Date Published: February 27, 2019
Publisher: Public Library of Science
Author(s): Stephen R. Midway, Tyler Wagner, George H. Burgess, Athanassios C. Tsikliras.
Shark attacks are a global phenomenon that attracts widespread attention and publicity, often with negative outcomes for shark populations. Despite the widespread perceptions of shark attacks, trends in human water activities and shark populations are both dynamic, resulting in variable rates of shark attacks over space and time. Understanding variable trends in shark attacks may contribute to a better understanding of risk, and a more tempered response in the wake of an attack. We found that global shark attack rates are low, yet variable across global regions and over decades. Countries with low populations were found to have the highest rates of attack, while countries with high populations (U.S.A., Australia, South Africa) tended to have overall low attack rates, but also much more interannual variability. From the 1960s to the present, those countries with the highest populations also tended to be the places where attack rates have increased. Ultimately, shark attack risk is also driven by local conditions (e.g., time of day, species present); however, a global scale understanding of attack rates helps place risk into perspective and may contribute to a more scientifically-grounded discussion of sharks, and their management and conservation.
Sharks are a diverse group of cartilaginous fishes that have drawn considerable scientific and popular interest. Although most often associated with marine habitats, some species occupy brackish and even freshwater habitats [1, 2], and in many systems shark species are considered meso- or apex-predators [3, 4]. The equilibrium life history strategy  of most shark species leaves them particularly vulnerable to fishing mortality. While sharks are valuable contributors to fisheries around the world , their life-history limitations have contributed to resultant overfishing in most directed shark fisheries and in many multi-species fisheries that feature high bycatch of sharks . Our knowledge of shark population dynamics is limited compared to that of many teleost species, although good examples do exist of shark population monitoring–e.g., Bimini lemon sharks (Negaprion brevirostris), Gulf of Mexico blacktip sharks (Carcharhinus limbatus), and the Australian gummy shark (Mustelus antarcticus). Because of gaps in biological and population data, broad inferences occasionally have been drawn from limited data [8, 9], leading to contrasting evaluations of populations  which unfortunately promotes confusion surrounding the status and trends of individual shark populations. Despite the instances of confusion, there are examples of shark populations that are and are not doing well .
Our approach was to investigate two models for quantifying global trends in unprovoked shark attacks—a model based on the country in which an attack was reported and a model based on a region of a country in which an attack was reported. The regional-level analysis was based on a regionalization of habitats frequented by certain species of sharks. Although the country-level is useful for characterizing broad-scale trends, it often pools multiple habitats, multiple species, and even multiple oceans, which may mask species effects or smaller-scale patterns. For the country analysis, data were limited to the 14 countries with ≥10 shark attacks since 1960 (Table 1). For the regional analysis, country data were subset to include only those countries with ≥100 shark attacks (a value of 100 attacks was chosen to ensure adequate sample sizes within regions): the Commonwealth of Australia, Republic of South Africa and the United States of America. Regions were determined based on the dominant shark species involved in attacks, but also supported by the regional habitat (see Table 2 for regions and dominant species) because positive identifications of attacking sharks can often be difficult and uncertain.
A total of 14 countries were examined in the country-level analysis (Table 1; Fig 1), and seven regions (from a total of three countries) were examined in the region-level analysis (Table 2). Although not the focus of this contribution, the majority of all shark attack outcomes were non-fatal (85%), and typically less than about 25% of attacks were fatal in countries with >50 attacks since 1960.
Predicting individual shark attacks is impossible, and was not our task. We were more interested in long-term trends and investigating regional trends based on dominant species attributed in attacks. There is value in understanding both the regional risks associated with water-based activities, and perhaps more value in being able to interpret reports of high or low attacks in a given year. As expected, regions with very few attacks (but not necessarily low attack rates) tend to not change much annually. In regions with greater numbers of attacks, the probability of one year being different from the year before it can be minor, but also very large; for example, the eastern US/Gulf of Mexico has seen probabilities of annual increases in shark attacks fluctuate from nearly 0 to 1 between just one year, and changes >0.5 are common. Generally speaking, greater raw numbers of attacks means more data and therefore better estimation of uncertainty—you can’t have a very high or low probability of annual change if you are very uncertain about the attack rate because there are few attacks. For example, Australia and the USA both have increasing trends and less uncertainty than other countries, which means that their probabilities of annual increase in attacks is better estimated. The upshot is that areas with large numbers of attacks tend to be the most variable year-to-year.
Overall, trends in global shark attack rates vary substantially, are a function of many social-ecological interacting factors, and remain quite low. Despite the cultural perception of shark attack risk, the risk at larger scales is not very high, and where it is increasing the rates are low and preventative measures are more likely to take place. However, in highly populated regions—like the Eastern USA and Southern Australia—shark attack rates have doubled in the last 20 years, and while the rates remain relatively low, they should continue to be monitored. In locations with high raw number of attacks, any given year (or season) with an elevated or decreased number of attacks might not be that unusual, given the nature of any Poisson-distributed responses in which the variance increases directly to the mean. Ultimately, all shark attacks are local—risk is best assessed at the spatial and temporal scale of the beach or water body intended to be used and known risk factors at those scales should be considered over larger-scale patterns. Our work has shown that in some global locations, shark attacks are increasing over long time periods and are highly variable, and we recommend additional research at multiple scales to better articulate possible scale-dependent predictors of shark attacks.