Discovery of a multispecies shark aggregation and parturition area in the Ba Estuary, Fiji Islands

Abstract Population declines in shark species have been reported on local and global scales, with overfishing, habitat destruction and climate change posing severe threats. The lack of species‐specific baseline data on ecology and distribution of many sharks, however, makes conservation measures challenging. Here, we present a fisheries‐independent shark survey from the Fiji Islands, where scientific knowledge on locally occurring elasmobranchs is largely still lacking despite the location's role as a shark hotspot in the Pacific. Juvenile shark abundance in the fishing grounds of the Ba Estuary (north‐western Viti Levu) was assessed with a gillnet‐ and longline‐based survey from December 2015 to April 2016. A total of 103 juvenile sharks identified as blacktip Carcharhinus limbatus (n = 57), scalloped hammerhead Sphyrna lewini (n = 35), and great hammerhead Sphyrna mokarran (n = 11) sharks were captured, tagged, and released. The condition of umbilical scars (68% open or semihealed), mean sizes of individuals (±SD) (C. limbatus: 66.5 ± 3.8 cm, S. lewini: 51.8 ± 4.8 cm, S. mokarran 77.4 ± 2.8 cm), and the presence of these species over recent years (based on fishermen interviews), suggest that the Ba Estuary area is a critical habitat for multiple species that are classified as “Near Threatened” or “Endangered.” Specifically, the area likely acts as a parturition ground over the studied period, and potentially as a subsequent nursery area. We identified subareas of high abundance and found that temperature, salinity and depth acted as small‐scale environmental drivers of shark abundance. The data suggests a tendency for species‐specific spatial use, both horizontally (i.e., between sampling areas) and vertically (i.e., across the water column). These results enhance the understanding of shark ecology in Fiji and provide a scientific basis for the implementation of local conservation strategies that contribute to the protection of these threatened species.

The establishment of marine protected areas is a popular conservation strategy that has been shown to support shark populations, or at least to mitigate detrimental human activities in critical nearshore areas (Aburto-Oropeza et al., 2011;Henderson, Jourdan, & Bell, 2016;Knip, Heupel, & Simpfendorfer, 2012).
Selecting appropriate locations, however, requires the identification of shark habitats, which may not only differ between species and across regions, but may also shift with the requirements of certain life-history stages (Grubbs, 2010;Ward-Paige, 2014).
Such basic information is often scarce, particularly in many rural and developing coastal areas. One example of such a region are the Fiji Islands. At least 17 shark species are known to occur in Fijian waters (Seeto & Baldwin, 2010), but generally little is known about exactly where different species concentrate, and how and when they make use of the available habitats. Using data collected from dive operators, citizen scientists and local fishermen, an increasingly clear picture of shark species abundance throughout Fiji is emerging (Brunnschweiler, Abrantes, & Barnett, 2014; Glaus, Adrian-Kalchhauser, Burkhardt-Holm, White, & Brunnschweiler, 2015;Rasalato, Maginnity, & Brunnschweiler, 2010;Ward-Paige, 2014). Specific locations with confirmed species occurrence in the scientific literature are only available for Viti Levu (Brown, Seeto, Lal, & Miller, 2016;Brunnschweiler & Earle, 2006;Cardeñosa, Glaus, & Brunnschweiler, 2017;Marie, Miller, Cawich, Piovano, & Rico, 2017) and Vanua Levu (Goetze & Fullwood, 2013), the two largest islands of Fiji. In the former case, this has led to the establishment of the Shark Reef Marine Reserve, Fiji's first national marine park, and the Fiji Shark Corridor which comprises approximately 30 miles of coastline (Brunnschweiler, 2010).
The main threat to sharks in Fijian waters is their frequent occurrence in the bycatch of artisanal and subsistence fisheries in the inshore fishing grounds (Glaus et al., 2015). This includes not only coastal waters but also rivers and river deltas, as shown by Rasalato et al. (2010) who collected interview-based evidence of shark occurrences in all of Fiji's rivers. Ecological studies recently confirmed the usage of riverine and estuarine habitats by juvenile sharks in both the Navua and Rewa River in southern Viti Levu (Brown et al., 2016;Cardeñosa et al., 2017). There is currently no systematic data on shark occurrence in estuaries on the northern coast of Fiji's main island Viti Levu.
Thus, in this study we investigate for the first time the Ba Estuary on the northern coast of Viti Levu and aim to assess (a) the composition of shark species occurring in the area, (b) their abundance and life-history stages, (c) spatiotemporal differences in habitat use over 4 months, and (d) environmental drivers of abundance. Furthermore, through semi-structured interviews with local fishermen, we provide socio-economic context that also explores community support for potential management options.

| Study site
The study was conducted in a shallow bay environment (depth < 15 m) in north-western Viti Levu, the main island of the Republic of Fiji ( Figure 1). The sampled area around the Ba River mouth is part of a larger bay that is sheltered from the open sea by patches of fringing reefs and from the mainland by mangroves. The sea bottom predominantly consists of muddy substrate and seagrass beds. The area is under strong tidal influence, with a tidal range of approximately 2 m (www.tide-forecast.com, 2016). There is activity by artisanal and subsistence fishermen from surrounding villages in the estuary. While sharks are by tradition not explicitly targeted in fishing operations, they regularly occur as low-value bycatch.

| Sampling methods
Over 6 days in November 2015, a pilot shark-fishing survey was conducted, informed by participatory mapping with local fishermen who could indicate spots where they had previously caught sharks. The pilot study consisted of 26 gillnet deployments without a clear spatial sampling scheme in order to test the sampling methodology and procedure, and to identify suitable areas for sampling within the Ba Estuary. Deployments were conducted between 16:00 and 02:00, at varying tides and with checking intervals of 15-25 min. Total gillnet soak time of the pilot survey was 30.2 hr, during which a total of 12 sharks where caught. The catch was comprised of nine C. limbatus (65.5 ± 3.8 cm; seven males, two females; umbilical scar condition: five open, three semihealed, one healed) and three S. lewini (51.9 ± 0.7 cm; one male, two females; umbilical scar condition: one open, two semihealed). These sharks were not included in any further analyses.
Based on the results of the pilot survey, seven 1.13 km 2 circular sampling areas in the immediate vicinity of the river mouth were selected. Each sampling area featured contrasting environmental features (e.g., depth, distance to mangroves, turbidity) and overlapped with areas that local fishermen identified as having higher shark abundances. Sampling areas 1-6 contained 10 sites each, and area 7 contained nine sampling sites ( Figure 1).
The main shark-fishing survey was conducted on 26 days from December 2015 to April 2016. Bottom-set gillnets and longlines were set at depths ranging from 1 to 15 m in the seven sampling areas, with a total of 73 and 30 deployments, respectively. All deployments were carried out between 18:00 and 02:00 from a 7 m fiberglass boat with a 40 HP engine. Two assistants and a captain were present at all times. Bait used on longlines consisted predominantly of Indian mackerel (Rastrelliger kanagurta), and occasionally of red snapper (Lutjanus argentimaculatus), squid (Loligo sp.), and mullet (Mugil cephalus). Up to two gillnets (100 m length and 3 m width, ~10 cm mesh size) were deployed simultaneously with a soak time of 1-6 hr. To minimize animal stress and mortality, gillnets were regularly checked in intervals of 15-25 min. When feasible, a longline (75 m) with 27 hooks was additionally deployed at the same sampling sites to assess potential catch differences attributed to gear selectivity. Distance between gangions attached to the floater line varied from 2.4 to 2.8 m. Gangion length ranged between 0.6 and 3 m, with the last 0.5 m consisting of 1.5 mm steel wire and a baited 13″ circle hook. In total, fishing effort ranged from 6 to 10.36 hr (longline) and 15 to 23.08 hr (gillnet) per sampling area. Total soak time of gillnet and longline deployments varied from 30 min to 6 hr, and from 45 min to 3 hr, respectively. Sampling effort was intended to be uniformly distributed among the seven sampling areas and ranged from 24.5 to 33.3 hr/area (mean: 28 ± 3 hr/area).

| Shark handling
All captured sharks were sexed and tagged with an internal Passive Integrated Transponder (Beijing KingDoes RFID Technologies Co., Ltd., China), as well as an external nylon spaghetti tag (Hallprint Pty. Ltd., Victor Harbor, Australia) below the first dorsal fin. Sharks were examined for umbilical scar condition (open, semihealed, healed) and measured (see Supporting Information Appendix S1 for example of open scar). Measurements were taken by placing the shark laterally on a 10 cm wide wooden board with measuring tapes attached to either side. Precaudal and fork length were read at the lower tape, while stretch-total-length was read from the upper tape by stretch-

| Environmental data
To determine differences in abiotic conditions between sampling sites, and to characterize their influence on shark abundance, we measured a variety of environmental parameters selected in accordance with previous studies on drivers of habitat selection of juvenile sharks (Yates, Heupel, Tobin, & Simpfendorfer, 2015), including tide, which may also affect shark movement (Ackerman, Kondratieff, Matern, & Cech, 2000;Wetherbee, Gruber, & Rosa, 2007). Depth was recorded at the beginning and end of each gear deployment using a weighted rope and taken as the mean of both measurements. To measure turbidity, a Secchi disk was lowered in the water column until it became indistinguishable. In case of darkness, a headlight (LiteXpress liberty 120 sensor) was used to assist in determining depth. Salinity (PSU), and sea surface temperature (°C), were measured using a Manta 2 (Eureka to Water Probes, www.
waterprobes.com). Furthermore, tide was assessed and categorized into either (a) incoming or high, or (b) outgoing or low. Geographic coordinates were determined using a Garmin Etrex 40 at the beginning and ending locations of a catch event. Distance of sampling site to mangroves (km) were calculated in QGIS 2.14.3 (Essen, www. qgis.org/de/site/forusers/download.html) using the distance matrix tool to measure a straight line from each sampling site to the nearest mangrove polygon. Before all executions, the coordinate system was set to EPSG:3141, Fiji 1956/UTM zone 60S. Due to logistical constraints, a complete set of environmental parameters could only be measured in 67 of the 103 deployments.

| Analysis
We supply a descriptive analysis of species abundances in relation to sampling site and month, and of shark biodata (sex, length, umbilical scar condition). Furthermore, we statistically compared the shark catch per unit effort (CPUE) per deployment between sampling areas with a Kruskal-Wallis test, due to the non-normal error structure of the CPUE data. For post hoc pairwise comparisons between areas, we used the non-parametric multiple comparisons procedures provided in the R package nparcomp (Konietschke, Placzek, Schaarschmidt, & Hothorn, 2015). This procedure corrects for multiple hypotheses testing via multiple contrast tests and not via adjustment of significance cut-offs (like Bonferroni correction), such that conventional levels of significance (α = 0.05) can be maintained without increasing the risk of Type I errors. The simultaneous twosided confidence intervals and p-values were calculated with Tukeytype contrasts and multivariate t-distributions.
For each species, we assessed tendencies of vertical distribution in the water column by analyzing gillnet position at capture (lower, middle, or upper third, see Section 2.3) with ordinal logistic models (R package ordinal, Christensen, 2015), treating net positions as ordered categories.
Species-specific time trends in umbilical scar condition over the study period (138 days from first to last deployment) were analyzed using univariate ordinal logistic models, where open, semihealed, and healed condition were treated as ordered categories that represent degree of healing. Linear models were used to analyze speciesspecific time trends in length.
Finally, we assessed the association of shark abundance with environmental parameters using zero-inflated Poisson (ZIP) models (R package pscl, Jackman, 2011), due to many excess zeros in the catch data (59 of 103 deployments yielded no shark catch). As environmental parameters were not measured for all deployments, models were based on the subset of n = 67 observations. Turbidity was excluded as a predictor, as in 13 cases of measurement the Secchi disk reached to the seafloor (we decided not to exclude these cases from analyses, but rather turbidity as a predictor, in order to maintain the already confined sample). The remaining variables-temperature, salinity, depth, distance to mangroves, and tide-were not strongly correlated with each other (all Pearson r < 0.3; Supporting Information Appendix S3) and thus were suitable for simultaneous inclusion in the full model. All models also included the log-transformed effort in minutes as an offset variable. For ecological inference, we selected the best-performing models based on AIC. We chose this information theoretic approach to assess the relevant importance of different models and predictors because of the rather exploratory nature of the study.
The models with the highest predictive accuracy were selected separately for each species based on the lowest AIC values from all possible combinations of predictors. One of the sampled species, Sphyrna mokarran, had insufficient abundance in the reduced dataset and was excluded from ZIP analysis.

| Interviews
A total of nine interviews were conducted with fishermen who use inshore and offshore areas around the Ba River mouth, and who inhabit the coast of the estuary. Interview information on shark occurrence was collected following the methods of Rasalato et al. (2010) and Glaus et al. (2015). Fishermen were either previously identified and approached after acquiring the consent of the headman of the respective village, or directly designated by the headman himself. Interviewees' oral consent was obtained prior to each interview, and fishermen were informed about the project and the purpose of the survey. All interviews were conducted on a voluntary basis and anonymity and confidential treatment of all obtained data was explicitly assured. A local Fijian translator who was fluent in English and Fijian (Bauan dialect) was present at all times and assisted whenever necessary. During the semi-structured interviews (Supporting Information Appendix S4), a visual identification poster of common inshore and offshore elasmobranch species (http://fijisharkcount.com/the-activity/all-materials/idposters) was used to confirm species recognition. Information was collected concerning shark species occurrence, history of shark abundance over the last 15 years, and where sharks are frequently caught by operating fishermen, as a proxy for preferred habitat types. Types of fishing gear used, as well as targeting and utilization of sharks, were also assessed.

| Catch composition
A total of 103 gear deployments were conducted. Gillnets (n = 73) and longlines (n = 30) were deployed in the seven selected sampling areas in the Ba Estuary, totalling 196.13 hr of fishing effort (Table 1) and resulting in 103 shark captures (Carcharhinus limbatus: n = 57, Sphyrna lewini: n = 35 and Sphyrna mokarran n = 11; see Figure 2). No sharks were recaptured during the study period.
Visual species identification could be confirmed using DNA barcoding for all 100 individuals for which a fin clip was stored. Thirtyfour sequences were positively identified as C. limbatus (100% bootstrap support), 30 sequences as S. lewini (100% bootstrap support), and six sequences as S. mokarran (100% bootstrap support; see Supporting Information Appendix S2 for parsimonious tree).

Highest monthly CPUE for gillnets was recorded in December
(  (Table 2). Specifically, CPUE was higher in sampling areas 4 and 7 than in sampling area 6 (p < 0.05), and tended to be higher in sampling area 4 than in sampling area 2 (p < 0.1). Although sampling area 5 had a mean CPUE even slightly above that of sampling area 4, catch variability was almost twice as high in sampling area 5 compared to sampling area 4. TA B L E 1 Overview of number of sharks caught per sampling area, the corresponding longline and gillnet effort and the resulting overall Catch per Unit Effort (CPUE) for each gear type and area (sharks gear −1 hr −1 )

Sharks caught in gillnet
Gillnet hours (shots)

| Vertical net positions
For 62 of 103 captured sharks, capture position along the vertical length of the gillnet was documented and subsequently used to explore potential partitioning of species in the water column.
As also indicated by Figure 3, juvenile C. limbatus were more frequently caught in the higher positions of the net (that is, closer to the surface), as compared to juvenile S. lewini (ordinal logistic model, p = 0.01; see Supporting Information Appendix S7). Most individuals of C. limbatus were caught within the top third of the net (62 %).
No difference in vertical occurrence was observed between S. lewini and S. mokarran (p = 0.29).

| Biological shark data
Of the 103 sharks, 52 were males, 49 females, and two could not be sexed due to damage to the sharks inflicted by predatory bites while in the gillnet (Table 4). When all three species are combined, 46% of sharks captured were found to have an open umbilical scar (n = 47), 22% were classified as semihealed (n = 23), 30% as healed (n = 31), with the remaining two individuals being unidentifiable due to the aforementioned damage. For statistics on umbilical scar condition and length by species, see Table 4. Length distributions ( Figure 4) differed significantly between all pairs of species (Kolmogorov-Smirnov tests, all p < 0.001).

| Environmental drivers of shark abundance
For both, S. lewini and C. limbatus, temperature and salinity were the most important predictors of abundance, as they appeared in all best-performing models (Table 6). For C. limbatus, there are four models that have almost identical predictive accuracy and some of them also include the predictors depth and distance to mangroves.
They indicate that slightly less C. limbatus were caught at deeper sampling sites and at those located further from mangrove forests (Table 6) (Table 6). Note the high uncertainty for most regions of our predictions as it is apparent in the plots (Figure 6).

| Interviews
This section reports some of the main findings from the interviews, while others will also occur in the discussion to provide context. When asked about the amount of sharks caught per week and per F I G U R E 5 Umbilical scar condition plotted over months including mean total stretch length (in cm) for (a) Carcharhinus limbatus and (b) Sphyrna lewini. Error bars depict standard deviation boat, numbers varied between four to 20 sharks for the fishermen using the study area. One of the hook-and-line fishermen who fishes further offshore reported to capture up to 100 sharks as bycatch per trip (4-5 days) and boat (29 feet, 40 PS).

| D ISCUSS I ON
This study is the first fisheries-independent survey on shark oc-

| The Ba Estuary as a multispecies parturition ground
As exclusively juvenile sharks were encountered, it is likely that the studied nearshore environment constitutes another parturition  . Size ranges of C. limbatus and S. lewini (66 ± 4 and 52 ± 5 cm, respectively) were in accordance with size ranges of neonate and young-of-the-year sharks from previously published studies (Castro, 1996;Castillo-Géniz, Márquez-Farias, Rodriguez de la Cruz, Cortés, & Cid del Prado, 1998;Brown et al., 2016 for S. lewini). For example, size of newborn C. limbatus range between 55 and 65 cm total length (Castillo-Géniz et al., 1998;Castro, 1996) (Heupel, Simpfendorfer, & Hueter, 2003). Note. All models within the range of two ΔAIC from the best-performing model are shown for each species, along with their Akaike weight w (weight is calculated from the set of all possible models, not only from the subset of best-fit models presented in the table). Models also contain the log-transformed effort in minutes as an offset variable.
ecological data, further studies are needed to investigate the importance of the Ba Estuary for S. mokarran.
Additionally, the main author was able to document three juvenile bull sharks (Charcharhinus leucas) during the study period which had been caught in gillnets by local fishermen in the Ba river several kilometres upstream the estuary. While only two could be measured (76.1 cm, 78.2 cm), fishermen confirmed fairly regular catches of small sharks of different species within the river during informal discussions and interview sessions. Bull sharks are classified as "Near Threatened" (Simpfendorfer & Burgess, 2009) and despite having been documented in other river systems of Fiji (Cardeñosa et al., 2017), no scientific record had been made in the study area before.

| Discussion of nursery ground criteria
In addition to its likely role as a parturition ground, does the Ba Estuary also serve as a nursery ground? Nursery grounds are essential habitats for sharks, usually located in shallow inshore waters, which provide juveniles with high intake of energy and little risk of predation. The capture of individuals with healed scar conditions, especially in the later stages of our sampling time frame, suggests utilization of the Ba Estuary by juvenile sharks even after the parturition period. However, this does not yet satisfy the definition of a nursery area. According to Heupel et al. (2007), a nursery area is defined by a higher mean density of neonate or young juvenile shark abundance than in surrounding areas (criterion 1), the utilization of the area over extended periods of time (criterion 2), and the repeated use over years (criterion 3).
Our study was too locally confined to prove that shark abundance was higher in the Ba Estuary than in surrounding areas (criterion 1).
The continued presence of sharks over the study period, however, fulfills criterion 2 for at least the sampled time frame. Regarding criterion 3, interviews conducted with local fishermen, additional informal talks with a range of village inhabitants, fishermen and elders, as well as the study conducted by Rasalato et al. (2010), strongly suggest that the Ba Estuary is utilized by juvenile sharks over multiple years. Distinct nursery areas of S. lewini and C. limbatus have been described (Bush & Holland, 2002;Heupel & Simpfendorfer, 2002) and both species exhibit some degree of philopatric behavior (Chapman, Feldheim, Papastamatiou, & Hueter, 2015). Anecdotal accounts by local fishermen of relatively high abundances of neonate and juvenile sharks repeatedly over years support the argument that this area is a nursery ground.
Follow-up studies should investigate the Ba Estuary over a longer timeframe and across all seasons (wet and dry) to further substantiate these findings and systematically test all three nursery ground criteria of Heupel et al. (2007) with long-term data. This will enable informed decision-making about management measures, such as temporal closures or protected areas, for the maintenance of ecologically valuable shark habitats (Knip et al., 2012).
Even without final proof of whether the Ba Estuary constitutes a nursery area, Yates, Heupel, Tobin, and Simpfendorfer (2012) argue that many diverse locations might serve as important habitats to young sharks, and thus to the maintenance of populations, despite not fully meeting the three criteria of a shark nursery as defined by Heupel et al. (2007).

| Fine-scale distribution of species in the Ba Estuary
We found differences in total shark abundance between sampling sites, which suggests variability in the use of parts of the estuary.
Sampling areas 4 and 5 yielded the highest total catches and CPUEs, while areas 2 and 6 yielded the lowest. This distribution is consistent with the reports of fishermen during the interviews. Whereas in some sampling areas (3, 4, 5) C. limbatus strongly dominated the catch over S. lewini, in others (1) the ratio was reversed.
Such differences in shark composition between areas are unlikely to be artefacts of differential gear use, because we tried to spread longline and gillnet sampling effort equally across sampling areas, which was approximately accomplished (Table 1). Furthermore, differential catchability between gears was only found for S. lewini, as they were exclusively caught with gillnets (Table 3). Even so, the proportion of longline effort in a sampling area does not correlate with the CPUE of S. lewini in an area (Pearson r = −0.08). Thus, we are confident that differences in S. lewini abundance across areas are not due to the (minor) differences in effort per gear across areas. This could be indicative of species segregation in the estuary, at least to some degree. In line with that, C. limbatus was almost never caught simultaneously with S. lewini (four cases from 46 in which at least one of both species was caught). There was also a difference in the depths at which C. limbatus (closer to the surface) and S. lewini were captured by gillnets. These instances of spatial segregation can be the result of either differential habitat selection based on physical factors (Yates et al., 2015) or direct interspecific processes like competition for space and food resources (White, Platell, & Potter, 2004). Competitive interactions (and thus selection) are theorized to occur with higher intensity within nursery areas (Heithaus, 2007). Also, the three juvenile bull sharks caught upstream by fishermen during the study period might avoid competition and predation risk by occupying freshwater that is inaccessible to other shark species (Heupel & Simpfendorfer, 2011), although this remains speculative. Osmoregulation is energy-consuming for sharks, with the largest energy expenditure presumably required when surface to volume ratio is lowest, that is among juveniles  Up to a point, warmer temperatures can induce faster growth and boost metabolic rates by increasing rates of biochemical reactions (Froeschke et al., 2010;Heupel et al., 2007). Thus, the overall high sea surface temperatures of the Ba Estuary (29-32°C) may benefit the juvenile sharks by maximizing physiological performance, as long as they do not surpass a critical threshold. Accordingly, catch rates for all shark species increased with temperatures from 20 to 33°C in a study conducted along the Texan coast, before declining again above 33°C (Froeschke et al., 2010). Given this information and based on our own models (Figure 6), the Ba Estuary might represent a habitat at, or in some parts even slightly above, the upper limit of tolerable temperatures for these sharks. Rising ocean temperatures in coastal waters, as is projected with climate change, might thus lead to altered spatial distributions or higher mortality rates (Bangley et al., 2018;Chin et al., 2010).
Turbidity can also affect habitat choice in juveniles (Yates et al., 2015), but measurements in this study were not sufficient to be included in our analyses. Turbidity is considered to facilitate predator avoidance for young sharks . Other factors like prey availability can also influence habitat use (Torres, Heithaus, & Delius, 2006). However, the majority of the bycaught teleosts in our study exceeded the size of potential prey, such that we lack a proxy for prey density. Strikingly, and opposed to other artisanal fisheries where sharks have high economic and consumption value (e.g., the Gulf of Mexico, Castillo-Géniz et al., 1998), the main anthropogenic threat to sharks in the study area results from bycatch. Shark bycatch is a problem of global magnitude (Bonanomi et al., 2017). However, compared to the situation in communities with shark-targeting fisheries, where conservationists and resource users experience conflicting interests, this has the positive implication that an agenda to reduce shark catch is not against the economic interests of fishermen in the Ba Estuary. Indeed, our interviews showed that fishermen desire to avoid catching sharks due to their low economic value, and that they would largely support spatiotemporal closures in conjunction with financial compensation schemes.

| Management implications
Temporal closures are not a new concept to Fijians, as their traditional tabu system refers to the part-time prohibition of fishing within selections of the qoliqoli following events of social significance (e.g., death of a chief) to allow recovery of certain fish species and maintain overall ecosystem health (Caillaud et al., 2004).
Contemporary governance approaches in Fiji often incorporate area management based on customary systems (Jupiter et al., 2017).
Thus, there is strong potential for community support. Such support and understanding by the local population is crucial for the successful implementation of fisheries closures (Bennett & Dearden, 2014).
Both our ecological and our fishermen-based survey indicate suitable time frames (November to February) and areas (sampling areas 4 and 5) for such closures due to high concentrations of juveniles and potential parturition.
Importantly, closures are not a panacea to integrate biodiversity conservation and development (Adams et al., 2004), and other measures to reduce shark bycatch exist, such as gear modifications (Bonanomi et al., 2017). Ultimately, the success of any strategy will depend on whether a co-management regime is successful in maintaining fisheries or alternative livelihoods and, at the same time, in being adjusted to the life-histories of the local shark species. This can be especially challenging for species like S. lewini, which has a particularly low potential for population recovery (Branstetter, 1987;Smith et al., 1998). Both will require further research, or even experimentation with policy schemes and continued monitoring (i.e., adaptive management; Folke, Hahn, Olsson, & Norberg, 2005). If, for example, fishing restrictions result in protection of sharks but also transparent co-benefits for fishermen through the replenishment of fished teleost stocks (Aburto-Oropeza et al., 2011), such an intervention has good chances of being enforced and institutionalized even by the communities themselves (Ostrom, 2000).
The fact that interviewed fishermen reported additional species to occur in the estuary that we did not sample, such as whitetip reef or nurse sharks, further emphasizes the need for appropriate local conservation policies and potentially the incorporation of fishermen's catch or landing data into assessments of local shark occurrence. Interestingly, the majority of interviewees (67%) reported to mainly catch hammerhead sharks as bycatch, while our own sampling predominantly yielded blacktips (62%). This difference might be partly due to the characteristic appearance of hammerhead sharks making them more prone to be remembered, and highlights the need for complementarity of indirect (Rasalato et al., 2010) and direct shark surveys such as this study.
While there is currently no practical solution to eradicate shark bycatch, there are several possibilities to minimize it if policy and decision-making processes incorporate scientific data into their agenda. Countries such as the Republic of Fiji, where sharks naturally occur within the national territory, have a responsibility to ensure the long-term survival of these endangered species by adopting national management plans that support global biodiversity, such as the protection of critical habitats.

ACK N OWLED G M ENTS
Sampling and interviews were conducted under a research permit issued by Fiji Immigration Department to T.V. and approved by the Justin Rizzari and another anonymous referee for their constructive reviews.

CO N FLI C T O F I NTE R E S T
Authors have no conflict of interest to declare.