This study sought to utilise data sourced from a shark diving ecotourism company to study the behaviour of great white sharks (GWS) in Mossel Bay, South Africa and to critically evaluate the value of this alternative data source for scientific research. By removing explanatory variables from the model via backwards selection, we were able to discern that the environmental conditions and the sampling methodology were both significant (p < 0.1) for predicting GWS sightings in Mossel Bay. Yet, as several significant variables retained in the model mirrored the findings of previous studies, the authors are confident that the method of employing a robust hurdle model, allowed elucidation of genuine patterns in these data, whilst countering the potential biases associated with the data collection methodology. It is our conclusion that collaboration with ethical ecotourism companies could have great value for scientific research and offer an alternative, fisheries-independent source for scientific data in the future.
The Effect of Environmental Conditions on GWS Sightings
The moderate to very strong significance of the cyclical spline indicated that both shark presence and shark sighting rate (SSR) followed a cyclical seasonal pattern in Mossel Bay. Yet, the model also showed that environmental conditions did have an impact on GWS sightings. There was weak to moderate evidence that probability of sighting at least one GWS, was correlated to wind direction, minimum air temperature, water visibility and how long the boat stayed at anchor. Comparatively, the SSR was associated with lunar phase, maximum air temperature, seal activity, time of arrival, and wind direction and speed.
As has been previously reported, GWS were sighted year-round in Mossel Bay, South Africa [14, 49, 50, 68]. Throughout sampling, SSR averaged 2.4 ± 2.0 sightings/hour, which was higher than rates previously reported in Mossel Bay [49]. Yet the maturity stages and the sex ratio of the sharks sighted were consistent with those reported in previous studies of the region [49, 50, 68]. Juveniles dominated the demographic structure in these data, supporting the hypothesis that Mossel Bay is not used for GWS parturition or mating [14, 49].
Despite their year-round presence, the strong to very strong correlation of the spline coefficients in this model, indicated that GWS sightings cycled seasonally in Mossel Bay, with shark presence more assured and SSR higher during the winter months (June - August). At many aggregation sites, seasonal peaks in GWS abundances have been attributed to prey availability [7, 10, 15, 39, 46, 48, 69]. Whilst they are generalist predators, after their ontogenetic diet shift, high calorie marine mammals make up an important component of the GWS’s diet [9, 10]. As the cape fur seal pupping season occurs from November to December in Mossel Bay, it seems logical to conclude that the seasonal abundance of GWSs in the region, was caused by the immigration of sub-adult sharks into Mossel Bay, to exploit a seasonally abundant prey resource [17, 35, 51, 69].
Yet, the average size of GWSs sighted in this study was only 2.5 m TL and 81.0% of sharks were smaller than the 3 m TL associated with the ontogenetic diet shift. There have been reports of GWSs smaller the 3 m TL threshold predating seals in South Africa [18], and reporting similar findings, Milankovic et al (2021) hypothesised that Mossel Bay may act as a training ground; with smaller GWSs observing the hunting behaviour of conspecifics in anticipation of their shift to predating marine mammals. It is possible that the seasonal fluctuations in GWS sightings modelled here were driven by the immigration of pre-ontogentic-shift juvenile GWSs into Mossel Bay during the winter months to take advantage of social learning [14].
Yet seal abundance alone may not explain shark presence and the fluctuations in SSRs reported in these data. Seal activity was negatively correlated with SSR and the model highlighted the significance of the sampling location both for predicting GWS presence and counts. The higher probability of observing GWSs at Seal Island compared to Kleinbrak, supports the hypothesis that GWSs immigrated into Mossel Bay to target seal prey. Yet Grootbrak boasted on-average almost double the SSR compared to Seal Island and Kleinbrak. Stomach contents analyses have identified dusky sharks (Carcharhinus obscurus), chub mackerel (Scomber japonicus) and sea bream (Sparidae spp.) make up a significant proportion of the South African GWSs’ diet [10]. Several estuaries in Mossel Bay house nursery habitats for many teleost fish species [49], and the Grootbrak river mouth particularly has been identified as a core habitat for smaller GWS targeting teleost and elasmobranch prey [47]. Therefore, it seems likely that seasonal migrations and/or seasonal breeding of teleosts and other elasmobranchs were also driving GWS immigration into Mossel Bay during the winter months, and that GWS habitat selection based on prey availability occurred on a fine spatial scale within the bay [17, 23, 35, 48, 50].
The significance of several other variables in the model implied that seasonal fluctuations in environmental conditions were also responsible for driving GWS sightings in Mossel Bay to some degree. For instance, modelling indicated a relationship between the minimum air temperature and GWS sightings. Despite their regional endothermy, environmental temperatures have been found to be a significant driver in GWS distributions, as in low temperatures, GWSs must maintain their internal body temperature at a slightly higher metabolic cost [11, 70, 71]. Therefore, these findings indicate that GWSs vacated the study site during cold conditions, to avoid physiological stress. On the other hand, if our conclusions that GWSs immigrate into Mossel Bay during the winter holds true, it seems doubtful that a higher number of sharks were present during periods of warmer maximum air temperatures. Rather it seems these findings indicate that warm conditions had an impact on GWS behaviour [30]. When colder, great whites experience significant physiological changes, including slower metabolic rate and cooling muscles [11, 70, 71]. Under these conditions, with less energy readily available for high-octane displays, GWSs would be less likely to be sighted by observers. Comparatively, in warmer conditions, excess energy would be readily available to mount surface attacks on the bait, making the sharks particularly conspicuous, and resulting in the higher SSR recorded in these data. Therefore, this finding may be highlighting a change in GWS behaviour, as opposed to a direct relationship between air temperatures and GWS abundance.
The arrival time was found to be significant in the counts portion of this model, with highest SSRs reported when sampling began earlier in the morning. GWSs practice a diel pattern of behaviour and habitat use; most actively hunting and patrolling at dusk and dawn, and spending the day at rest, digesting food and conserving energy [7, 29, 47, 51]. Hammerschlag et al (2006) reported that GWS predations on cape fur seals were most frequent during low light intensities and predation success rates dropped dramatically as light increased during the day. High octane hunting behaviours and high activity levels of the early morning would have made GWSs especially conspicuous to observers, leading to higher SSRs during early morning sampling trips. Therefore, the significance of arrival time when modelling shark sightings was likely caused by the diel pattern of behaviour practiced by GWSs.
The model also indicated that GWS sightings fluctuated on a lunar cycle, with higher SSRs recorded when the previous night experienced a full moon. These findings conflict with those of Weltz et al (2013), in Cape Town, South Africa, who reported that GWS sightings off bather beaches increased four-fold during darker new moons. GWSs have been reported to hunt during the night [75] and Klimley et al (2001) have suggested that a full moon may create optimal hunting conditions for GWSs, as the light conditions mean seals become silhouetted against the relatively bright night sky. The increased GWS sightings on days precluded by a brighter night highlighted by this modelling, may have been caused by GWSs concentrating within Mossel Bay to exploit the optimal foraging conditions experienced during the full moon [75, 18, 72].
This model indicated there was a complex relationship between the wind conditions and GWS sightings in Mossel Bay. The presence portion of the model implied a weak correlation between shark presence and wind direction, yet the count part indicated a moderate correlation between wind direction and wind speed with SSR. As ambush predators, surface chop and swell generated in strong winds, would presumably make GWSs more cryptic to their prey at the surface [73, unpublished data]. Therefore, we would expect GWS presence in Mossel Bay to peak during high winds, as sharks would enter the bay to exploit favourable hunting conditions [18]. Yet, this model indicated a negative correlation between SSR and wind speed. It is possible that our model is highlighting a sampling bias here. If observers were unable to perform as rigorously during high winds, low shark sightings may have been recorded even when GWS were present in higher numbers. Yet, the significance of wind conditions in both phases of the model, suggest we should not discount the effects of wind without further investigation, as wind directions may have genuinely had an impact on GWS behaviour. The presence portion of the model suggested there was a slightly lower probability of sighting a GWS when winds were from the west. Yet highest GWS counts were reported for winds from the south and/or west. Winds coming from the south (y-direction) had approximately twice as large an effect on SSR compared to the effect of winds coming from the east (x-direction). In combination, these findings indicate that certain wind conditions were associated with GWS absence in the Bay. During winds from the west, olfactory stimulants would blow from the prey-rich Mossel Bay out into open ocean, potentially attracting GWSs from offshore into Mossel Bay and resulting in the high SSRs recorded. Conversely, winds coming from the east or north, from open ocean, would lack such intensive attractants, potentially resulting in a lower probability of GWS sightings in the bay [51].
Water visibility was also shown to be correlated to GWS sightings in Mossel Bay. GWS presence peaked when waters were very clear, although this effect was substantially less marked compared to seasonality. These findings mirror those of Milankovic et al (2021), who reported that GWS sightings in Mossel Bay peaked with vertical water visibility between 3 and 7 m. Yet, this finding is somewhat surprising, as studies have shown a negative correlation between water clarity and GWS attack frequency and success [17, 18, 20]. It has been suggested that GWSs favour murkier, turbid conditions, because clear water makes them all too visible to their prey [21]. Milankovic et al (2021) hypothesised that, as they are yet to shift to predating upon marine mammals at the surface, the small GWSs observed in Mossel Bay did not attempt to remain cryptic, and so were abundant when waters were very clear. The findings reported here could be caused by the inexperience of the GWSs sighted. Yet, it also seems very plausible that this finding is an artefact of the sampling methodology; with increased visibility improving sampling conditions for the observers and therefore biasing GWS counts to be higher.
Similarly, the stay duration was also found to have a significant impact on shark presence, with shark counts higher during shorter stay durations. This is confounding and is likely an artefact caused by the sampling methodology. As WSAUK assured their tourists a sighting of at least one GWS, the team stayed at anchor for longer periods when shark activity was low. Therefore, higher GWS counts and SSRs recorded during shorter stay durations are unlikely to have been driven directly by GWS behaviour, but by a bias associated with the data collection method [35].
Whilst modelling highlighted some potential biases created by the sampling methodology, it is the authors’ opinion that these data collected from by ecotourism company had great value. The relationships between significant environmental conditions and GWS behaviour highlighted in this model were repeatedly corroborated by scientific research previously conducted in the region. These data can and will be used in future studies of GWS behaviour, with applications ranging from GWS personality, to ethology, and movement ecology and philopatry.
Ecotourism as a Source for Scientific Data
Projects to better understand the ecology and behaviour of endangered species are absolutely critical if we are to develop effective conservation initiatives [1, 45]. Yet, researchers face numerous logistical and financial challenges. Namely finite funding opportunities severely limiting data availability and stunting the advancement of vital knowledge. Collaboration with ecotourism companies could hugely cut the costs associated with data collection for this vital work. Citizen science projects have already been utilised throughout ornithology, ichthyology, palaeontology, and astronomy, and are spreading into include herpetology and botany [74]. Shark diving ecotourism companies are capable of generating colossal data sets, over long-term study periods, throughout the world, and they also have an enormous influence to educate tourists, and generate public support for critical shark conservation measures [34, 39–41, 43, 45].
The nature of data sets sourced in this way, is that they are generally large (larger than could be collected by a single scientific team) and that they range in their reliability depending on the individual who collected the data (as it is likely the majority of participants will not be professional scientists). Therefore, from a statistical standpoint, the biases and inaccuracies generated by the methodology of data collection should be drowned out by statistical power of the large sample size [74]. Yet, the authors strongly recommend that researchers employ powerful statistical methods, which can manage the biases associated with the sampling methodology when conducting their research using ecotourism data. We also advise cultivating an open dialogue with the ecotourism company. During this project, close collaboration with the ecotourism team proved vital in both formatting the historical data and for interpreting the results of statistical modelling.
Ecotourism professionals have a huge amount of experience and knowledge, and amass an enormous amount of contact time with their subject animals in the wild. The WSAUK shark expert spent 12 years observing GWSs in the wild; experiencing an estimated 6000 GWS sightings over approximately 3500 trips. Utilising these expertise will be very beneficial to future scientific research.
WSAUK was a conscientious GWS cage diving company, which adhered to a strict policy of ethics in accordance with South African law [53, 54]. Baits were removed and/or bait lines dropped upon approach, to ensure minimal provisioning, and to dissuade unnaturally aggressive behaviour. Chumming was also only conducted in designated areas. Hence, the data were collected in accordance with strict ethics and dedication to accuracy. If data are to be sourced from other ecotourism companies, we believe it will be imperative to screen and evaluate potential contributors, to ensure that collaborations only occur alongside companies that are as ethical as WSAUK. From an ecotourism professional’s perspective, collaboration alongside scientists could validate their company and elevate public perception of their work.
Ecotourism could also be critical to the conservation initiatives that scientists are advocating for. The enormous local and national capital generated by ecotourism companies may have the power to switch economic gain from extractive industries, towards non-consumptive exploitation of sharks [7, 40, 44, 45]. The inevitable increase in shark sightings that would come with a healthier population size, would benefit both ecotourism businesses and conservationists.
In the future, when employing data sourced from ecotourism in a scientific setting, the authors would advise that the data collection protocol is: 1) designed with care and rigour, 2) structured to incorporate the knowledge and experience of the ecotourism crew (in order to ensure that the research goals are feasible and realistic), 3) data collection sheets are provided for consistency, 4) blank pre-formatted spreadsheets are provided to reduce data formatting times, and )5all aforementioned methods are piloted to ensure adequate training and to allow for revision of poor methodology. With these provisions, there is no reason that data sourced from ecotourism cannot be used for scientific research. If scientists would reach out and collaborate with ecotourism companies, a mutually beneficial alliance could evolve into something remarkably powerful for science, conservation and economy alike.