The Australian Shark-Incident Database for quantifying temporal and spatial patterns of shark-human conflict

There are two phases involved in supporting the technical quality of the dataset: (i) the process used by Taronga when collecting data for each shark-bite incident, and (ii) the consistency modifications that we made during manuscript development.

Phase one

For each shark-bite incident, Taronga attempts to contact the most-relevant person involved in the event as possible — e.g., victim, victim’s family, witnesses. Those contacted are asked to complete a questionnaire with information relating to the shark bite. If applicable, questionnaires can also be completed by a fisheries officer in the relevant State and sent to Taronga. Taronga works closely with experts in each State’s fisheries department to validate information sourced in media reports. Each shark-bite case is unique, so the validation process varies depending on details specific to each incident. Forensic shark scientists within each State department are contacted after each shark bite to confirm details related to the incident. For example, validating species responsible for the bite often requires forensic analysis through expert examination of bite marks or artefacts^{21,22,23,24,25}. When available, video footage is analysed for validation of information, such as confirmation of shark species and length.

The database is cross-checked annually for the previous year with the International Shark-Attack File in Florida, USA, as well as with fisheries officers to ensure consistency. The database is cross-checked with the acknowledgement that there are discrepancies between versions due to differences in inclusion criteria. For example, the International Shark-Attack File includes bites where the victim was bitten aboard a boat, whereas the database we present here does not include bites aboard a boat.

We acknowledge the limitations associated with this database, such as differences in reporting over time. For example, incidents might be reported more in recent decades due to technological advances making reporting more accessible or media publicising these events more widely¹⁸. There might also be reporting biases, for example, victims could be more likely to report a bite by a large, potentially dangerous shark (e.g., white, tiger, or bull shark) rather than a smaller, less-dangerous shark species (e.g., wobbegong shark). We also completed a quality assessment of the original database fields and redesigned the data acquisition and entry process (see Phase two) to allow exploration of shark-bite trends and patterns in Australia.

Phase two

We identified errors and inconsistencies in database fields. To avoid additional errors and inconsistencies, and to obtain a quality-controlled database, we redesigned the process for gaining and entering information into the database. This included creating a data descriptor (Supplementary File 2) used as a protocol to inform which questions to ask in the questionnaire. The data descriptor also directs the format of database entries by specifying information required in each field and by indicating the format of each entry (i.e., numeric, descriptive, or categorical). We manually inspected all previously entered data and adapted them to match the data descriptor. We checked each entry using the filter function in Microsoft Excel to identify any spelling and grammatical errors in the fields and ensure that all categories were grammatically identical. We standardised all metrics during this process (e.g., the data descriptor now stipulates that all length measurements should be recorded in metres).

We validated and standardised the geographical locations of shark bites by converting all coordinates into decimal degrees using Microsoft Excel. We subsequently plotted all coordinates using the ggmap library²⁷ in R (Version 4.0.2) (R Core Team 2020) (Fig. 2). We corrected any unusual coordinates (e.g., outside of Australian waters) and crosschecked them with site descriptions and states to ensure validity.

During the development of the data-descriptor protocol, we converted some previous descriptive columns into categorical columns. These columns included (but are not limited to) victim activity, attractant, injury location, injury severity, and weather condition. Converting these columns into categories facilitates analysis to investigate shark-bite patterns. For example, we converted victim activity into a categorical field to restrict answers to the following: snorkelling, motorised boating, unmotorised boating, boarding, swimming, standing, diving, fishing, or other, rather than allowing answers in any format. We used this information to create a time series to show the activity of shark-bite victims in Australia over time (Figs. 3 and 4). Shark bites have increased for boarders (including surfboarding, bodyboarding, kiteboarding, sailboarding, wakeboarding, and stand-up paddle boarding) over time, particularly since 1960 (Figs. 3b, 4). This is likely due to the increase in popularity of board sports, particularly surfing, since the 1960s²⁸. This trend is likely not reflected in Fig. 3a because shark bites are unlikely to be classed as ‘provoked’ during board riding.

The column representing a shark-bite victim’s recovery status also requires a categorical response, restricting answers to: fatal, injured, or uninjured (Fig. 5). Since 1900, the proportion of shark-bite-related fatalities has decreased (Fig. 5). This trend is also true for the three species most attributed to shark-bite-related fatalities, white (Carcharodon carcharias), tiger (Galeocerdo cuvier), and bull (Carcharhinus leucas) sharks. The decrease in shark-bite-related deaths is likely due to advancements in medical responses to shark-bite victims over time¹⁴ and better understanding among surfers about using tourniquets to stem bleeding following increased certification as first responders in workplace occupational health and safety requirements. Bites resulting in an uninjured victim includes interactions where the shark might have bitten the victim’s equipment (i.e., surfboard, bodyboard, kayak) rather than biting the person.

Previously, entries in the injury location column were descriptive. We converted the column to a categorical field restricting answers to the following: arm, hand, lower arm, upper arm, shoulder, neck, head, torso, leg, foot, calf, thigh, pelvic region, or other. During the analyses process, we further categorised these injury locations into four body areas (head, arm, torso, and leg) to assess how injury location affects recovery status (fatal or injured) (Fig. 6). Fatality most often occurred following shark bites to the torso (Fig. 6). This is likely due to the injuries to organs and major arteries resulting in blood loss, which is a leading cause of shark-bite fatalities²⁹. This is the first time that the location of a shark bite on the body has been assessed relative to recovery status.

Understanding how the location of a shark-bite wound relates to victim recovery has value in informing the development of shark-bite mitigations. For example, the development of shark-bite-resistant wetsuits could potentially result in higher survival rates of the user if the fabric is concentrated around the torso region^30,31. Redesigning data acquisition and entry process to allow for categorical columns permits these types of analyses.

Some detail of a shark-bite incident might be lost by converting previously descriptive columns into categorical columns. We addressed this by complementing categorical columns with accompanying fields to retain details of the incident. For example, injury location and injury severity columns are both categorical and allow for user-friendly data analysis, whereas the injury description column is descriptive and provides added detail about the victim’s injuries if applicable. Individually, all three columns address certain aspects of the victim’s injuries, and together, all three columns comprehensively summarise injuries to the shark-bite victim.

Our analysis of the Australian Shark-Incident Database suggests that tiger sharks are proportionally responsible for the most fatalities of all shark species in Australia (38% of all tiger shark bites result in fatality), followed by bull sharks (32% of all bull shark bites result in fatality), and white sharks (25% of all white shark bites result in fatality) (Fig. 6). We emphasise that these figures represent the overall percentage of bites resulting in fatality since 1791 and do not account for possible changes over time. These figures are also proportional to the number of bites by each respective species. In Australia, white sharks are responsible for the largest number of bites on humans (361 total) compared to tiger (229 total) and bull sharks (197 total). At the time of publication, white and tiger sharks were each responsible for 91 and 86 total fatalities on humans in Australia, respectively.

There were 540 incidents in which time of day was recorded. We used these data to assess whether particular shark-bite incidents are more likely to occur at specific times of day (Fig. 7). To standardise reported 24-hour times, we took the location (latitude, longitude) information and reported time and date of each incident using the getSunlightTimes function in the suncalc library in R³² to calculate whether the incident occurred in one of four light-availability categories: dawn, day, dusk, or night. Shark bites occur mostly during the day, which likely reflects time of day when there are more ocean users present. However, there is a slightly higher proportion of bites at dusk for bull sharks compared to tiger and white sharks (Fig. 7). Identifying these trends can assist authorities in developing data-driven educational messaging as a shark-bite mitigation measure. This is important considering that enhanced education is the preferred mitigation measure scored by ocean users in New South Wales³³.

These examples demonstrate that the Australian Shark-Incident Database will be useful for scientists to analyse environmental, social, and biological related shark-bite patterns in Australia. Use of the newly developed data descriptor to standardise future applications and account for quality assurance and control will aid in keeping the database consistent for ease of analysis and interpretation. Ultimately, the publishing of this database will improve our understanding of shark-bite incidents in Australia and will equip us with the knowledge to aim to avoid or predict these events in the future.

Source: Ecology - nature.com