Estimating Natural Mortality of Atlantic Bluefin Tuna Using Acoustic Telemetry

Atlantic bluefin tuna (Thunnus thynnus) are highly migratory fish with a contemporary range spanning the North Atlantic Ocean. Bluefin tuna populations have undergone severe decline and the status of the fish within each population remains uncertain. Improved biological knowledge, particularly of natural mortality and rates of mixing of the western (GOM) and eastern (Mediterranean) populations, is key to resolving the current status of the Atlantic bluefin tuna. We evaluated the potential for acoustic tags to yield empirical estimates of mortality and migration rates for long-lived, highly migratory species such as Atlantic bluefin tuna. Bluefin tuna tagged in the Gulf of St. Lawrence (GSL) foraging ground (2009–2016) exhibited high detection rates post release, with 91% crossing receiver lines one year post tagging, 61% detected after year two at large, with detections up to ~1700 days post deployment. Acoustic detections per individual fish ranged from 3 to 4759 receptions. A spatially-structured Bayesian mark recapture model was applied to the acoustic detection data for Atlantic bluefin tuna electronically tagged in the GSL to estimate the rate of instantaneous annual natural mortality. We report a median estimate of 0.10 yr−1 for this experiment. Our results demonstrate that acoustic tags can provide vital fisheries independent estimates for life history parameters critical for improving stock assessment models.

natural mortality (M), which has traditionally been difficult to estimate 22 . Natural mortality has been estimated for Atlantic bluefin tuna using electronic tagging data 23 , although values of age-specific M for western Atlantic bluefin tuna are still considered uncertain by ICCAT. In 2017, the western assessment used an age-varying rate derived from the Lorenzen method 23 scaled to M = 0.10 at ages 14-16+, while the eastern assessment used a Lorenzen curve scaled to M = 0.10 at ages 20+ 24 . A sensitivity analysis found that a lower rate of terminal M for the western stock is associated with lower estimates of recruitment and SSB.
The degree of mixing between eastern and western stocks is a further key uncertainty in the Atlantic bluefin tuna assessment. ICCAT currently uses separate assessment models for eastern and western Atlantic bluefin tuna (i.e. population mixing is not accounted for). Estimates of population-specific depletion can be biased when catch removals are not attributed to the correct stock of origin [18][19][20] . An analysis using simulated data has indicated that accouting for mixing is particularly important for the western stock to obtain unbiased estimates of absolute stock size 25 . Information about the movement and relative abundance of different populations in time and space is key to correctly quantifying the contributions of those populations to fishery catches in mixing models. Electronic tagging of Atlantic bluefin tuna has emerged as a powerful tool for learning about many aspects of the biology and ecology of bluefin tunas [26][27][28][29][30][31][32][33][34][35] . Data obtained from electronic tags has the potential to improve estimates of key model parameters, such as rates of fishing and natural mortality, and migration and mixing. Conventional and electronic tagging have revealed the details of large-scale migrations of juvenile, adolescent and mature bluefin tuna, the understanding of which is central to the proper management of this species [3][4][5][6][7][17][18][19][20][26][27][28][29][30][31][32][33] . Information from electronic tagging and biological markers that provide origin of the tagged fish, allows estimation of population-specific movement patterns. Tagging, otoliths and genetics indicate that the amount of trans-Atlantic crossing varies depending upon population of origin, year examined, age of catch, and possibly sex. Tagging studies have indicated higher rates of trans-oceanic movements of eastern origin fish to the western Atlantic than vice versa, likely due to the much larger size of the eastern population 5 and the more limited distribution of the GOM population. Together with genetic markers, results from electronic tagging can also be informative about catch composition, demonstrating for example that western Atlantic bluefin tuna fisheries target mixed populations along the eastern seaboard of North America 5,18 .
Using this new knowledge from electronic tagging data to build biologically plausible models is vital to minimizing bias in assessments of stock status. Despite the rapid advances in our understanding of Atlantic bluefin tuna biology, key questions remain about population mixing, productivity, recruitment dynamics, maturity schedules, abundance trends and the number and stock origin of fish harvested by western Atlantic fisheries. To date, almost all electronic tagging of Atlantic bluefin tuna has been focused on deployments of archival and pop-up satellite archival tags. Acoustic tags have the potential to provide valuable information about the biology of Atlantic bluefin tuna, given their high detection rate, their longevity, and the independence of detections from fishing activity.
In this paper, we examine the use of acoustic tags in combination with Ocean Tracking Network (OTN)deployed acoustic receiver lines across entrances of the Gulf of St. Lawrence to: a) examine timing of arrival and departures of Atlantic bluefin tuna foraging in the GSL, b) determine the fidelity to foraging grounds annually to estimate how many fish return to the GSL, c) estimate survivorship using a multistate Bayesian mark-recapture model, and d) test if fish acoustically tagged and released in North Carolina waters recruit into the GSL. The GSL may serve as a unique location for long term monitoring of the Atlantic bluefin tuna fishery, due to the extraordinary investments Canada has made in the OTN infrastructure in this region. Strategic underwater receiver lines are now in place in many location of Canadian coastal waters and opportunistic investments have placed additional receivers along the US coastline from Maine to Florida. Conventional tags placed simultaneously on fish tagged with the acoustic tags provide an additional set of long-term marks necessary to generate estimates of natural and fishing mortality similar to previous studies conducted on Atlantic and Pacific bluefin tuna (Thunnus Orientalis) using archival and pop-up satellite tags 23,35 .  (Table 1). Fish were caught on commercial Atlantic bluefin tuna fishing vessels, permitted to conduct scientific tagging, in the fall months off Port Hood on Cape Breton Island, Nova Scotia. The fish were all caught on rod and reel with live or freshly caught dead Atlantic mackerel (Scomber scombrus) or Atlantic herring (Clupea harengus) bait. During the tagging campaigns, one vessel was designated a tagging boat and multiple fishing vessels caught bluefin tuna on rod and reel. The fish were "transferred" to the designated tagging boat, the F/V Bay Queen IV, which had a large deck and transom. All bluefin tuna were brought on board the vessels using methodologies described previously 3,5,17 . In addition to these Canada deployments, four fish were caught on trolling lures in North Carolina waters in March 2013 using sport fishing vessel, tagged and released 4 .

Methods
Once a bluefin tuna was caught by rod and reel, the fish was reeled in and leadered to the open transom door. By placing a titanium or stainless steel lip hook, carefully behind the lower jawbone, we are able to pull the fish through the transom door and onto a wet vinyl mat. A saltwater hose was inserted in the mouth to oxygenate the tuna gills while on deck and a soft cloth soaked in a fish protectant solution (PolyAqua ® ) was placed over the eyes to keep the fish calm 3,33 . Curved fork length (CFL) of the fish was measured to the nearest mm with a flexible tape measure, fish were also sampled for fin clips for genetics, tagged and released. When possible pictures of the electronic tag positions were obtained upon release (Fig. 1 For this experiment all Vemco acoustic tags were packaged in a plastic "shark case" with 5 mm holes drilled in at both ends of the tag (Fig. 1). The holes were enlarged for the attachment leaders by hand boring with a file or dremel tool, and carefully smoothed to prevent any interaction of the edges of the hole with the materials used to construct the leaders. The tags were secured to the fish externally using a two-point attachment technique, with a custom titanium dart on each end of the acoustic tag. Tags were inserted into the dorsal musculature of the fish at depths of 15.2 to 17.8 cm depending upon the size of the bluefin tuna. The materials in the leader consisted of a single layer of 180 kg monofilament (Moi Moi Hard), a cover layer of aramid braided cord that provided increased abrasion resistance over the monofilament, and up to two layers of heat shrink wrap. Pop-up satellite archival tags (Wildlife Computers MK-10 and mini-PATs) were attached to a subset of the acoustically tagged tuna and tracks from these satellite tags were reported on previously 17,33 . Information from these pop-up satellite archival tags are not used in the present model and analysis.
Acoustic receiver lines using VR4 UMs receivers were deployed and maintained by OTN. They were initially placed across a portion of the Cabot Strait and across the Scotian shelf off Halifax, Canada in the summer of 2007 (Fig. 2). The receiver array used to enclose the GSL was partially installed when the project was initiated. This line was completed in late 2008 and spanned the entire Cabot Strait and the Strait of Belle Isle, which together provides an electronic "gate" that the Atlantic bluefin tuna must cross prior to reaching the GSL foraging ground. The completion of the OTN lines enabled us to record long-term movements of bluefin tuna acoustically tagged on their GSL foraging grounds. In addition, the previously deployed Halifax Line, completed in 2007, provided a line of complete coverage across the Scotian Shelf (Fig. 2). Additional deployments opportunistically of receivers along the eastern seaboard of North America from Newfoundland to the Gulf of Mexico, Bahamas and in the Strait of Gibraltar provided opportunistic detections (Fig. 3a).   Table 1). 91% of the acoustically tagged fish were subsequently detected by a receiver post deployment (Table 1). From these acoustic tag deployments, 31,822 acoustic detections were acquired by receivers located along the eastern seaboard of North America from Newfoundland to the Florida keys, the Bahamas, and in the Strait of Gibraltar (Figs 1-3). We used 101 acoustic tags for development of a bluefin tuna mortality model and the mean mean curved fork length for tagged fish in the model was 250 cm (5 th percentile 193 cm, 95 th percentile 294 cm).
The original deployment years (2009-2013) were designed to test whether the Vemco acoustic tags (V16-4h, 6 L) were detectable from bluefin of the size class tagged, and we scheduled these tags to transmit coded acoustic pulses for a period of ~2.5 years with a predicted maximum of 858 days. The tags were designed with a kill switch for 865 days per manufacturer specifications, however some variation occurs due to battery life and temperature. Up to one year post release, 91% of the acoustic tags were detected at the OTN lines in Canadian maritime Shelf waters (Fig. 4). By year two, 61% of the fish carrying acoustic tags were detected across the OTN lines. As many as 34% of the tags were still detected in their third year post release indicating that the battery life extended beyond the manufacturer specifications (Fig. 4). Two bluefin tuna tags were detected for four years post release from this first release of Vemco tagged bluefin tuna and a single fish had five years of detections also indicative that tag attachments worked.
Based on results from pop up satellite archival tagging 33 , and the recapture history of tagged bluefin tuna in the Mediterraenan Sea (Table 2) (Fig. 3b). Based on acoustic detections bluefin tuna entered the GSL by crossing the Cabot Strait Line during the summer months from 4 June to 22 October, (mean date 10 July) (Fig. 5). Bluefin exited the GSL when crossing the Cabot Strait Line from 2 July to 19 November (mean date 12 October) after spending 7 to 166 days (mean GSL residency 94 days) on the GSL foraging grounds. Bluefin tuna usually crossed the Halifax Line on the Scotian Shelf (located ~400 km southwest of the Cabot Strait Line) before crossing the Cabot Straight Line in the early summer and after in the fall. Bluefin tuna crossed the Strait of Belle Isle Line, located to the north of Newfoundland, from 7 July to 23 September (mean date 6 August), including one fish that crossed the Strait of Belle Isle Line in four consecutive years. It appears these fish were exiting the GSL via this route as most had been detected earlier entering the GSL via the Cabot Strait. Entry and exit dates, and residency days were calculated from detections post deployment year.
Transit durations between the crossing of the Halifax Line on the Scotian shelf on the northern journey, and the Cabot Strait Lines ranged from 2.85 to 77 days (mean duration 14.90 days). The shortest distance that a fish could swim between the two lines is approximately 460 km, suggesting a minimum sustained speed of approximately 6.73 km/hour, in the case of the bluefin tuna with the shortest duration between subsequent recordings. www.nature.com/scientificreports www.nature.com/scientificreports/ Transit durations between the Cabot Strait and Halifax Lines were longer, ranging from 3.57 to 127 days (mean duration 37.53 days). Inshore receivers on both the Cabot and Halifax lines received significantly more hits than offshore receivers (Fig. 6) indicative that the fish are moving along the coastal shelf waters in relatively shallow depths.
Additional detections of tagged Atlantic bluefin in Canadian waters were obtained from Vemco Mobile Transceivers (VMTs) attached to free swimming grey seals located in the southern GSL and on the Scotian Shelf An interesting finding of the current study was the large number of detections that we observed on the Canso Causeway receiver (detections). The Strait of Canso, linking the GSL to the Atlantic Ocean was the historic migration route of these fish, and has been blocked by the Canso Causeway since late 1952. The longevity of giant bluefin would suggest that current year classes of GSL fish are only a few generations removed from the last bluefin tuna that might have used this passage as the primary migration route for entering and exiting the southern GSL prior to 1952. There are anecdotal reports of large numbers of bluefin seen in close proximity to the causeway in the years immediately following its construction. The receiver in the vicinity of the Canso causeway obtained over 4000 detections. To exit the GSL, bluefin must swim around Cape Breton Island to reach the Atlantic side of the Strait of Canso, a detour >450 km or longer or go north thru Bell Isle.
Tagged Atlantic bluefin tuna were also detected by individual moored Vemco receivers (    www.nature.com/scientificreports www.nature.com/scientificreports/ of the four North Carolina acoustic tags were subsequently detected by acoustic receivers (Table 1) on the Halifax line and off Sable Island (Fig. 7). One of the fish was detected off Cape Cod in June 2013.
Bayesian mark-recapture model. Using a spatially-structured state-space model, we obtained a posterior median estimate of the instantaneous annual natural mortality rate in Atlantic bluefin tuna of 0.10 yr −1 (standard  www.nature.com/scientificreports www.nature.com/scientificreports/ deviation of log x, SD 0.34). (Fig. 8). The acoustic tagging data were also informative about rates of seasonal movement into and out of the Gulf of St. Lawrence, upating the prior distribution in most months (Fig. 9). The estimated rate of movement into the GSLwas highest during June and September (Fig. 9b), while the high estimated rates of departure from the GSL in October and November (Fig. 9a) are consistent with observations among receivers at the Cabot, Canso and Belle Isle Straits in those months.
Estimated detection probabilities at acoustic receivers were much higher in the GSL box than outside (Fig. 10), reflecting a higher density of receivers in this area, and the fact that tagged Atlantic bluefin tuna must cross receiver lines to enter and exit the GSL. Acoustic detection probabilities were estimated to have increased during the first years of the study in both areas, probably reflecting recruitment of receivers in the OTN and other projects over the study's duration. Acoustic detection probabilities in the GSL were estimated to have decreased in the final 2 years of the study (Fig. 10a), possibly reflecting attrition and re-deployments of receivers to new areas, or lags in the acquisition of receiver data annually. See supplementary material for additional model results.

Discussion
Electronic tagging of long-lived highly migratory fishes with coded acoustic tags permits conducting long-term studies that can provide valuable information about rates of mortality and migration. For Atlantic bluefin tuna, this technology can potentially provide monitoring capacity and address significant questions such as: a) the timing of arrival and departures of Atlantic bluefin tuna foraging in Canadian waters, b) the natural mortality rate of mature fish based on Bayesian modelling approaches 17,23,35 . Acoustic tagging data can inform current population models on the status and assessment of the Atlantic bluefin tuna populations. The original battery life of the acoustic tags used in this study was programmed to be ~2.5 years. More recent tags have projected battery lives of 5-10 years. The tags showed significant reliability when placed externally, with double titanium dart attachments, indicating the technology is capable of showing fidelity to a specific geographic area. We anticipate that with the long periods of occupation evident in the GSL waters (Fig. 7b) it may be possible to routinely obtain 5 year acoustic records for Atlantic bluefin tuna. This can be utilized for long-term monitoring of the assemblage of fish in these waters and could be used to assess recruitment of juvenile fish utilizing Carolina waters into the GSL.
Atlantic bluefin tuna have a complex population structure and there remain significant questions concerning the status, the structure and dynamics of Atlantic bluefin tuna populations, especially in the North Atlantic where mixing is known to occur on foraging grounds. The availability of a network of receivers covering the Cabot Strait provided the initial opportunity to test the role of acoustic tags in improving fisheries management of these valuable fish. Development of methods to provide empirical estimates of natural mortality is of high priority for bluefin tuna stocks, since all else being equal, using a lower rate of natural mortality in the stock assessment can often lead to  www.nature.com/scientificreports www.nature.com/scientificreports/ lower estimates of the ratio of current to unfished stock size (i.e. greater depletion), and more conservative projections of future stock development. Survival estimates from the multistate mark-recapture model for Atlantic bluefin tuna suggest a low rate of mortality from natural causes, consistent with the fact that most individuals in this study had a curved fork length ≥240 cm at tagging, corresponding to an age of ~14 years or more 39,40 . For comparison, ICCAT uses a natural mortality rate of 0.10 yr −1 for eastern Atlantic bluefin aged 20 years and older, and for western Atlantic bluefin tuna aged 14 and over 23 . Values used in the stock assessment are thus consistent with the natural mortality estimates obtained in this study using acoustic tag recapture histories. Acoustic tagging methods appear to have good potential to improve estimates of natural mortality in the stock assessment, where conventional tagging data have so far proven insufficient to distinguish between alternative hypotheses about natural mortality 23 .
The multistate mark-recapture model we applied provides a robust and flexible framework for estimating rates of survival and seasonal movement in long-lived migratory fish species. Disentangling non-detection, fishing vs. natural mortality and tags reaching the end of their programmed transmission life presents a challenge with acoustic tag data sets, particularly for long-lived species where relatively long recapture histories are needed to accurately estimate survivorship. Using Bayesian approaches can help to alleviate this problem by allowing incorporation of prior knowledge from other studies or sources. For example, in this study, prior information from earlier published studies was utilised for rates of natural and tagging-related mortality, while an empirical prior was developed for acoustic detection rates in the Gulf of St. Lawrence (see Supplementary Material for details). As noted above, the tags deployed from 2009-2013 had a programmed transmission life of approximately 2.5 years, which is likely not long enough to discriminate over a range of low values of natural mortality with a high degree of precision. Despite the use of prior knowledge, there is likely some conflation of natural mortality, tagging related mortality, tag loss, and non-functioning tags in model parameter estimates. Adding a further tag type to the model for which information about the reporting rate is available (e.g. tags with a large monetary reward such as the pop up satellite archival tag or surgically implanted archival tags) could help to inform estimates of acoustic tag loss and tag transmission time. The precision of the natural mortality rate estimate is also expected to improve once detection histories from tags with longer programmed transmission times (5-10 years) start to accrue.
A potential limitation of the model applied in this study is the coarse spatial resolution. Permanent (i.e. over the duration of the study) emigration out of regions of high detection probability, for example return of Mediterranean origin fish to the eastern Atlantic may affect estimates of other model parameters. This phenomenon could potentially lead to estimates of natural mortality and tag shedding rates that are biased high, although its effect is not expected to be significant given the low frequency of observations of satellite tagged Atlantic bluefin tuna that ended in the eastern Atlantic or Mediterraean (2 out of 48 over the duration of the study). Future work will extend the model to a higher spatial resolution. This could be implemented by splitting the outside-GSL box into e.g. 3 or 4 areas, allowing more detailed patterns of movement to be estimated. Improving prior information or adding auxiliary data on detection probabilities and rates of fishing mortaliy is of high priority: extension of the multistate mark-recapture model to both acoustic and satellite tag detection histories is ongoing. This is expected to improve estimates of area-specific detection probabilities and acoustic tag transmission times for acoustic tags with short detection histories. Both have potential to improve the accuracy and precision of natural mortality estimates. Given additional data on the genetic origin of tagged Atlantic bluefin tuna from fin clips (i.e. Gulf of Mexico vs. Mediterranean spawners), accounting for stock-of-origin would be straightforward within the model framework presented, whereby movement and other parameters can be estimated separately for each origin. While the results above apply to a limited number of year classes (e.g. corresponding roughly to the terminal age group in ICCAT's western bluefin tuna assesment), there has been a trend towards smaller lengths at tagging in recent years, so that development to an age-structured model could also be of interest in future. By increasing acoustic tagging effort in North Carolina, it might also be potentially possible to determine when a fish recruits into the GSL foraging ground from this lower latitude foraging area.
Testing acoustic tagging on the GSL foraging grounds was critical as this sea is a semi-enclosed region and the OTN has strategically placed two fully closed receiver lines at Cabot Strait, and Belle Isle. This placement of receivers ensures capture of the tuna's electronic signals when they leave the region and return. An additional line on the Scotian Shelf (Halifax Line), across the continental shelf provides valuable information in concert with the Cabot Strait line on arrival and departure. Together these receiver lines permit continuation of a long-term study both on resident and new arrivals. The GSL may serve as the best long term site for monitoring western Atlantic bluefin, due to the investment Canada has made in placing strategic underwater receiver lines here and the diligent effort they have in maintaining these lines and downloading the data. Our study has demonstrated a high detection probability within the GSL, which supports estimation of detection probabilities in other areas with lower receiver densities.
Importantly, the use of external acoustic tags was made possible only by deploying on the deck, and carefully anchoring the tag in two places. From recapture results, we know that we have succeeded in constructing a 5 year attachment tether that keeps tags on the fish reliably. Given that the V16 tags have met the 2.5 year specifications of the manufacturer in tag transmission rates we predict 5 and 10 year data detections times will be possible with the current deployment techniques (2016-present) and receiver arrays, yielding improvements in the precision of survival estimates. New models incorporating valuable information from double tag experiment (satellite and acoustic tags) data sets, as well as genetic identification of the population origin of the fish from fin clips, should improve our capacity to model the survivorship of bluefin tuna by population, providing important information on their annual foraging patterns, and potentially enabling an assessment of the efficacy of increased protections on the spawning grounds in the Gulf of Mexico.

Data Availability
Telemetry data will be made available via our public website at tagging of pelagic predators (https://oceanview. pfeg.noaa.gov/topp/map) upon publication, or by request to the corresponding author. All model data is provided in the supplement.