Movement behaviour and fishery interaction of silky sharks (Carcharhinus falciformis) in the tropical tuna purse seine fishery in the Western Indian Ocean

Abstract

The silky shark Carcharhinus falciformis regularly associates with floating objects in the open ocean, resulting in relatively high levels of bycatch in industrial tuna purse seine fisheries using drifting fish aggregating devices (FADs). This bycatch has contributed to concerns regarding the sustainability of this fishery and its impact on silky shark populations. To investigate fishery interactions, movements of 28 silky sharks (86–235 cm TL, mean = 118 cm) fitted with pop-up and archival tags in the western Indian Ocean, between 2010 and 2012, were examined. Monthly overlap between probability surfaces of sharks and two fishery metrics (FAD-tuna catches and FAD positions) were calculated. Vertical habitat use overlapped almost entirely with operational gear depth. Horizontal movements were extensive (3–5024 km) and covered large areas of the western Indian Ocean. Monthly overlap with FAD distributions was consistently high (64.03–100%) highlighting the need for compliance with FAD design regulations to avoid entanglement. Monthly overlap with tuna catches was more variable (8.43–51.83%). The observed movement patterns suggest static spatial management measures would be have limited conservation impact, however dynamic approaches could be appropriate. Limiting fishery activities directly will likely have the greatest conservation outcomes for silky sharks in the purse seine fishery.

Bycatch, Fish Aggregating Device, FAD, pelagic, satellite tag, tagging

Introduction

The silky shark (Carcharhinus falciformis) occurs throughout the tropical and subtropical regions of the world's oceans. They are captured throughout their distribution by industrial and artisanal fisheries (Bonfil, 2008) where they are both targeted and taken incidentally as bycatch. Historic and recent studies have found that the silky shark is the second most important species (after the blue shark, Prionace glauca) supporting the shark fin trade in Hong Kong (Clarke et al., 2006; Fields et al., 2018). A range of stock indicators have shown population declines in this species across all oceans (Aires-da-Silva et al., 2014; Rice et al., 2015; Clarke et al., 2018; Ortiz de Urbina et al., 2018). In light of the widespread exploitation in industrial and artisanal fisheries and incidental capture of this species the International Union for Conservation of Nature (IUCN) has designated it a global status of vulnerable (Rigby et al., 2017). Additionally, in 2016 the species was included in Appendix II of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CoP17, Notification No. 2016/063), which forces strict control over the international trade of any part of the species.

The silky shark is known to associate with floating objects found in the oceanic environment, especially during the juvenile phase (Bonfil, 2008; Filmalter et al., 2015; Tolotti et al., 2020). This behavioural trait has led to the species forming an important component of the bycatch in industrial tropical tuna purse seine fisheries that utilize floating objects to facilitate the capture of tunas. Vessels deploy floating rafts known as fish aggregating devices (FADs) which drift freely around the oceans and over time aggregate large schools of tunas and other non-target fishes. FADs are then revisited after weeks or months and the aggregated biomass captured using a purse seine net. The use of this method has increased during the past three decades with the majority of the worlds purse seine caught tropical tunas now caught around FADs (Dagorn et al., 2013a). In the Indian Ocean the silky shark comprising 95.5% of the elasmobranch bycatch in the tropical tuna purse seine fishery and occurs in 65% of fishing sets on FADs (Clavareau et al., 2020). These sharks are generally early juveniles with an average size of 1 m (González et al., 2007; Amandè et al., 2008) and age range of 0–5 years (Oshitani et al., 2003; Rabehagasoa et al., 2014). At a global scale, it is estimated that almost 200000 silky sharks are taken as bycatch each year in all FAD-based purse seine fisheries (Filmalter et al., 2013a). In the Indian Ocean alone, this figure is approximately 80000 (Filmalter et al., 2013a).

Conservation efforts for this species have now begun to take shape but still require further attention. In recent years all of the tuna regional fishery management organisations (tRFMOs) have developed conservation and management measures related to silky sharks. These include retention bans for silky sharks caught by purse seine vessels in the Atlantic and Pacific Oceans (ICCAT Rec 11-08; WCPFC CMM 2019-04; IATTC Res C-19-05). However, the conservation efficacy of these measures is questionable where capture is incidental (Tolotti et al., 2015a) and post-release survival rates are low (Poisson et al., 2014a; Hutchinson et al., 2015; Eddy et al., 2016). Furthermore, all tRFMOs have adopted regulations prohibiting the use of entangling materials in the construction of FADs (IOTC Res. 19/02; ICCAT Rec. 19-02; IATTC Res. C-19-01; WCPFC CMM 2018-01). These measures aim to mitigate the incidental mortality of silky sharks (Filmalter et al., 2013a) and other marine animals such as sea turtles and have likely had a major positive impact on their conservation, but require appropriate control and enforcement to ensure their efficacy. Additionally, some spatial and effort management measures have been applied in different oceans, largely aimed at reducing fishing effort on tuna stocks (WPCFC CMM 2018-01; IATTC Res C-17-02; ICCAT Rec 16-01). While not directly aimed at conserving silky sharks, these measures could indirectly reduce fishery related mortality. However, for the efficacy of proposed spatial management measures to be determined, information on the movement behaviour of the target and non-target species is critical.

Little is known about the large scale movement behaviour of silky sharks, especially in the Indian Ocean, with information on the location and spatial extent of nursery or feeding areas particularly lacking. Both persistent use of seamounts and wide-spread movements have recently been observed from the central Indian Ocean (Curnick et al., 2020). In the eastern Pacific Ocean (EPO), juvenile silky sharks have been found to be more common north of the equator (Watson et al., 2009) however this pattern can be influenced by environmental regimes (Lennert-Cody et al., 2019). In the western Atlantic, juvenile silky sharks are believed to occupy shelf water with new-borns having a more demersal lifestyle (Bonfil, 1997). Fishery data modelled with environmental variables in the eastern Atlantic Ocean detected both persistent and seasonal zones of increased silky shark abundance linked to productive oceanographic systems (Lopez et al., 2020a).

An important aspect determining the vulnerability of pelagic sharks to purse seine activities is the intensity of their interactions with the fishery, i.e. interactions with FADs and interactions with vessels when fishing. Here, we investigate the spatiotemporal overlap between silky sharks and the FAD based purse seine fishery, both in terms of the distribution of FADs and active purse seine fishing. Furthermore, by examining their vertical movement behaviour, we investigate their theoretical overlap with purse seine gear. In the Indian Ocean, the tropical tuna purse seine fishery is known to exert effort in different areas, following a predictable seasonal pattern (Davies et al., 2014; Kaplan et al., 2014) while the movements of silky sharks remain unknown. In an attempt to evaluate the vulnerability of silky sharks to the purse seine fishery, we tested the following hypotheses: (H₀) silky sharks have a spatially defined area use in the Indian Ocean and are vulnerable to the fishery seasonally or alternatively (H₁) area use is not spatially explicit and their movement patterns dictate their vulnerability to fishery exposure. Following these hypotheses, the expected outcome of H₀ would be that sharks could be afforded some indirect protection when the fishery is operating in another part of the ocean. Alternatively, under H₁ if their movements consistently overlap with the fishery, then their probability of capture would be non-zero throughout the year. To do this we used fishery-independent movement data from silky sharks equipped with Pop-up Satellite Archival Tags (PSATs). Coupling this movement data with information on FAD positions and fishing effort data, this study aimed to obtain novel insights into the vulnerability of silky sharks to the industrial tuna purse seine fishery in the Indian Ocean.

Methods

Capture and tagging

Sharks were caught or using baited handlines from a small vessel in the immediate vicinity of a FAD. Tagging operations were conducted between 2010–2012, with tags being deployed in two areas; in oceanic regions surrounding the Seychelles islands or in the northern half of the Mozambique channel. Sharks were brought onboard, placed in a padded cradle and a hose pumping clean saltwater was placed in their mouth to irrigate the gills. PSATs (model: miniPAT, Wildlife Computers, Redmond, USA) were attached to the sharks (n = 8) by inserting a plastic anchor, to which the tag was tethered, into the dorsal musculature of the shark. Initially a small puncture was made in the shark's skin such that the tag anchor could be inserted with ease. For some sharks (n = 6), the tags were attached using a threaded nylon rod, which was passed through a hole drilled into the anterior base of the dorsal fin. All PSATs were programmed to release from the animal after either 100 or 120 d at liberty. PSATs recorded hydrostatic pressure (to estimate depth), external temperature and light level (used for geolocation).

In addition to PSATs, captured sharks were also tagged with internal acoustic tags, either V13 or V13P (V13-1L-64K,  Vemco, a division of Amerix Systems), transmitting at frequency of 69 kHz. Acoustic tags had a nominal delay of 90 s (range: 50–130 s) and an expected battery life of 879 d). As an alternative to PSATs, four sharks caught from small vessels were tagged with internal archival tags (MK9, Wildlife Computers, Redmond, USA). All sharks were released at the FAD where they were captured. The entire tagging operation usually lasted approximately 3–5 min. Satellite linked acoustic receivers (VR4-GLOBAL, Vemco, Canada) were attached to FADs were sharks were double tagged. These receivers recorded and transmitted acoustic detection information (time of detection and tag identity) from sharks with acoustic tags when they were within the reception range of the receiver. Effective reception range of the receivers was approximately 400 m (Filmalter et al., 2015). Acoustic receivers also transmitted their location multiple times per day.

In order to increase the size of the data set analysed in this study, additional silky sharks tagged with PSATs as part of a post-release survival study in the tropical tuna purse seine fishery in this region, described in Poisson et al. (2014b) were also included. For this study, only sharks that were deemed to have survived the capture process were included. These sharks were caught under regular tuna purse seine operations using FADs and tagged on deck of the purse seine vessel on a foam mattress. Detailed descriptions of capture, tagging and fate determination are provided in Poisson et al. (2014b).

Horizontal movement behaviour

Data received from tags were pre-treated and any tags that released from the animal less than six days after tagging were not considered in further analyses.

Most probable tracks were generated using the proprietary online software tool GPE3, provided by Wildlife Computers. This software uses a state-space modelling approach to generate time-discrete and gridded probability surfaces throughout the deployment period based on light level data collected and transmitted by the tag, or in the case of archival tags, downloaded from the tag upon recovery. These 12-h likelihood surfaces are output at 12-h intervals and correspond to 50, 9% and 95% location probabilities. Geolocation estimates are refined by matching recorded surface temperatures and depths with sea surface temperature (NOAA OI SST V2 High Resolution) and bathymetrical (ETOPO1-Bedrock) databases within GPE3. In additions to probability surfaces, most probable locations are also output by the model on a 12-h basis. Known locations, (for double-tagged sharks which were detected by acoustic receivers on drifting FADs) were also included in the geolocation model, further refining the tracks. A speed filter was applied within the track estimation process that limited the potential distance that the model could predict future positions from previous positions. The selection of appropriate speeds was determined through an iterative process on an individual basis, where multiple tracks were generated for an individual, over a range of speeds at 1 km h⁻¹ increments, until successive tracks converged. The lowest the speed that provided a convergent tack was then selected for the final model output of each individual. Final speed values ranged from 1–6 km h⁻¹

Vertical habitat use

Swimming depth were recorded by the tags at five-minute intervals. Depth data records were assigned to bins of 10 m intervals for the first 100 m and then a 50 m bin from 100–149 m followed by 150 m bin from 150–299 m, a 300 m bin between 300 m and 600 m and lastly any records >600 m were placed together. This was done on an individual basis and then the average values and standard deviations for each bin were calculated across all individuals. Using this information the proportion of time that tagged sharks spent at depths less than the average operational depth of tuna purse seine nets used in the Indian Ocean was assessed. Tuna purse seine nets typically have a constructed hanging depth of 200–300 m (Ben-Yami, 1994), but the operational depth is considerably shallower. In the Indian Ocean, purse seine sets are typically made shallower than 150 m with the majority occurring between 60  and 70 m (IOTC, 2015). For the assessment of theoretical shark overlap with this fishing gear two depths were considered. The proportions of time spent shallower than 150 m and shallower than 20 m were both calculated. Vertical data were combined from all individuals. Data were divided between day and night periods using the time of sunrise and sunset at the location of the animal according to geolocation model estimates each day during the deployment period. For the analysis of potential diel variation in depth distribution, only day or night period where > 50% of the depth records in the period were received, were included. Deep diving behaviour, where sharks descended below 150 m, was also investigated. The effect of day period (dawn, day, dusk and night) as well as sex and size, were assessed using uni- and multivariate Poisson regression models.

Dispersal analysis

The dispersion of juvenile silky sharks from their tagging locations was assessed on a regional basis, to compare movements between the vast oceanic environment of the Seychelles area with the more confined area of the Mozambique Channel. Considering the inaccuracy of light-based geolocation estimates (approximately 110 km), site fidelity was assessed by evaluating the amount of time the animal spent within 200 km of its tagging location.

Fishery interaction

Currently, detailed information on numbers of fishing sets on FADs are not publicly available for all fleets. To obtain the broadest representation of fishing effort on FADs we used catch data reported by member states to the Indian Ocean Tuna Commission (IOTC) during the period when the tagged sharks were at liberty. These data are reported by year, month, species and set type in a 1 × 1 degree grid. For our analyses we used the total catch of the three target tuna species (skipjack, yellowfin, and bigeye) caught at floating objects across all fleets by 1º grid during each month. These data were used to create monthly kernel densities aggregated across all study years (2010–2012) of total target catch at floating objects. Kernel densities were estimated using the kde2d function in the MASS package in R (R Core Team, 2020) using a bandwidth of 2° and 500 grid points.

As a second indicator of fishery interaction, we examined the spatial distribution of tracked floating objects (FADs) on a monthly basis during the months when sharks were at liberty. Locations of FADs and tracking buoys deployed by French-flagged vessels were obtained from the “Observatoire des Ecosystèmes Pélagiques Tropicaux exploités” (Ob7) from IRD/MARBEC. These data provided daily positions recorded from GPS tracking buoys attached to FADs by owner vessels. Monthly kernel densities (Figure S1 in Supplementary Material) were generated using these data following the same method described above and aggregated across years.

To investigate the spatial interaction between tracked sharks and these two fishery metrics the 90% probability surfaces of monthly kernels from both fishery data sets were extracted. The 90% probability surface was chosen as this provided sufficient representation of the fishery's spatial extent, but did not overinflate the overlap with tagged sharks. For shark space use, the 50% probability surfaces obtained from GPE3 were used. This level of probability was chosen as we believed it provided sufficient information on both the space utilized by tagged sharks and also encompassed the uncertainty of the of the geolocation model without overinflate potential fishery interactions. These surfaces were merged for all sharks during each calendar month and the percentage overlap with the 90% fishery surfaces obtained. As such the interaction between tagged sharks and the fishery could be compared across all months during which tagged sharks were at liberty.

Results

A total of 28 silky sharks fitted with PSATs were included in this study (14 tagged solely for ecological investigation and 14 which survived in the post-release survival study). An additional four sharks had archival tags internally implanted (Table 1). All sharks survived the capture/tagging process however four individuals succumbed to entanglement in FADs during their monitoring periods (see Filmalter et al., 2013b). Tagged sharks ranged in size from 86–235 cm TL (mean = 118 ± 33 cm SD). Satellite transmitted data were received from all PSATs with 27 deployed for > 5 d and thus included in the analyses. Three PSATs and one archival tag were recovered from recaptured sharks and the data downloaded directly, resulting in a total of 28 tags used for data analysis. The vast majority of PSAT tags (92%) released prior to their programmed date with only two tags reaching full term. Tag deployments ranged from 6 to 141 d (mean = 51.3d ± 34.7d SD) with a cumulative observation period 1437 days. Due to the timing of tag deployments, transmitted data covered the months of April–August from 2010–2012. In total, six tagged sharks were recaptured, four by purse seine vessels, and two by artisanal fishers (tags not returned), resulting in an overall recapture rate of 16%.

Table 1.

Tagging, morphometric, and movements details of silky sharks tagged in the Western Indian Ocean.

Tagging date	Total length (cm)	Sex	Tagging latitude	Tagging longitude	Tag ID	Deployment duration (d)	Track length (km)
13-Mar-2010	88	F	−13.211	41.731	34419a	22	462
15-Mar-2010	109	M	−13.843	44.586	34420a	100	2595
28-Mar-2011	86	F	−14.200	45.317	94255	44	367
01-Apr-2011	116	F	−15.067	45.783	104659	36	435
01-Apr-2011	87	M	−15.350	44.767	104658b	21	273
01-Apr-2011	90	M	−15.350	45.417	94259c	17	405
02-Apr-2011	155	M	−15.350	44.767	94254	13	379
15-Apr-2011	91	M	−13.119	44.967	34206a, b	27	579
20-Apr-2011	103	F	−12.175	44.948	98719a	12	124
20-Apr-2011	99	M	−12.175	44.948	34366a, c	98	1983
25-May-2011	119	F	−14.250	42.350	104663b	41	2308
25-May-2011	122	F	−14.550	42.617	104662	53	3306
27-May-2011	235	M	−14.267	42.500	98 717	45	400
18-Jun-2011	98	F	−5.506	54.395	94261a	6	3
20-Jun-2011	102	M	−5.794	53.953	94251a, c	75	882
20-Jun-2011	93	M	−5.794	53.953	34415a	33	440
02-Apr-2012	104	M	−6.576	54.067	104665	29	223
02-Apr-2012	114	F	−6.576	54.067	104664b	41	422
03-Apr-2012	132	M	−6.905	54.518	104667	78	613
03-Apr-2012	155	M	−6.905	54.518	98724	100	2242
03-Apr-2012	130	F	−6.905	54.518	98723	43	564
13-Apr-2012	109	F	−8.973	49.868	94260a	119	5024
13-Apr-2012	112	F	−8.973	49.868	104678a	31	572
13-Apr-2012	102	F	−8.973	49.868	990030d	–	NA
13-Apr-2012	93	M	−8.973	49.868	990036d	–	NA
13-Apr-2012	103	F	−8.973	49.868	990028d	–	NA
13-Apr-2012	99	F	−8.973	49.868	990026b, d	141	3642
14-Apr-2012	111	–	−8.973	49.868	104674a, b	30	538
14-Apr-2012	116	–	−8.973	49.868	94253a	59	1225
14-Apr-2012	116	M	−8.973	49.868	94255a, c	4	NA
03-May-2012	225	M	−4.747	61.483	98721	45	1469
06-May-2012	104	–	−7.293	60.334	104656	78	1191

^aDouble tagged (PSAT + acoustic).

^bRecapture.

^cEntangled in FAD.

d

Double tagged (archival + acoustic).

Vertical habitat use

Data on vertical movements were obtained for 1089 days, with data gaps resulting from messages not being received by the ARGOS satellite system before the tags’ batteries expired. As a result of these gaps, only data from 1039 observation days were included. Silky sharks depth records ranged between 0 and 600 m, however, they typically displayed very shallow depth distributions and spent an average of 100.0% (± 0.2% SD) of their time in depths less than 150 m (Figure 1). Furthermore, tagged sharks spent 75.7% (± 26.8% SD) of their time at depths less than 20 m. These shallow depths were occupied regularly both during the day and night (day = 79.4 ± 30.6% SD, night = 72.1 ± 22.1% SD). Occasional deep dives were observed to a maximum depth of 600 m, however, these forays were rapid, with a mean duration of 20.8 min (± 6.9 min SD). In total 130 dives beyond 150 m were observed during the 1039 days that sharks were monitored, equating to 0.17% of the observation time. Although most deep dives were observed during the night (64.5%), the highest frequency (per hour) occurred during dusk (an average of 0.014 deep dives/per hour). The number of deep dives observed during each period of the day (dawn, day, dusk, and night) differed significantly (χ² = 98.364, df = 3, p < 0.001). However this behaviour was linked to only a few individuals, and no significant relationships between number of dives and period of the day, as well as sex or size, were found at an individual scale (tested using uni- and multivariate Poisson regression models). It was found that eight of the 28 individuals accounted for 75% of all dives observed. While these results indicate that deep dives are not particularly common among silky sharks, they show a preference for conducting deep dives during dusk. It also appears that larger individuals (> 120 cm) tend to conduct more deep dives, although no significant size effect could be identified in the model, probably due to the small number of large sharks included in this study.

Figure 1.

Average time at depth from 28 silky sharks fitted with pop-up (n = 27) and internal archival (n = 1) tags. Error bars indicate standard deviations, periods refer to nights or days where depth records were obtained.

Horizontal movement behaviour

The tracks generated using the GPE3 model provided by Wildlife Computers ranged between 3–5024 km (mean = 1167 km, ± 1238 km SD) (Figure 2). Daily movement rates averaged 20.4 km d⁻¹ (± 14.1 km d⁻¹ SD). For individuals tagged in the Seychelles region, the majority of the large-scale movements were initially westward and then northwards once the sharks reached waters off the African continental coast. Only two individuals (Tag IDs 98 724 and 98 721) moved in an easterly direction with movements centred around the 5°S latitude. The track of one of these sharks (Tag ID 98 721) ended close to the southern end of the Maldives while the other (Tag ID 98 724) moved towards the Chagos archipelago before turning back east. It is noteworthy that these two individuals included the largest and the third largest tagged in this study. Three sharks tagged in the Seychelles region were recaptured by purse seine vessels fishing on FADs. Two were caught within the Seychelles area while one was captured offshore of Somalia.

Figure 2.

Most probably tracks of 28 silky sharks tagged with pop-up and internal archival tags. Track widths indicate uncertainty in daily location (50% probability as generated by the GPE model from Wildlife Computers), while colours represent time as per colour scale.

Sharks tagged in the Mozambique Channel either remained in the northern portion of the Mozambique Channel throughout their monitoring period or moved out toward the Somalia area travelling offshore but roughly following the continental coast of East Africa. Two sharks from this area were captured by local fishers. The first was in southern Somalia and the second in northern Mozambique. This was evidenced by the direct (straight-line) trajectories and increased track speeds obtained from the tag transmissions. Furthermore, the final ARGOS locations indicated that the tags were stationary in small coastal villages. In addition, one shark (ID 104 658) was recaptured after 21 days at liberty by a purse seine vessel fishing in the northern Mozambique channel.

Dispersal

A large degree of variation was observed in the dispersal patterns of tagged sharks. However, there is evidence that tagged sharks generally remained within approximately 500 km of their tagging location for between 20 and 40 days in both the Mozambique Channel (Figure 3a) and Seychelles region (Figure 3b). Dispersal from the Mozambique channel was slower than from Seychelles region. Two individuals from the Mozambique Channel showed very rapid long-range movements away from the tagging location after approximately 30 days. The other 11 tagged sharks in this region displayed longer residency times with the majority still within 500 km when their tags released. In the Seychelles area, a more consistent dispersal pattern was apparent, with sharks moving further from their tagging location with time (Figure 3b).

Figure 3.

Dispersal of 28 tagged silky sharks from original tagging locations in (a) the Mozambique Channel and (b) the Seychelles.

Fishery interaction

A total of 189766.7 tons of the three target tuna species were reported to the IOTC as caught in floating object sets during the months that the tagged sharks were at liberty from 2010 to 2012. The average monthly kernel density distribution of these catches is shown in Figure 4a. During the same period, 138966 daily FAD positions were recorded in the region from 2444 unique FADs. The average monthly kernel density distribution of FAD positions are shown in Figure 4b. The probability surfaces from tuna catches on FAD revealed a high degree of overlap with the tracks of tagged sharks for several of the observed months. Overlap with FAD locations was extremely high during April (Figure 5a), with all shark tracks falling within the 90% probability surface of FAD locations (Table 2). FAD catches were concentrated in the Mozambique Channel, an area occupied by many of the tagged sharks, resulting in a high degree of overlap. In May catches on floating objects were spread out between the Northern Mozambique Channel and oceanic areas surrounding the Seychelles (Figure 5b). In May, FADs were more homogenous spread throughout the study area, with their distribution covering areas off the Somali coast and central-western Indian Ocean and northern Mozambique Channel. Overlap between shark tracks and floating object catches remained high in the Mozambique channel during May but was lower in the central region of the study area, when several tagged sharks were located.

Figure 4.

Average monthly kernel distribution of (a) purse seine catches from FADs declared by all fleets and (b) positions of drifting FADs from French flagged tuna purse seine vessels. All data spans the study months when tagged silky sharks were at liberty (April–August in 2010–2012).

Figure 5.

Monthly spatial interaction of tagged silky sharks with two fishery metrics. Probability surfaces (50%) of tagged silky sharks are shown in green. The monthly 90% probability surfaces generated from kernel densities of drifting FAD positions are shown as blue dashed lines. The monthly 90% probability surfaces of catches of target tuna species on floating objects from all purse seine fleets in the Indian Ocean, generated from kernel densities, are indicated by red dashed lines. All data cover the period 2010–2012.

Table 2.

Overlap between tracked sharks and 90% probability surface of two fishery metrics for calendar months when sharks were at liberty from 2010–2012.

Month	% Overlap with FAD positions	% Overlap with FAD catches	Tracks included (N)	Shark observation days
April	100	51.85	20	812
May	85.31	39.5	21	763
June	64.03	8.43	15	676
July	82.65	12.44	10	415
August	97.79	12.79	3	147

By June, the distribution of FADs covered a large part of the central western Indian Ocean and spread further north off the Somali coast. The overlap of this distribution with shark tracks was still high (Table 2) during this month. FAD catches were concentrated in the north and north-western parts of the Seychelles and off southern Somalia (Figure 5c). The overlap between tracked sharks and FAD catches was low during June as sharks spent much time outside of the focal fishing grounds.

In July, FADs were largely concentrated west of 55°E and north of 10°S (Figure 5d), with overlap with shark tracks remaining high (Table 2). While the FAD distribution reached further to the west and north-west of the Seychelles, FAD catches remained spatially similar to June. Shark tracks showed the lowest spatial overlap with the catch distribution during July, with several individuals moving away from the top of the Mozambique Channel and along the Somali coast.

In August, the distribution of FADs continued to spread further northwards off Somalia and the overlap with shark tracks remained high (Table 2). FAD catches became more concentrated north-west of Seychelles and off Somalia with limited overlap with tagged sharks (Fig 5e).

Overall overlap between tracked sharks and the 90% probability surface of FAD positions was extremely high, ranging from 64.03–100% between calendar months (Table 2). A significant difference in overlap was observed between months (Chi-squared test, χ² = 9.6488, df = 4, p < 0.05). Spatial overlap with the 90% probability surfaces of monthly reported tuna catches form floating objects was lower and also varied significantly between months (Chi-squared test, χ² = 60.498, df = 4, p < 0.01), ranging between 8.83–51.85%.

Discussion

The deployment of satellite tags on pelagic sharks is generally limited by their remote habitats and high expenses associated with accessing them. As such sample sizes are typically small. This study represents the largest electronic tagging efforts on silky sharks in the Indian Ocean and also reports the longest movements for the species globally.

Fishery vulnerability is a direct function of both the vertical and horizontal space use patterns of a species and the fishery with which it interacts. While the horizontal component relies on spatial patterns of fishing effort, the vertical component relates to the physical characteristics of the fishing gear. Understanding vertical swimming patterns can allow for modification in fishing practices that could reduce the incidental capture of non-target species. For example, setting tuna-targeted longlines deeper has been suggested to reduce the number of oceanic whitetip sharks captured in the western Atlantic Ocean (Tolotti et al., 2015b). The tropical tuna purse seine fishery focuses on surface orientated tuna schools and uses nets of 150–350 m in depth (Lopez et al., 2020b) but the effective depth when fishing is generally much shallower. In the Indian Ocean, effective fishing depth is generally less than 150 m (IOTC, 2015). Results on vertical behaviour of silky sharks from this study showed that tagged sharks occupied depths less than 30 m for much their time, irrespective of location or time of day. Furthermore, almost all of their time was spent above 150 m. As such it is clear that reductions in incidental capture of this species cannot be achieved through vertical modifications of fishing operations as the target tunas typically occupy depths similar to silky sharks when associated with FADs (Forget et al., 2015).

Understanding horizontal movement patterns of pelagic animals is important for developing spatial management and conservation measures. Information on the horizontal movement patterns of silky sharks is limited globally and largely lacking in the Indian Ocean. Some results from tagging studies conducted in other oceans have shown that adults silky sharks can move large distances (> 1200 km) (Kohler et al., 1998; Musyl et al., 2011), however, patterns of residency and philopatry in association with islands, sea mounts and continental shelves have also been reported (Kato and Carvallo, 1967; Stevens, 1984; Clarke et al., 2011). Recently, a small number of subadult silky sharks tagged in the central Indian Ocean showed divergent movement patterns with some individuals showing restricted horizontal space use with fidelity to the same sea mounts spanning several months, while one individual moved over 4700 km from its tagging location (Curnick et al., 2020). The movements observed in the current study were similar in their maximum extent but little evidence of spatial fidelity was observed. It is possible that the behaviour of association with a physical structure in the pelagic environment leads to these divergent patterns. In one case, the structure is fixed in place (a seamount) while in the other it is mobile (a floating object).

Knowledge on the large-scale movement behaviour of FAD-associated silky sharks is limited. FAD-associated juvenile silky sharks in the central Pacific Ocean have recently been found to undertaking movements of up to 3800 km in 129 days (Hutchinson et al., 2019). Similar results were obtained in this study in the Indian Ocean. Several tagged sharks covered distances in excess of 3000 km with individuals moving at speeds of up to 62 km d⁻¹. The movements recorded here clearly show that silky sharks in the size range commonly caught at FADs, utilize the entire tropical region of the western Indian Ocean, from latitudes of 15S° to 10°N. While there appears to be some short term fidelity to the regions where individuals were tagged, both sharks from the Mozambique Channel and those from the Seychelles moved away from these areas when observation periods were sufficiently long (>30 d). These dispersal patterns are contrary to findings from the EPO and eastern Atlantic Ocean where areas of higher abundance of juveniles have been identified (Roman-Verdesoto and Orozco-Zoller, 2005; Lopez et al., 2020a). It is likely that the benefits derived from associating with floating objects allow these small sharks to utilise massive oceanic areas, characterized by patchy prey fields, where other associated species serve as a reserve food source (Filmalter et al., 2017). In this way, juvenile silky sharks might maximize their feeding success by targeting rich prey patches when they are encountered and falling back on FAD-associated species when the surrounding prey environment is poor.

Knowledge on the fishery interaction through time and space is critical for understanding the broad-scale impacts of both active fishing activity and passive impacts such as potential entanglement in FADs. An ecological risk assessment for pelagic sharks in the Indian Ocean showed that there was a 100% overlap between the purse seine fishery and the distribution of silky sharks (Murua et al., 2018). This assessment, which relied heavily on observer data, categorised the silky shark as moderately vulnerable to the purse seine fishery in this region, largely due to assumed increased post-release survival following the fleet-wide adoption of best handling practices for sharks. Post-release survival studies such as that conducted by Poisson et al., (2014b) are essential to validating such assumptions.

The results obtained here suggest that direct fishery interactions are not constant through time and space. It is important to note that interactions reported here give insight into general patterns. High levels of overlap in space and time between tracked sharks and the fishery metrics do not necessarily translate to high levels of capture. This is largely due to the resolution of the various data sets used in this analysis. Despite this limitation, the outcomes of this study provide novel insights into the potential interactions of this fishery with its primary elasmobranch bycatch species. Overlap between shark space use and purse seine fishing activity was highest during the Mozambique Channel fishing season (April–May) and decreased as effort shifted towards the Seychelles area in June–August. It is noteworthy that during the period when sharks were at liberty purse seine fishing effort did not follow its typical seasonal shifts (Kaplan et al., 2014). Due to the threat of piracy off the coast of Somalia fishing effort was constrained to the area north of the Seychelles, which likely impacted on the interaction probabilities of sharks in that area. As such, if this pattern of fishery effort were maintained, through active spatial management interventions, silky shark bycatch could be effectively reduced.

It is noteworthy that the patterns of space use and periods of highest overlap with the fishery differ substantially from a recent examination of predicted bycatch hotspots of silky sharks in the purse seine fishery in the Indian Ocean (Mannocci et al., 2020). Using onboard observer data, this study found a persistent hotspot for silky shark bycatch in the Arabian Sea, north of 5°N. Interestingly none of the sharks tagged in the current study utilised this area, possibly due to truncated observation periods or a the lack of tag deployments in this area.

The recapture rate of silky sharks tagged in this study was very high for a pelagic species (16%), confirming both the vulnerability of this species to capture and the high level of fishing pressure applied to the stock in the western Indian Ocean. The largest dart tagging effort for this species, undertaken in the Atlantic Ocean (Kohler et al., 1998), reported a recapture rate of 6.6%. While the majority of recaptures in the current study were made by purse seine vessels fishing on FADs it is noteworthy that a third appeared to have been caught in coastal artisanal fisheries. This assumption was based on the location of satellite transmissions in conjunction with vertical movements of the tags before they surfaced prior to their programmed release date. It is unlikely that dart tagging methods would have provided such insights into fishing pressure as reporting of tags from remote artisanal fisheries is highly improbable. As such it would appear that the use of PSATs could provide a reliable method for assessing fishing pressure for this highly mobile species.

Indirect fishery interactions between juvenile silky sharks and floating objects were found to constantly occur at high levels. Although the degree of spatiotemporal overlap between drifting FADs and tagged silky sharks differed slightly among months, there was no period when tracked sharks were located in areas void of drifting FADs. The deployment of large numbers of FADs has significantly modified the floating object environment in the Indian Ocean from its natural state prior to industrial fishing. There are now areas with high densities of man-made FADs where densities of natural floating objects, such as logs, are low (Dagorn et al., 2013b). While the results of this study show that the extensive movements undertaken by silky sharks always fall within areas where FADs occur, it is not possible to know if these movements are a result of this habitat modification, or if they occurred prior to the massive use of FADs. Accordingly, the ecological impact of such modifications remains unclear. Recent work from the Indian Ocean has suggested that juveniles silky sharks move independently of FADs at least 30% of the time, however quantifying the exact impact of drifting FADs on silky shark movements is complicated by the strong correlation between FAD drifts and ocean surface currents (Bonnin et al., 2021). While it has been suggested that the increased deployment of FADs could potentially have negative impacts on the ecology of tunas (Marsac et al., 2000; Hallier and Gaertner, 2008), the provision of additional refuge and food reserves may benefit juvenile silky sharks. Furthermore, juvenile silky sharks may benefit from associated tuna schools as they more efficiently locate patchy prey in the areas surrounding FADs. However, their increased vulnerability to fishery mortality could mask such benefits.

An investigation into the unobserved entanglement of juveniles silky sharks in the netting used to construct FADs suggested that between 480000 and 960000 individuals could be killed in the Indian Ocean alone (Filmalter et al., 2013a). Since then, tRFMOs have introduced management measures preventing the deployment of entangling FAD. The high overlap of silky sharks and FADs highlights the necessity of these regulations. Furthermore, the need to constantly monitor the construction of deployed FADs is very apparent. It is noteworthy that during the period when the sharks in this study were tracked, the use of none entangling FADs had not begun. Considering their constant overlap with FAD distributions, the extent of this risk at that time is highlighted. These results also illustrate the need for further tagging studies of this kind to assess the efficacy of new FAD design regulations.

Understanding animal movement behaviour is critical to the planning and development of spatial management tools such as marine protected areas (MPA). The concept of Pelagic MPA that cover portions of the high seas has been debated for some time (Game et al., 2009; Kaplan et al., 2010). Considering the results of the current study, for a pelagic MPA to effectively protect juvenile silky sharks in the Indian Ocean, it would have to be unrealistically large. The design and implementation of such MPAs requires compromises and trade-offs between fishery production and protection of target or vulnerable species. The key knowledge provided here can aid in evaluating such trade-offs. Recently, the concept of dynamic spatial management of ocean areas using the environmental variable to predict bycatch hotspots has gained momentum (Hazen et al., 2018; Mannocci et al., 2020; Lopez et al., 2020a). Using the spatiotemporal information obtained in this study in combination with environmental variables can facilitate the development of predictive models for identifying hotspots in near-real time. Such models can then be used to develop more effect conservation and management measures.

Finally, from the results presented here, it appears that effort restrictions on FADs would likely be the most effective measure to reduce silky shark catches in this fishery. Such restrictions could take a variety of forms, including limiting the number of FADs sets per vessel, or imposing stringent limits on the number of active FADs/tracking buoys deployed at any time. Naturally, any such measure would require appropriate monitoring, control and enforcement initiatives.

Conclusion

Silky sharks tagged in the Western Indian Ocean undertook extensive movements and utilised large areas of the open ocean environment. Their area use patterns overlapped with purse seine fishing grounds for much of the year with a peak in interactions occurring in the northern Mozambique Channel during April and May. Despite their wide-ranging movements, silky sharks are constantly moving through environments rich in drifting FADs, suggesting that both passive and active fishing activities have impacted on this species. Owing to these broad movement patterns, it is unlikely that implementation of static spatial conservation and/or management initiatives would be effective for protecting or conserving silky sharks in the western Indian Ocean. Rather, dynamic spatial management initiatives could be more effective, but will require considerably more information on habitat preferences of silky sharks in this region. Finally, regulations that manage fishing effort on FADs are likely to be most appropriate for mitigating the impacts of the purse seine fishery on this species.

Funding

This work was supported by the Commission of the European Communities, Framework Programme 7, Theme 2–Food, Agriculture, Fisheries and Biotechnology, through the research project MADE, contract no. 210 496. Some research reported in this article was funded by the International Seafood Sustainability Foundation (ISSF) and conducted independently by the authors. The paper and its results, professional opinions and conclusions are solely the work of the authors. There are no contractual obligations between ISSF and the authors that might influence the article's results, professional opinions and conclusions.

Data availability statement

The data underlying this article were provided, in part (FAD positions), by Ob7 (IRD) under data exchange agreement with fishing companies. These data will be shared on a request to corresponding author and Ob7 (IRD) with permission of the owners. All other data will be be shared on reasonable request to the corresponding author.

Authors’ contributions

JDF conducted field work, assisted with data analysis and wrote manuscript. PDC Contributed to the interpretation of data and to writing manuscript. FF assisted with the data collection and interpretation. RKB contributed majorly to data analysis. LD conceptualized and coordinated the study, acquired funding and assisted with manuscript writing. All authors read and approved the final manuscript.

Acknowledgements

The authors would like to acknowledge the “Observatoire des Ecosystèmes Pélagiques Tropicaux exploités” (Ob7) from IRD/MARBEC for the provision of data related to fishing activities and FAD tracking buoys. The authors would like to sincerely thank the skippers and crew from the six vessels from which tagging was conducted. We also extend our thanks to the three anonymous reviewers who's comments and suggestions greatly improved this manuscript.