile monitors (Varanus niloticus).

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

At the end of the 10-d period, control nests suffered 30% predation by Nile monitors (19 nests in 64). All treatments led to a similar reduction in nests predated, with predation rates as low as half of those upon the control nests. There was a decrease to a 14% predation on the scent-treated nests, 13% on track-covered nests, and 12% on the metal net–protected nests. The differences in the proportion of predated nests between each of the treatments and the controls were not statistically significant, although for the metal net and the track-covering treatment, they were very close to the 0.05 threshold (Table 3).

Table 3. Predation rates of green turtle (Chelonia mydas) nests by Nile monitors (Varanus niloticus) in the different treatments. The p-values of the 1-way Fisher exact tests are displayed in the last column; these tests compared each treatment with the control (unmodified) nests.

View Fullscreen

The lack of significance of the previous experiment, given its nearly significant value and our significant result in the delaying action of the treatments, is likely due to our small sample size. According to the RR results, all of our experiments resulted in a risk reduction similar to that of other studies (Lei and Booth 2017a; Nordberg et al 2019; Hof et al. 2020), with metal nets being the most effective method (risk of predation was reduced by 61%) and scent covering the least effective (risk of predation was reduced by 52%; Table 3). The reduction described in Lei and Booth (2017a) for the first sampling season of their experience was statistically significant, even though their results for metal nets wielded a smaller risk reduction than the results obtained in our experience. This is probably a result of the high nest predation suffered by their control nests (100%) compared with the relatively low predation of the control nests in our study (30%).

The use of metal nets, which has been very successful for protecting nests from land predators such as canids (Yerli et al. 1997; Lovemore et al. 2020), is not as effective with varanid lizards, probably due to their ecology. While dogs and other carnivores can dig efficiently, they are not burrow dwellers and so are not adapted to dig large burrows or tunnels. On the other hand, monitor lizards, which use burrows daily (Blamires 2001), have longer and more-complex tunnel systems and are perfectly able to dig horizontal tunnels. This gives them the ability to bypass the net because it cannot be placed very deep in order not to disturb the nests. This problem could be reduced with the use of larger nets, as was shown with some success in Queensland, Australia (Lei and Booth 2017a). However, the metal nets that we implemented had a noticeable effect on the time it took for the lizards to predate the protected nests—the average number of days until predation rose from 1.89 ± 0.94 d (SD) to 8.67 ± 4.16 d (SD). Taking into account that Nile monitors predate 75% of the nests within the first week after they were laid (Gonçalves et al. 2007; Leighton et al. 2009, 2011; Barbosa et al. 2018), this may narrow down their window of opportunity.

Metal cages are expected to protect the nests more efficiently than other treatments, including the use of a simpler metal net (O'Connor et al. 2017; Hof et al. 2020) like the one we used. However, comparing the RR of our metal net experiment with the one of the metal cages used by Hof et al. (2020), there is not much added benefit with the cages. The amount of effort to place metal cages is much higher than for simple nets, making this technique unsuitable for rookeries like the one at Cavalos.

Special care should be taken when selecting the material for a net. Logistic and financial matters should be considered, but the possibility that the use of a magnetically active material such as iron or steel might disrupt the hatchlings magnetic imprint, hampering their ability of later returning to the rookery to lay eggs (Irwin et al. 2004), is more important to the livelihood of the hatchlings. An alternative to metal nets would be the use of cheaper and lighter plastic mesh nets, which were relatively effective preventing predation by varanid lizards in Australia (Lei and Booth 2017a) (Table 4). However, the addition of plastic material on a nesting ground raises obvious environmental concerns, and if the net is lost or damaged it might trap turtle hatchlings and other wildlife. Furthermore, erosion will create plastic debris that with time may degrade down to microplastics, both of which are a hazard to sea life, including sea turtles (Lusher 2015).

Table 4. Relative Risk (RR) compared with other similar studies involving sea turtle nest protection. A negative result means that the treatment applied reduced the risk of predation for the nests, a positive result means that the risk increased. Values for RR in bold mean that the reduction in predation was considered statistically significant by the authors of this article.

View Fullscreen

As far as we are aware, track covering has never been tested before as a protection technique for sea turtle nests. There is evidence that visual cues in turtle nests might be used by mammalian predators like racoons (Buzuleciu et al. 2016). Covering the tracks and signs of sea turtle activity might reduce predation rates by making the nests harder to find, although it did not fully prevent Nile monitors from identifying them. Revolving the sand might spread the fluids released by the turtle during oviposition, and these are a strong scent cue for most sea turtle nest predators (Oddie et al. 2015). This means that not only the visual cues are masked by this action, but some scent cues will be as well. It has been reported that some predators rely partially on visual cues to detect turtle nests (Oddie et al. 2015), so this method might make the nests harder to detect, forcing predators to spend more time searching (Table 2) but not preventing them from finding the nests. This technique has the disadvantage of being extremely time-consuming and damaging if the nests are located in dunes, so it is not the best suited for rookeries like Cavalos.

The use of scent masks to protect turtle's nest from predation is a novel approach that tries to focus on the predator's physiology; in this case, the reliance of monitor lizards on scent to read environmental cues (Cooper et al. 2019). This approach, as far as we are aware, has only been tried once to deter monitor lizard predation and, even then, the approach was not to try and disguise the smell but to spread something in the sand that might be unpleasant for the lizards to smell (in this situation, chili powder) (Lei and Booth 2017a). That approach did not yield any significant results; however, their sample size was very small (n = 10 and n = 15 in 2 experiments). In our study, the use of a scent mask to protect the nests lead to an RR of –52%, similar to that of other techniques in other studies (Table 4). As varanids have a very acute sense of smell and use chemosensory cues in prey detection (Cooper 1989), it is expected that scent masking the nests would be an effective method to reduce their detectability, which it did, at least for the first few days. Given that the delay on the timing of the predation was significant, with lizards taking twice as many days to prey on these nests than they did on the control, it would be very interesting to test repeated applications of this deterrent.

A possible factor, not assessed in this study due to its impracticability and that could be affecting predation on treated nests, is the scent left by humans while applying the protection techniques. The inhabitants of the Bijagós archipelago have a particular distaste for Nile monitors and often kill them on site, and so it is probable that the monitors have learned to associate human scent with danger and that may have influenced their likelihood of visiting a human-visited nest.

More than 75% of natural predation by varanid lizards occurs when the nests are less than 1 wk old (Leighton et al. 2009; Barbosa et al. 2018). This timing of varanid predation events supports the theory that the Nile monitors rely mainly on scent cues, because nests remain visible for much longer than a week. Contrastingly, the scent, especially in areas like Guinea-Bissau where the turtle breeding season coincides with the rainy season (Catry et al. 2002), may be washed away quickly. Scent masking may thus be the best of the 3 methods tested to be used as a management tool in large rookeries. Besides being a technique adapted to the behavior of the predator, it is also the cheapest and the least time-consuming. Additionally, it does not require the disturbance of the habitat, so it can be used even in sensitive areas such as small dunes. However, more tests are necessary before applying scent masking on a larger scale. For instance, it is important to test what is the most efficient compound to be used and how much deterrent should be applied on each nest. It might also be appropriate to frequently change the scent mask used, given that varanid lizards in general can learn from association with a stimulus where profitable prey can be found (Firth et al. 2003; Cooper et al. 2019). It would also be important to conduct a more comprehensive study on the cues Nile monitors and other varanids use to detect the nests to adapt the protection techniques to the predator's physiology. Finally, in this study not all the nests were protected, so the treatments might just have diverted the predators to other nests that were easier to find. Further research should include the comparison of predation rates of all the nests before and after protecting a large proportion of a breeding ground.