Juvenile survival and movements of two threatened oceanic sharks in the North Atlantic Ocean inferred from tag‐recovery data

Abstract Understanding population dynamics, movements, and fishing mortality is critical to establish effective shark conservation measures across international boundaries in the ocean. There are few survival and dispersal estimates of juveniles of oceanic shark species in the North Atlantic despite it being one of the most fished regions in the world. Here we provide estimates of dispersal, survival, and proportion of fishing mortality in the North Atlantic for two threatened oceanic sharks: the blue shark (Prionace glauca) and the shortfin mako shark (Isurus oxyrinchus). Our results are based on multi‐event models applied to tag‐recovery data of 700 blue sharks and 132 shortfin makos tagged over a decade. A total of 60 blue sharks (8.57% of tagged) and 30 makos (22.73%) were recovered by the longline fishery between 2009 and 2017. Tag‐reporting rate (percentage of returned information when a tagged shark was caught) was estimated to be high (0.794 ± 0.232 SE). Mean annual survival, as predicted from the models, was higher for blue shark (0.835 ± 0.040 SE) than for shortfin mako (0.618 ± 0.189 SE). Models predicted that fishing caused more than a half of total mortality in the study area for both species (0.576 ± 0.209), and more than a third of tagged individuals dispersed from the study area permanently (0.359 ± 0.073). Our findings, focused mainly on juveniles from oceanic areas, contribute to a better understanding of shark population dynamics in the North Atlantic and highlight the need for further conservation measures for both blue shark and shortfin mako, such as implementing efficient bycatch mitigation measures and static/dynamic time–area closures in the open ocean.


| INTRODUC TI ON
Oceanic sharks are among the widest-ranging animals in the ocean, typically moving across whole ocean basins and throughout a major part of the water column (0-2000 m; Queiroz et al., 2019).
As for most elasmobranchs, the life-history strategies of oceanic sharks are characterized by slow growth and late sexual maturity, which results in low fecundity and population productivity . Surviving through the long juvenile phase is therefore crucial to ensure the sustainability of populations (Kinney & Simpfendorfer, 2009). This is especially important for populations of oceanic sharks hampered by human activities such as fisheries, which can reduce reproductive opportunities for adults under scenarios of high fishing mortality (Camhi & Pikitch, 2008;Dulvy et al., 2021;Pacoureau et al., 2021).
Due to a lack of demographic and life-history information, existing stock assessments of pelagic sharks are most commonly based on catches and/or catch-at-age data, which usually results in great uncertainty around the estimated parameters (Carvalho et al., 2018;Cortés & Brooks, 2018). Understanding the fate of sharks (e.g., their survival, mortality, and dispersal) is also required to accurately estimate population growth and total allowable catch for harvested oceanic sharks. In particular, determining the fate of the juvenile portion of the stocks of oceanic sharks with low fecundity is needed to understand which proportion of the population reaches the mature stock and can therefore contribute to the subsequent generation (Benson et al., 2018).
Traditionally, mark-recapture studies have been based on adults and coastal areas (Kohler & Turner, 2001). In this study, we used mark-recapture data to investigate the fate of two species of oceanic sharks in the North Atlantic with a main focus on juveniles. The blue shark (Prionace glauca) and the shortfin mako shark (Isurus oxyrinchus) are distributed throughout tropical and temperate waters from the surface to ~1800 m depth (Mucientes, 2023;Sims et al., 2018;Vedor et al., 2021). In the North Atlantic, both species are heavily fished (Campana et al., 2016;Queiroz et al., 2016Queiroz et al., , 2019Sims et al., 2018), with catches of 36,500 tonnes and 3800 tonnes for blue and mako shark per year respectively ICCAT, 2023), which has resulted in severe population declines in the last four decades Pacoureau et al., 2021;Sims et al., 2021). Among the oceanic sharks, blue sharks have one of the highest population growth rates, with an age of maturity of 4-6 years and a litter size of 35-44 embryos (Dulvy et al., 2008). This life-history strategy has likely contributed to a slower decline in the relative abundance of blue sharks in the North Atlantic over the past 50 years compared with other oceanic sharks, despite high fishing intensity . Currently, there is a limitation in place, based on total allowable catches (TAC), for North and South Atlantic and, according to ICCAT (2020), the stock is "not overfished" and "overfishing is not occurring." However, the species has been classified as "near threatened" globally by IUCN (Rigby, Barreto, Carlson, Fernando, Fordham, Francis, Herman, et al., 2019). In contrast, shortfin mako matures at a remarkably late age (7.5 years in males and 18-22 years in females (Natanson et al., 2006(Natanson et al., , 2020Rosa et al., 2017;Yokoi et al., 2017)) and have a litter size of 8-12 embryos (Dulvy et al., 2008), which results in slow population growth. As a result, populations of shortfin mako have shown marked declines in abundance since 1970 that are attributed to overfishing ; indeed, ICCAT considers that the North Atlantic population is "overfished" with "overfishing still occurring" (ICCAT, 2019).
Furthermore, shortfin mako is considered "Endangered" globally in the IUCN Red List assessment .
Mark-recapture studies represent a valuable and cost-effective means to obtain information about the life history and behavior of oceanic sharks (Kohler & Turner, 2001, 2019. Mark-recapture has been used to analyze the distribution of sizes and sex ratios in populations of coastal and oceanic sharks, such as Caribbean reef shark Carcharhinus perezi (Talwar et al., 2022) and great white shark Carcharodon carcharias (Kanive et al., 2021), to develop indices of relative abundance in zebra shark Stegostoma fasciatum (Dudgeon et al., 2008(Dudgeon et al., , 2013, to provide data on the population structure of whale shark Rhincodon typus (Rohner et al., 2022), and to inform international fisheries management organizations (Cortés & Brooks, 2018). Mark-recapture studies on blue shark and shortfin mako conducted in the Atlantic Ocean have been successful in collecting information on both short-and long-term movements and migrations (Queiroz et al., 2005), growth rate, reproductive behavior, and for identifying mating and nursery areas (Kohler & Turner, 2019).
Here, we expand the existing knowledge by specifically addressing movement behavior, to determine survival, dispersal, and mortality of juveniles of blue shark and shortfin mako in the Atlantic Ocean.
Our results contribute to a more complete understanding of population growth and thus sustainability in these threatened species.

| Study area and tagging
Tagging of blue shark and shortfin mako was performed between 2007 and 2017 under the framework of the Cooperative Shark Tagging Program (CSTP, https://repos itory.libra ry.noaa.gov/view/ noaa/22731). The CSTP is a collaborative effort between recreational anglers, the commercial fishing industry, and scientific researchers to understand the movements and the life history of Atlantic shark species. It is managed by the Northeast Fisheries Science Centre, of the National Oceanic and Atmospheric Administration (NOAA).
Blue shark and shortfin mako were captured as target species (together with other species such as swordfish, Xiphias gladius). Both species were tagged by commercial fishers on board the Spanish longline fleet in the central North Atlantic (mainly west and south of the Azores islands); and mainly, by sport fishers (rod and reel) in coastal areas of Iberia ( Figure 1). Fishers were trained in handling, tagging, and collecting data according to the procedures of the CSTP.
The information recorded during tagging included species, size (fork length, FL), sex, date, gear type, and location of tagging. Based on size at maturity of blue (215 cm total length, TL; Dulvy et al., 2008) and shortfin mako, (200/280 cm TL male/female; Dulvy et al., 2008), most of the tagged individuals were likely juveniles at the time of capture. Conventional numbered dart tags (Kohler & Turner, 2001) were implanted in the dorsal musculature near the base of the first dorsal before sharks were released. This type of tag is highly visible to fishers and observers to increase the likelihood of sighting the tag upon the capture of the shark; furthermore, it has a small capsule at the posterior end containing detailed return instructions. Longline vessels and scientific observers reported the recoveries (Figure 1).
Our study area thus corresponds to the area of the North Atlantic where both tagging and recoveries occurred.

| Tag-recovery, data analysis, and modeling approach
To estimate survival of the tagged sharks, the tag-recovery data of blue shark and shortfin mako were used to construct two encounter history datasets (one for each species) that contained, for each year of the study period, information on whether the individual remained tagged or had been captured and the tag returned in that year. Since the data were collected opportunistically without a well-defined annual sampling season, we adapted our recovery records to the classical encounter history format of discrete annual sampling occasions.
The months of February through October were chosen as our annual sampling season because most tagging events occurred during that period of the year (100% of shortfin makos and 85% of blue sharks were initially captured during that period; Tables 2 and 3). Tag recoveries within a sampling season were assigned to the season's year, whereas recoveries taking place out of the sampling season were assigned to the next year (for a similar procedure see Fernández-Chacón et al., 2015). Multi-event modeling approach (Pradel, 2005), a type of hidden Markov model, was used to link tag recoveries to a series of underlying individual states defined in the model structure (see below and Appendix S1). This modeling approach has been successfully applied to mark-recapture data of other marine species such as Atlantic cod (Fernández-Chacón et al., 2015Kleiven et al., 2016).
Our encounter data consisted of three types of observations or "events," codified as follows: "not encountered" (0), "captured for the first time" (1), and "recovered dead" (2). From this set of events, we estimated annual individual survival, fishing mortality proportions, dispersal probabilities, and tag-reporting rates. We did so by constructing a model pattern based on transition matrices that linked the F I G U R E 1 Capture and recovery locations (dots with external white line) of shortfin makos (red dots, bottom picture) and blue sharks (blue dots, upper picture). Yellow lines join the tagged and recovery locations. observed events to transitions between possible underlying states, in which individuals may be found at a given occasion ( Figure 2). In this model individuals could transition among six states: alive in the study area ("I"), alive outside the study area ("O"), dead by fishing in the study area ("DFI"), dead by other (unknown) causes in the study area ("DUI"), dead outside the study area ("DUO"), and dead for a long time (" †"). By "inside the study area" we mean the area of the ocean where sampling occurred, whereas "Alive outside the study area" is a mathematical concept, rather than a geographical area, that allowed us to account for the possibility of some tagged individuals moving into a state where they remain alive but unobservable. Note that states "O", "DUI", and "DUO" are not observable and can only be linked to the event "not encountered" (see below and Figure 2): here, DUI and DUO states reflect unobservable but recently dead individuals, whereas state "O" indicates that the individual is alive but unavailable for sampling. The state " †" is an additional unobservable dead state that was also included in the model definition to distinguish the observed recoveries or "newly dead" individuals from the unobservable "long-time dead" ones (see Lebreton et al., 1999). This classification allows estimating mortality proportions due to fishing and tag-reporting rates (see below). Between each sampling occasion, sharks can change state according to the transitions shown in Ψ I→O : emigration (from inside to outside the study area, that is, areas where vessels that participated in the study traveled making tag recoveries still feasible).
Ψ O→I : immigration (from outside to inside the study area).
f: the probability of death due to fishing given that an animal has died in the study area.
These model parameters could be estimated separately by splitting the full transition matrix into a three-step series of transition matrices representing dispersal, survival, and cause-specific mortality processes, respectively (see Appendix S1). Our model pattern assumes that ecological processes occur before the observational ones, with dispersal being the first step in our sequence of transition matrices and survival the second. If an individual dies in the study area, it can transit to two dead states (one observable and one unobservable, see Figure 2), thus estimating the proportion of deaths due to fishing separately from other causes of mortality. Finally, the third and last step corresponds to the observational process and allows us to estimate event probabilities. Matrix E shows the event probabilities that link the biological states (rows) with the observations (columns).
where, p: the recapture probability of a marked animal that is alive; r: the reporting probability of a marked animal dead by fishing.
Events "1" and "2" are directly linked to model states "I" and "DFI" (i.e., they can only happen in these states) but event "0" (not encountered) arises from imperfect detection (see also Figure 2) and can be related to any possible underlying state in our probabilistic model. Because non-fishing deaths and those occurring Diagram showing the model pattern used in the analysis of the encounter data. Each step represents a different model parameter and the whole sequence links both ecological (ψ, S, f) and observational processes (r) to the different events contained in the individual encounter histories (the numbers between brackets).
outside the study area were never reported, their corresponding states can only be linked to event "0" (see also Figure 2). Given that no animals were recaptured alive in our study (only dead sharks were reported), the recapture probability (p) was always fixed to zero in our modeling.

| Goodness-of-fit test and model construction
Multi-event models were built and fitted to the data using the program E-SURGE (Choquet & Nogue, 2010). Prior to the model selection process, a Goodness-of-fit test was conducted to check if our data met the assumptions of a departure model that considers all parameters to be state and time-dependent, namely the Arnason-Schwarz model (Pradel et al., 2003). Goodness-of-fit tests were performed using U-CARE (Choquet et al., 2009), a statistical program that by means of contingency tables helps users to detect sources of lack of fit in their encounter data, which are mainly caused by differences in survival and detection probabilities among individuals.
To correct for those sources of lack of fit, we calculated an overdispersion coefficient or ĉ (the sum of chi-square results for each test divided by the total number of degrees of freedom) that was applied to the analyses in E-SURGE.
Model selection was based on Akaike's information criterion corrected for overdispersion (Quasi-AIC or QAIC), and we retained as good candidate models those showing the lowest QAIC values (Beier et al., 2001). Models differing in <2 points of QAIC from the top-ranked one (ΔQAIC <2) were also considered good candidate models (i.e., statistically equivalent).
Encounter data from both species were analyzed together under the same multi-event modeling approach. By analyzing both species together, we increased the amount of data available for making statistical inference allowing us to build models with more mathematical parameters, testing biological hypotheses, and quantifying rates that would not have been tested nor quantified otherwise. The model selection process departed from a general Modeling of r consisted of removing group and time interactions ("*") and on testing constancy ("."), time-only (t), group-only (species), and additive (+) time and group effects on this parameter until the most parsimonious (i.e., lowest QAIC) model structure was determined. Once a best structure for r was found, we kept that structure and repeated the same modeling process with f and S parameters until a consensus model, with the best supported structure for S, f, and r, had been retained. In both our departure model and in the subsequent modeling of S, f, and r parameters, we always kept immigration transitions fixed to zero (i.e., a permanent emigration structure). Alternative hypotheses regarding Ψ were also tested on the consensus model, to check whether they improved, or not, the retained model structure.

| RE SULTS
A total of 700 blue sharks and 132 shortfin makos were tagged (  Table 3).

| Goodness-of-fit testing and model selection results
The multistate Goodness-of-fit tests performed for the twospecies encounter history dataset yielded significant results and the ĉ coefficients resulting from each subset of data were all >1 (see Appendix S1). Such results indicated that the departure model used in the test (Arnason-Schwarz model) did not fit our data adequately and that overdispersion was present, yielding a global ĉ value of 1.89 that was applied as a correction factor when running the multi-event models in E-SURGE. In the multi-event modeling we departed from a more complex model (model 16,

| DISCUSS ION
By using an extensive tag-recovery dataset of more than 800 individuals, mainly juveniles, we were able to estimate important demographic parameters of two heavily exploited oceanic sharks: blue shark and shortfin mako shark. Survival rate was moderate for shortfin mako and high for blue shark; fishing mortality rep-  (Wood et al., 2007)]. The age and size at 50% maturity for blue shark is around 4 years and 210 cm TL for males and 5 years and 220 cm TL for females (Cailliet & Goldman, 2004;Dulvy et al., 2008;Yokoi et al., 2017 TA B L E 1 Summary of capture and recovery data for shortfin mako and blue shark.

F I G U R E 3
Length-frequency distributions of shortfin mako (Isurus oxyrinchus) and blue shark (Prionace glauca) tagged in this study.

TA B L E 2
Tagging and recovery information of blue shark (Prionace glauca) obtained during this study. Diff FL= difference in fork length between tagging and recovery; Sex (female, 0; male, 1).  (Dulvy et al., 2008;Natanson et al., 2006;Semba et al., 2009;Yokoi et al., 2017). This information suggests that only 20.0% of the male population and 2.1% of the female population would reach the age at which 50% are mature in the North Atlantic (34% of males and 9% of the female population according to Wood et al., 2007). based on mark-recapture methods (Wood et al., 2007), and 0.19-0.56 based on satellite tagging data (Byrne et al., 2017

TA B L E 2 (Continued)
TA B L E 3 Tagging and recovery information of shortfin mako (Isurus oxyrinchus) obtained during this study. Diff FL=difference in fork length between tagging and recovery; Sex (female, 0; male, 1).  (Camhi & Pikitch, 2008;Queiroz et al., 2016). In fact, the estimated global fishing capture of blue sharks reached 100,000 tons landed in the period 2016-2022, with a peak in 2016 (more than 110,000 t) and slight decrease over last years (FAO, 2023). The high proportion of mortality due to fishing in both blue shark and shortfin mako is not surprising given the high overlap between these species' spatial distribution and preferred fishing areas of vessels, having on average 62% and 76% of their space use, respectively, overlapped by longlines each month (Queiroz et al., 2016(Queiroz et al., , 2019. Furthermore, in the North Atlantic fishing-induced mortality (catch per unit effort) of pelagic sharks has been demonstrated to be higher where the overlap between shark space-use hotspots and longline fishing effort is greater , which underlies the long-term declines in abundance of these species attributed to overfishing .
Our results suggest that more than one-third of the tagged sharks may have moved out of the study area permanently. The long-distance, wide-ranging movements observed in this study and the known highly migratory nature of these sharks suggest, in agreement with previous studies, that there is a single well-mixed population in the entire North Atlantic for both species (Schrey & Heist, 2003;Veríssimo et al., 2017), including global panmixia (Corrigan et al., 2018). Habitat selection and use of space studies of blue sharks have provided evidence for the existence of a central North Atlantic nursery where blue shark juveniles can reside for up to at least 2 years (Vandeperre et al., 2016). After birth, juveniles spatially segregate with different ontogenic movements, where females travel toward tropical latitudes and males display diverse behavioral strategies (Vandeperre et al., 2014). In the case of shortfin mako, newborns and juveniles may be dispersed over a broad geographical area from the Gulf Stream in the west (Kohler et al., 2002) to the African coast in the east (Dinkel & Sánchez-Lizaso, 2020). In this work, differences between sexes or sizes were not explored due to data limitations, although they represent a natural next step.
Our results show that two thirds stayed in the study area, indicating that there are preferred areas of space-use hotspots in the North Atlantic, explaining and support the findings about overlap between fishing effort and blue and shortfin mako space use (Queiroz et al., 2016(Queiroz et al., , 2019. The tag-reporting rate (percentage of returning information when a tagged shark is caught) in our study was relatively high (0.794), considering the possible loss of information during longdistance movements and lack of motivation for reporting by some fishers, and was similar to reporting rates of coastal shark species like the sand tiger shark (Carcharias taurus; 0.753; Dicken et al., 2006). Given the highly migratory nature of the blue shark and shortfin mako a lower rate could be expected; however, the result is consistent with high spatial overlap between fishing activity of longliners (between 67% and 76% per overlap per month) and the range of oceanic shark species where higher tag-reporting rates are feasible (Mucientes et al., 2022;Pacoureau et al., 2021;Queiroz et al., 2016Queiroz et al., , 2019. The recovery rate for both species was also relatively high (18.11% and 8.84% for makos and blue sharks) compared with other studies in Atlantic Ocean that reported recovery rates ranging from 9.4% and 13.5% for mako, and from 4.9% and 11.9% for blue shark (Casey & Kohler, 1992;Kohler & Turner, 2001, 2019Wood et al., 2007 (Rigby, Barreto, Carlson, Fernando, Fordham, Francis, Herman, et al., 2019;. The results of our study (high fishing mortality rates, particularly among juveniles, and low chances to reach maturity) support the

DATA AVA I L A B I L I T Y S TAT E M E N T
The data (shark recoveries) that supports the findings of this study are available in the Tables 2 and 3 of this article. All tagging information (tagged sharks) of this study are available from the corresponding author upon reasonable request.