Incidence of shark‐inflicted bite injuries on Australian snubfin (Orcaella heinsohni) and Australian humpback (Sousa sahulensis) dolphins in coastal waters off east Queensland, Australia

Abstract The ecology and evolution of prey populations are influenced by predation and predation risk. Our understanding of predator–prey relationships between sharks and dolphins is incomplete due to the difficulties in observing predatory events directly. Shark‐inflicted wounds are often seen on dolphin bodies, which can provide an indirect measure of predation pressure. We used photographs of Australian humpback and snubfin dolphins from north, central, and south Queensland to assess the incidence of shark‐inflicted bite injuries and to examine interspecific differences in bite injuries and their relationship with group sizes, habitat features, and geographical locations characteristic of where these individuals occurred. The incidence of shark‐inflicted scarring did not differ between species (χ 2 = 0.133, df = 1, p = .715), with 33.3% of snubfin and 24.1% of humpback dolphins showing evidence of shark bites when data were pooled across all three study sites. Generalized additive models indicated that dolphins closer to the coast, with greater photographic coverage, and in north Queensland were more likely to have a shark‐inflicted bite injury. The similar incidence of shark‐inflicted wounds found on snubfin and humpback dolphins suggests both are subject to comparable predation pressure from sharks in the study region. Results highlight the importance that habitat features such as distance to the coast and geographical location could have in predation risk of dolphins from sharks, as well as the importance of considering photographic coverage when assessing the incidence of shark‐inflicted bites on dolphins or other marine animals. This study serves as a baseline for future studies on shark‐dolphin interactions in Queensland and into how predation may influence dolphin habitat usage, group living, and behavior.

The coastal waters in which these two species reside are also inhabited by several shark species known to prey on small cetaceans. In Queensland, tiger (Galeocerdo cuvier), bull (Carcharhinus leucas), and white (Carcharodon carcharius) sharks overlap in spatial distribution with snubfin and humpback dolphins (Green et al., 2009;Heithaus, 2001b;Heithaus et al., 2017;Monteiro et al., 2006). The probability of predation of small dolphins by sharks is ultimately influenced by the predator's ability to encounter, ambush, and overpower its prey, and the prey's ability to detect, avoid, and escape its predator (Heithaus et al., 2009;Martin & Hammerschlag, 2012). Prey species have evolved a variety of strategies to combat the threat of predation, including active defense (fight or flight; Lima, 1998;Lima & Dill, 1990), grouping (Clark & Mangel, 1986;Norris & Dohl, 1980) and predator avoidance through changes in habitat use , 2006. There are, however, environmental factors such as water depth Long & Jones, 1996) and turbidity (Heithaus, 2001b;Turesson & Brönmark, 2007) influencing the preys' detection abilities and the predators' ambush abilities, and thus ultimately affecting the success of predation attempts by sharks (Martin & Hammerschlag, 2012). Similarly, habitat characteristics are also likely to influence predation success. Waters adjacent to the coast could have a higher density of predators and hinder the detection ability of dolphins (Cameron, 2010;Heithaus, 2001a;Heithaus et al., , 2007Meyer et al., 2009), and hence pose a more dangerous environment for dolphins. Habitats close to estuaries and river mouths may be more dangerous habitats for dolphins due to the spatial overlap with predatory species frequenting this environment such as bull sharks (Heupel & Simpfendorfer, 2008;Melillo-Sweeting et al., 2021). Additionally, anthropogenic factors such as hunting and culling of sharks might also influence predation of dolphins by sharks, altering the composition of both predator and prey populations and encounter rates (Baum et al., 2003;Holmes et al., 2012).
In this study, we used photographic evidence of shark-inflicted scarring on individual snubfin and humpback dolphins from northern (Cleveland Bay and Halifax Bay), central (Bowen), and southern (Keppel Bay and Gladstone) Queensland, Australia ( Figure 1) to (1) assess the prevalence of shark-bite scars on snubfin and humpback dolphins; (2) assess if shark bite presence on dolphins differs among dolphin species, study sites and environmental variables (water depth, distance to coast, distance to estuary) associated with dolphin habitat use; and (3) identify which dorsal region where most shark-inflicted scarring occurs. Based on existing knowledge of predator-prey relationships, relative shark abundance (i.e., shark catches per unit effort), and the ecology of both delphinids and sharks, we predicted that the incidence of shark bites on dolphins would be greater in (1) northern Queensland as this area tends to have a larger shark population (indicated by higher catches per unit effort) than southern areas and thus an expected higher dolphinshark encounter rate; (2) shallower water and waters close to the coast and estuaries due to the preference for such environments by predatory shark species, hence higher encounter rate; (3) greater for snubfin dolphins due to their habitat preference for shallower water and waters close to estuaries; (4) greater with more photographic coverage due to the increased likelihood of observing a shark bite; and (5) less incidence of shark bites with increasing group size as this can improve detection of predators or dilute predation risk.

| Study sites
To measure the incidence of shark-inflicted bite injuries on snubfin and humpback dolphins, we used digital photographs of the two species collected across three study sites in Queensland (north, central, and south, Figure 1).

| Photo-identification of dolphins
Photographs of the dorsal side of dolphins identified individuals where possible, primarily using the dorsal fin shape, nicks, and scars (Würsig & Würsig, 1977), as well as loss of pigmentation in the upper region of the dorsal fin (Hunt et al., 2017). Only photographs considered excellent or good quality of dorsal fins with distinctive markings were used for the identification of individuals, development of the catalog and analysis (Parra et al., 2011;Würsig & Jefferson, 1990).
Images were then checked using DISCOVERY software (Gailey & Karczmarski, 2012) to be matched with individuals in the catalog.
Only marked individuals in the photo-identification catalogs were used for the analysis of the evidence of shark bites to ensure each individual was only counted once (see Parra et al., 2006 for further details).

| Presence of shark bite scars
We reviewed the capture history of each individual dolphin in the photo-identification catalogs to source multiple images of each individual's dorsal region and for assessments of shark-inflicted scarring. Scarring attributed to sharks is generally crescentshaped, jagged, and consisted of widely spaced tooth marks (Heithaus, 2001b;Scott et al., 2005;Smith et al., 2018). In the analysis, we did not include scarring that could not be clearly attributed to sharks, such as notches, linear scars and narrowly spaced, shallow rake marks. When shark-inflicted scars were identified, they were assigned to the body region they covered ( Figure 2) and the respective side of the animal (left or right). If individuals had shark bites in more than one region, one region was selected randomly to include in the analysis. We attempted to identify the species of shark responsible for the scarring using the conformation of wounds and spacing between teeth, however, decided against including this due to the unreliability of such methods.
We estimated the photographic coverage of the dorsal side of each individual in the photo-identification catalogs, regardless of the presence of shark bites, by recording which body regions (as indicated in Figure 2) had been photographed, and then calculating a percentage of the dorsal side of each individual photographed.
Photographic coverage was explored as a variable ( Figure A1), and then only individuals with ≥60% of their dorsal body photographed were selected for analysis to standardize the comparison of sharkinflicted wounds among individual dolphins and to minimize bias in the incidence of shark bites towards individuals with greater photographic coverage.

| Data analysis
All analyses were done in R version 4.0.2 (R Core Team, 2022).

| Univariate analysis
Average water depth, group size, distance to coast, distance to estuary, and photographic coverage were calculated using the average of all sightings of that individual across the study period (using the capture history) to ensure that a representative range of habitat use of each individual was reflected (Table 1). We examined the relationship between each explanatory variable (Table 1) and the incidence of shark-inflicted wounds using a chi-squared with Yates' continuity correction test to assess differences in shark bite incidence between dolphin species and study sites. A Fisher's exact test for count data was used to compare differences in shark bite incidence across the left and right side, as well as the different body regions of dolphins.
Randomization tests were used to compare the mean of each predictor variable (average water depth, group size, distance to coast, distance to estuary, and photographic coverage) between individuals with and without shark-inflicted scarring.

| Generalized additive modeling
We used generalized additive modeling (GAM; Hastie & Tibshirani, 2017) to model the relationships between the presence of shark-inflicted scarring and a suite of predictor variables including dolphin species, group size, photographic coverage, water depth, distance to coast, distance to estuary, and study site (Table 1). As the central study site only had five individuals with ≥60% photographic coverage, this site was excluded from models to avoid model overfitting. Correlation between variables was checked using Spearman's rank correlation test and by calculating the variance inflation factor using the udsm package (Naimi, 2015), with no correlation found between the variables. We standardized numerical data prior to analysis using the STANDARDIZE function in Excel (Microsoft Corporation, 2022), returning a normalized value (z-score), to allow for interpretation of the relative strength of parameter estimates in the averaged model (Grueber et al., 2011).
A total of 128 GAM models were built with binomial distribution and a logit link function using the mgcv package (Wood, 2001), including the null model, using all possible combinations of predictor variables. To prevent overfitting, gamma was set to 1.4 (Wood, 2017). Models were ranked using Akaike's information criterion corrected for small sample size (AICc) and final models were checked for patterns in the residuals. We adopted an informationtheoretic approach (described by Burnham & Anderson, 2002) and averaged the top competing models (ΔAICc < 1, as recommended by Burnham & Anderson, 2002). The sum of Akaike weights was then calculated for averaged top models using the qpcR package (Spiess, 2018) to determine the importance of the predictor variables. ( Table A1). Most animals included in the analysis were sighted several times (mean ± SE = 3.2 ± 0.24, range = 1-10 sightings), and throughout the study period (mean ± SE = 224.7 ± 28.4, range = 0-765 days; Table A2).
The average water depth, group size, distance to coast, and distance to estuary at which individual dolphins with photographic coverage of ≥60% with and without shark scars were sighted, as well as their photographic coverage, did not differ (randomization test, all p > .05; Table 2, Figure A2).

| Generalized additive modeling
GAM modeling of individuals with ≥60% photographic coverage in the north and south study site returned eight models within 1 delta AICc, including the null model. The top models (ΔAICc < 1.0) are listed in Table 3.
In general, there was an increase in the likelihood of an individual having a shark bite with increased photographic coverage and decreased distance to coast (Figure 4), with individuals being more likely to have shark-bite injuries in the north study site. The sum of weights of the averaged top models (ΔAICc < 1.0) suggested that for individuals with ≥60% photo cover, the presence of a shark-inflicted bite was best predicted by study site, average photographic coverage, and average distance to coast (Table 4). The deviance explained was extremely low for all models, suggesting that the variables included here are not sufficient to explain our data and there are other factors at play.

| Location of shark bites
There were no differences in the incidence of shark-inflicted bite

| DISCUSS ION
The ecology, evolution, behavior, population dynamics, and community structure of prey populations are influenced largely by preda-  (Heithaus, 2001b;Smith et al., 2018), and hence, our observations here should be considered a minimum estimate of predation pressure.
We acknowledge that the small sample sizes in this study may limit the generalisability of the findings and the statistical power of our analysis. Although the sample sizes were small, snubfin, and humpback dolphins occur at very low densities (Parra & Cagnazzi, 2015;Parra, Cagnazzi, & Beasley, 2017;Parra, Cagnazzi, Perrin, & Braulik, 2017), and thus, we believe that the findings provided here represent robust patterns and insights into shark bite prevalence on these species along the east coast of Queensland.
Analysis of shark-bite scars on the dorsal body of snubfin and humpback dolphins suggests that both species are subject to predation from sharks, that predation pressure is similar across the two species and appears to be influenced by distance to coast and the geographic location along the coast. Additionally, our analysis highlights the importance of considering photographic coverage when assessing the incidence of shark-inflicted bites on dolphins or other marine animals.

| Interspecific differences
Analysis of photographs from the dorsal regions of the body of snubfin and humpback dolphins indicate that both are subject to predation attacks by sharks. We found predation pressure (as inferred from the prevalence of shark-inflicted bite injuries) to be consistent between the two species in coastal waters of east Queensland, Australia. Interspecific variation in the incidence of shark bites on dolphins could be linked to their habitat use patterns as well as differences in shark abundance, shark sizes, or food availability among study sites. In Queensland, snubfin dolphins prefer shallower waters (1-2 m), occur closer to river mouths, and form larger groups than humpback dolphins (Parra, 2006 (Paterson, 1990), and there has been a decline in the number and average size of sharks because of culling (Holmes et al., 2012). In northwestern Australia no major shark  (Braccini et al., 2020). Therefore, snubfin and humpback dolphins in northwestern Australia may be subject to higher shark predation risks than those in Queensland and, thus, the difference in shark bite prevalence between studies. Furthermore, differences in sample size between the two studies (Western Australia = 152 snubfin and 26 humpback dolphins, Queensland = 56 snubfin and 36 humpback dolphins) may have also contributed to the contrasting findings.
Future research on the habitat use and spatial preferences of snubfin and humpback dolphins in north Western Australia, as well as the shark abundance across different study sites, should elucidate why predation risk of the two species differs between the populations in eastern and northwestern Australia.

| Distance to coast
Although distance to coast for individuals with and without sharkinflicted scarring was not different between the two groups, the likelihood of an individual bearing shark-inflicted scarring increased the closer it was observed to the coast, supporting the hypothesis that the incidence of shark bites would be greater close to the coast.
Coastal regions are productive areas, with combinations of estuary output, nutrient run-off, and upwelling increasing productivity and food availability in these areas (Webb, 2021). For example, Cleveland and Halifax Bays in the north site are productive mangrove habitats, supporting large populations of teleosts and attracting both sharks and dolphins to feed (Robertson & Duke, 1987, 1990Simpfendorfer & Milward, 1993). Additionally, due to the abundance of food, sharks often use coastal areas, such as Cleveland Bay, as a nursery habitat (Simpfendorfer & Milward, 1993). More sharks and dolphins in areas close to the coast would increase encounter rate (Heithaus et al., 2009) and presumably the risk of predation on dolphins. Therefore, it could be expected that dolphins occurring closer to the coast would be more likely to have shark-inflicted scarring as they are exposed to greater predation pressure in these areas. It is also possible that the selection of areas close to the coast happens after predation attempts have occurred; however, these species generally use shallow, estuarine, coastal areas along the east coast of Queensland (Parra, 2006).

| Study site
Snubfin and humpback dolphins were more likely to have shark-inflicted scarring in the northern study site. This may be due to differences in the relative shark abundance in these areas, with the abundance of predatory species such as tiger sharks having declined more in south Queensland compared with north and central Queensland (Holmes et al., 2012). Additionally, the north site (Cleveland Bay and Halifax Bay) is recognized as a nursery area for predatory species of sharks including tiger sharks (Simpfendorfer, 1992;Simpfendorfer & Milward, 1993). Areas with a higher abundance of predators would pose a greater risk of predation for dolphins due to a higher encounter rate (Heithaus, 2001a), hence it could be expected that dolphins occurring in the north study site would face greater predation pressure F I G U R E 4 Partial effect plots generated for each variable shown to be influential in the eight top (delta AICc < 1) models, relating the relative influence of (a) distance to coast (m) (negative), and (b) photographic coverage (%) (positive) on sharkbite presence. Solid lines are the fitted linear models. Shaded areas are approximate 95% confidence intervals. Data were standardized, representing the number of standard deviations a given data point is from the mean.  and, therefore, have a higher incidence of shark-inflicted bite injuries.
To infer why study site was an influential variable on the likelihood of an individual having a shark-inflicted bite injury, future studies should assess additional variables for each study site including shark size and abundance, as well as the health of ecosystems and the influence of urbanization and overfishing at each site on shark abundance.

| Location of shark bites
We found that the majority of shark bites on snubfin and humpback dolphins were in the mid-flank region, followed by the anterior and anterior peduncle regions. The mid-flank and dorsal fin regions of dolphins are the most commonly photographed body part due to the surfacing pattern of dolphins, with the remaining dorsal region (e.g., anterior, anterior peduncle, and posterior peduncle) photographed less often (refer to Figure A3 for photographic coverage of individuals). It is possible that fewer bites were observed in the anterior and posterior regions due to lack of photographic coverage of these areas. Furthermore, bites to the anterior and posterior peduncle are more likely to be lethal as they target vital organs and sever the tailstock, immobilizing dolphins and allowing sharks to finalize the kill (Cockcroft et al., 1989;Mann & Barnett, 1999;Smith et al., 2018;Turnbull & Dion, 2012). Therefore, scarring in these areas would not be observed as often on live animals compared with bites on the midflank region, with dolphins more able to escape and recover from bites to this area.

| Photographic coverage
Photographic coverage did not differ between individuals with and without shark-inflicted bite injuries; however, it was included in four of the eight top models, with a positive relationship to the likelihood F I G U R E 5 Percentage (%) of sharkinflicted scarring on each body region (anterior, mid-flank, dorsal fin, posterior peduncle, and anterior peduncle) of (a) Australian snubfin (Orcaella heinsohni) and (

| Additional factors
We found no difference in distance to estuary, water depth, or group size between individuals with and without shark bites. This suggests that dolphins face equal predation pressure across different distances to estuaries, depths, and group sizes in these areas, that dolphins did not change their habitat or grouping behavior after being attacked or that factors other than predation risk are influencing habitat selection and behavior of dolphins, such as prey availability.
Additionally, the small sample size of this study may potentially limit the power to detect the influence of these variables on predation pressure.

ACK N OWLED G M ENTS
We would like to thank all the research assistants and volunteers who participated in the collection of data for this study.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data supporting the results of this study will be archived in https:// figsh are.com/s/7cd05 c4136 84a19 b9e5d.
TA B L E A 2 Individual ID, number of sightings, and the time lag between the first and last sighting of individual Australian snubfin (oh) (Orcaella heinsohni) and Australian humpback dolphins (ss) (Sousa sahulensis) with photographic coverage ≥60%.

Number of sightings
Time lag between first and last sighting (days)