Elsevier

Fisheries Research

Volumes 119–120, May 2012, Pages 115-127
Fisheries Research

Estimating fishing and natural mortality rates for Pacific bluefin tuna (Thunnus orientalis) using electronic tagging data

https://doi.org/10.1016/j.fishres.2011.12.015Get rights and content

Abstract

This paper presents estimates of fishing and natural mortality rates derived from a spatially- and seasonally structured Bayesian mark-recapture model for electronically tagged Pacific bluefin tuna (PBFT) (Thunnus orientalis). Fishing mortality rates (F) were estimated by age group, year, quarter and area and ranged between 0.02 and 1.92 quarter−1 for the northeastern Pacific Ocean (EPO) and 0.18 and 0.54 quarter−1 for the northwestern Pacific Ocean. Annual Fs in the EPO were on average 2–3 times higher than the estimated rate of natural mortality for Pacific bluefin tuna aged 2 and 3 and 4–6 times higher than the estimated rate of natural mortality for Pacific bluefin tuna aged 4 and older. The estimate of M for PBFT aged 5 and above (median 0.15 yr−1, standard deviation = 0.10) was lower than the value currently used in the PBFT stock assessment (i.e. 0.25 yr−1). In addition to estimating age-group specific natural mortality rates (M), the plausibility of alternative values for M was evaluated by fixing it at the age-specific schedules tested in the PBFT stock assessment and computing a Bayesian model selection criterion (the Deviance Information Criterion, DIC) for alternative M configurations. For models in which M was fixed, the lowest DIC was obtained for the M scenario that assumed the lowest value of M for PBFT aged 4 and above (i.e. 0.12 yr−1).

Highlights

► We analyze electronic tagging data for Pacific bluefin tuna (PBFT). ► We use mark-recapture models to estimate fishing and natural mortality rates. ► Quarterly track positions from archival tag data were used to estimate movement rates. ► We obtained an estimate of natural mortality for adult PBFT of 0.15 yr−1. ► Estimated rates of fishing mortality may exceed sustainable levels.

Introduction

Pacific bluefin tuna (Thunnus orientalis) is a highly migratory species, having the largest home range of any tuna in the genus Thunnus. Pacific bluefin tuna, hereafter PBFT, are primarily distributed throughout the temperate waters of the northern Pacific Ocean but also range into the western South Pacific (Bayliff, 1994, Collette and Nauen, 1983). Two spawning areas are currently recognized; one in the southwestern North Pacific off Taiwan and the other in the Sea of Japan (Okiyama, 1974, Chen et al., 2006, Tanaka et al., 2007). PBFT in spawning condition have been taken in these waters in the months of April through August. Genetic structure has not been detected to date in PBFT, and one stock is currently recognized throughout the Pacific Ocean.

Commercial fisheries exist for PBFT throughout their range and are most intensive in the northwestern (WPO) and northeastern Pacific Ocean (EPO). Annual reported ocean-wide catches since the early 1950s have varied between 8000 and 35,000 tons (IATTC, 2005), averaging around 25,000 tons between 1950 and 1980 and 15,000 tons thereafter. Catches in the EPO have declined since the 1960s, although the last decade saw catches rapidly increase (Aires-da-Silva et al., 2007). Concomitantly, fishing effort in the EPO decreased between 1960 and 1990 but has rebounded since the late 1990s in concert with expansion of the PBFT ranching industry off Baja Mexico (Aires-da-Silva et al., 2007). PBFT are also exploited by a variety of fisheries in the WPO, including purse-seine, longline and gillnet fisheries (Bayliff, 2001).

A preliminary stock assessment was conducted for PBFT by the Inter-American Tropical Tuna Commission (IATTC) (Bayliff, 2001). However, the available catch and effort data and length-frequency data for the WPO were insufficient for calculation of abundance indices for the WPO or to conduct cohort analysis for the entire Pacific Ocean (Bayliff, 2001). A full stock assessment including updated catch and effort and length-frequency data for the WPO was conducted by the International Scientific Committee for Tuna and Tuna-like Species in the North Pacific Ocean (ISC) in 2008 (Anon., 2008). Estimates of PBFT fishing mortality rates (F) in recent years were above fishing mortality levels corresponding to target reference points such as FMSY (the fishing mortality that would achieve maximum sustainable yield) (ISC, 2008).

The stock assessment for PBFT is subject to uncertainty about several basic life-history characteristics, foremost among them the rate of natural mortality, M (Bayliff, 2001). M is of primary importance in fish stock assessments as it determines potential maximum sustainable yield (MSY) (Beddington and Cooke, 1983, Beamesderfer and North, 1995) and a population's response to harvest. Higher rates of natural mortality are associated with more productive stocks and their ability to sustain higher harvest rates. However, for most exploited fish populations, M is typically difficult to estimate because it is not feasible to monitor natural deaths (Quinn and Deriso, 1999) and it is difficult to separate the effects of natural mortality, fishing mortality and recruitment in observations of fish abundance (Hightower et al., 2001, Hilborn and Walters, 1992). Moreover, methods available for the estimation of M for fish populations have onerous data requirements and assumptions that are not easily satisfied. For example, estimation of M using catch curve analysis (Ricker, 1975, Vetter, 1988) requires data from an unexploited or very lightly exploited population and when catch-at-age data become available, exploitation rates are typically non-negligible. In contrast, tagging experiments with sufficiently rigorous tagging design and high tag recovery rates are often not feasible in many instances (Walters and Martell, 2004).

M has commonly been identified as a key source of uncertainty in fish stock assessments (Lapointe et al., 1989, Hilborn and Walters, 1992). The treatment of age-specific Ms as fixed parameters fails to capture this uncertainty, which could ultimately lead to loss of yield or overexploitation of the stock. Currently an age-specific vector for M, with M for ages 4 and above equal to 0.25 yr−1 is used in the PBFT stock assessment. Lower values for M for fish of age 4 and older yielded estimates of the depletion of spawning biomass between 1952 and 2005 (<5%) and unfished spawning biomass (about 1.4 million tons) that were considered implausible by the working group (Anon., 2008, Aires-da-Silva et al., 2009). Currently, only the value of M for age-0 PBFT has been empirically derived using tagging data (Takeuchi and Takahashi, 2006). Values for fish of age 1 and older were derived using life-history based methods and estimates for southern bluefin tuna (Thunnus maccoyi) (Polacheck et al., 1997). However, when considered together, values for M of 0.25 yr−1 and above with moderate to high fishing mortality rates for adult PBFT may be incongruous with their reported longevity (maximum age of 26 years, Shimose et al., 2009). More information on M for PBFT aged 4 and above may thus be useful to help reduce the substantial uncertainty associated with this parameter.

Tagging data can be informative with respect to rates of fishing and natural mortality. A major caveat to using tagging data is a requirement that the tag reporting rate is known or can be reliably estimated (Bacheler et al., 2009, Pollock et al., 1991, Pollock et al., 2001). Telemetry has been applied to estimate natural mortality rates (mortality rates are inferred from transmitters that stop moving over successive time periods) (Hightower et al., 2001, Waters et al., 2005). More recently, tag return data (e.g. from archival or conventional tags) and telemetry data have been combined (Pollock et al., 2004, Bacheler et al., 2009) to provide information about both return rates and natural mortality rates.

State-space Bayesian tagging models (Michielsens et al., 2006, Kurota et al., 2009, Taylor et al., 2011) offer a powerful framework for analysis of tagging data and provide a statistically rigorous means to quantify uncertainty, accounting for both observation and process error (Wade, 2000, Michielsens et al., 2006). We extend this methodology to the analysis of electronic tagging data for PBFT. In this paper we seek to use electronic tagging data to examine age-group specific rates of natural mortality and seasonal movement patterns of PBFT. The use of a sequential Bayesian approach facilitates the flow of information from different data sources whereby the posterior from one analysis becomes the prior for the next (Michielsens et al., 2008). In the context of the current analysis, posterior probability density functions (pdfs) for seasonal movement rates from analysis of pop-up satellite archival tag (PAT) data were used as the prior for the archival tag model (Kurota et al., 2009). Our analysis additionally includes quarterly locations for archival tagged PBFT between the release and recapture locations, thereby using the additional spatial information available from the electronic tag data.

The quantitative integration of conventional and electronic tag data into spatially structured assessment models is opening the door to an improved understanding of the population dynamics of highly migratory species (Goethel et al., 2011). Integrating tagging data with other fisheries data (typically catch-at-age and catch-per-unit-effort data) has advanced stock assessment models for several species (e.g. sablefish (Anoplopoma fimbria) (Haist, 1998); yellowfin tuna (Thunnus albacares) (Hampton and Fournier, 2001) and Atlantic bluefin tuna (Thunnus thynnus) (Taylor et al., 2011)).

The number of electronic tags placed on PBFT has increased rapidly in the past decade with several archival and satellite tagging projects on both sides of the Pacific basin (Kitagawa et al., 2000, Kitagawa et al., 2007, Boustany et al., 2010). In the EPO over 550 juvenile PBFT (primarily year classes 2 and 3) have been tagged with archival and pop-up satellite technologies, and a high archival tag recapture rate has resulted in a rich data set of over 58,000 tag days (Fig. 1), including track lengths exceeding 4 years (Kitagawa et al., 2007, Boustany et al., 2010, Block et al., 2011). These data provide new information on the migrations of PBFT and have revealed that PBFT in the EPO have predictable seasonal patterns of north-south movement that recur annually in the California Current in relationship to seasonal oceanographic changes (Kitagawa et al., 2007, Boustany et al., 2010, Block et al., 2011). The positional data also show that the California Current is retentive for young PBFT that migrate to the EPO, usually for one to four years with eventual return to the WPO along the North Pacific transition zone front (Boustany et al., 2010). Archival tagging experiments have also been carried out in the WPO, primarily to study ambient temperature preferences and migration patterns in young PBFT (Itoh et al., 2003a, Itoh et al., 2003b, Kitagawa et al., 2006, Kitagawa et al., 2009).

This paper presents results from a novel tagging model based on EPO electronic tag data sets where fishing mortality rates were directly estimated from archival tag returns. With the relatively long times at large for electronically tagged PBFT, particularly in the early years of the study (2002–2004, Fig. 2) and the high recovery rates of tags, the tagging data should also be informative with respect to age-specific Ms. The mark-recapture model is spatially structured and allows estimation of seasonal movement probabilities, which is desirable for robust estimation of exploitation rates for highly migratory fish such as tunas (Walters and Martell, 2004). Results are presented for informative and uninformative priors for age-group specific Ms. The consistency of alternative candidate M vectors proposed for the PBFT stock assessment with the structure of the tagging data set was evaluated by computing the Deviance Information Criterion (DIC) for each fixed M-vector (Spiegelhalter et al., 2002, Bolker, 2008).

Section snippets

Data

Electronic tagging data covering the period 2000–2009 were used in analyses described below. Tagging experiments, inclusive of deployments utilizing archival tags and pop-up satellite archival tags (PAT tags) were conducted under the Tagging of Pacific Pelagics (TOPP) Program (see Kitagawa et al., 2007, Boustany et al., 2010, Block et al., 2011). PAT tags (N = 35) were applied between 2000 and 2008 off the Baja California coast, with a maximum programmed time at large of 365 days. Pop-up

Model structure

The Pacific Ocean was subdivided into three areas (boxes) for the purposes of the model. These areas correspond roughly to Japan and trans-Pacific (box 3); California (box 1) and Mexico (box 2) (Fig. 1). The boundary between the northern and southern boxes in the EPO was assigned such that an approximately equal number of releases occurred on either side of the boundary to facilitate estimation of movement rates between the EPO boxes. This division was also chosen as it coincides approximately

PAT model

The model of PAT tags included only those tags for which pop-up satellite tag location data were available (about 90%). This model was primarily used to estimate prior distributions for seasonal movement rates between two areas (north and south EPO, boxes 1 and 2). Fishing and natural mortality rates were not estimated in the PAT model. Few (<5) tags were known to have been recovered by fishers; the majority of tags used in this analysis were recovered as a result of release from fish at or

Archival tag model

Fishing mortality rates were estimated by area group, where annual patterns in fishing mortality were estimated for the northeastern Pacific Ocean (boxes 1 and 2, Fig. 1) (EPO) and the northwestern Pacific Ocean (box 3, Fig. 1) (WPO). 94% of recaptured archival tags were recovered in the EPO boxes, so this analysis is intended primarily to estimate fishing and natural mortality rates for the EPO.

Population dynamics

For notational simplicity, subscripts for year, season and area of release are omitted from the following equations. The subscripts below refer to the year, season and area for the current quarter at large.

PAT tags

In the equations that follow, S denotes the number of seasons (4), while J denotes the number of areas (2) in the PAT tag model. The number of tagged fish in area k with tags programmed to pop-up after l quarters at large in year y and season s, in the first quarter at large (i.e. the quarter of release) is given by:Nk,l,y,s=Rj,l,y,sps,j,kwhere Rj,l,y,s is the number of tagged fish released in area j in year y and season s programmed to pop-up after l quarters at large and ps,j,k is the

Archival tags

In the equations that follow, S denotes the number of seasons (4), while J denotes the number of areas (3) in the archival tag model. The archival tag model describes the survival, movement and capture of archival tagged fish. The number of tagged fish of age a in area k in year y and season s, in the first quarter at large (i.e. the quarter of release) is given by:Na,y,s,k=j=1JRa,y,s,jps,j,k(1γ)where Ra,y,s,j is the number of tagged fish of age a released in area j in year y and season s

Results

This study examined the data set originating from deployment of 523 tags between August 2002 and October 2009. Over 270 archival tagged PBFT were recaptured during the study; recovery rates of archival tagged fish were substantial, varying between 43 and 75% by deployment year excluding the final year of the study (Table 3, final column). A summary of release and recapture numbers by year is given in Table 3. The mean time at liberty for archival tagged PBFT was 600 days; times at liberty

PAT tags

PAT tags were deployed most often in the third quarter (around two thirds). The estimated probabilities for a tag to pop-up and report in a given quarter showed that tags had the highest probability of releasing during the programmed quarter (Fig. 4). Tags were more likely to pop-up before than after the programmed quarter. The probability of releasing and surfacing 2 or 3 quarters before the programmed pop-up date was high (54%), while there was a low probability for tags to release after the

Archival tags

Priors for season-specific movement rates from the PAT model were further updated by the archival tagging data. The following results are shown for the base case model with informative priors for age group specific natural mortality rates Mg(a) and tag reporting rates (Table 1) unless otherwise stated.

Informative priors for age group specific Ms were updated by the tagging data, particularly for ages 5+ (age group 3, Fig. 5a). The estimated posterior median M for ages 2 and 3 (age group 1) was

Discussion

Electronic tagging data provide new opportunities for improving our understanding of the spatial and temporal components of fish population dynamics. They offer information on the movements, stock structure, life history and feeding ecology of tunas (e.g. Bestley et al., 2009, Kurota et al., 2009, Boustany et al., 2010) at finer spatial scales than were previously possible. While numerous papers have examined foraging ecology or diving behaviors, relatively few have used electronic tagging data

Acknowledgements

The tagging portion of this study was funded by the Tagging of Pacific Pelagics program (TOPP). Electronic tagging and recovery was supported by funds from the Sloan and Moore Foundations. The Monterey Bay Aquarium Foundation along with the Office of Naval Research (ONR) supported electronic tagging and shipboard activities. We thank K. Schaefer and D. Fuller and the Inter-American Tropical Tuna Commission for assisting in the return of electronic tags. R. Whitlock was funded by ONR and the

References (58)

  • A.M. Boustany et al.

    Movements of pacific bluefin tuna (Thunnus orientalis) in the Eastern North Pacific revealed with archival tags

    Prog. Oceanogr.

    (2010)
  • T. Shimose et al.

    Age determination and growth of Pacific bluefin tuna, Thunnus orientalis, off Japan and Taiwan

    Fish. Res.

    (2009)
  • Aires-da-Silva, A., Hinton, M.G., Dreyfus, M., 2007. Standardized catch rates for Pacific bluefin tuna caught by United...
  • Aires-da-Silva, A., Maunder, M., Deriso, R., Piner, K., Lee, H., 2009. A sensitivity analysis of alternative natural...
  • Anon., 2008. Report of the Pacific Bluefin Tuna Working Group Workshop (Annex 7). International Scientific Committee...
  • N.M. Bacheler et al.

    A combined telemetry—tag return approach to estimate fishing and natural mortality rates of an estuarine fish

    Can. J. Fish. Aquat. Sci.

    (2009)
  • Bayliff, W.H., 1994. A review of the biology and fisheries for northern bluefin tuna, Thunnus thynnus in the Pacific...
  • Bayliff, W.H., 2001. Status of bluefin tuna in the Pacific Ocean. IATTC Stock Assessment Report...
  • R.C.P. Beamesderfer et al.

    Growth, natural mortality, and predicted response to fishing for largemouth bass and smallmouth bass populations in North America

    N. Am. J. Fish. Manage.

    (1995)
  • J.R. Beddington et al.

    On the potential yield of fish stocks

    FAO Fish. Tech. Pap.

    (1983)
  • S. Bestley et al.

    Predicting feeding success in a migratory predator: integrating telemetry, environment, and modeling techniques

    Ecology

    (2009)
  • B.A. Block et al.

    Tracking apex marine predator movements in a dynamic ocean

    Nature

    (2011)
  • B.A. Block et al.

    Electronic tagging and population structure of Atlantic bluefin tuna

    Nature

    (2005)
  • B.M. Bolker

    Ecological Models and Data in R

    (2008)
  • K.S. Chen et al.

    Reproductive biology of female Pacific bluefin tuna Thunnus orientalis from the south-western North Pacific Ocean

    Fish. Sci.

    (2006)
  • B.B. Collette et al.

    FAO Species Catalogue. Vol. 2. Scombrids of the world. An annotated and illustrated catalogue of tunas, mackerels, bonitos and related species known to date

    FAO Fish. Synopsis

    (1983)
  • A. Gelman et al.

    A single series from the Gibbs sampler provides a false sense of security

  • D.R. Goethel et al.

    Incorporating spatial structure in stock assessment: movement modeling in marine fish population dynamics

    Rev. Fish. Sci.

    (2011)
  • V. Haist

    Integrated catch-age mark-recapture model: application to B.C. sablefish stocks, in Fishery Stock Assessment Models

  • J. Hampton

    Natural mortality rates ion tropical tunas: size really does matter

    Can. J. Fish. Aquat. Sci.

    (2000)
  • J. Hampton et al.

    A spatially-disaggregated, length-based, age-structured population model of yellowfin tuna (Thunnus albacares) in the Western and Central Pacific Ocean

    Mar. Freshwater Res.

    (2001)
  • J.E. Hightower et al.

    Use of telemetry methods to estimate natural and fishing mortality of striped bass in Lake Gaston, North Carolina

    Trans. Am. Fish. Soc.

    (2001)
  • R. Hilborn et al.

    The Ecological Detective: Confronting Models With Data

    (1997)
  • R. Hilborn et al.

    Quantitative Fisheries Stock Assessment: Choice, Dynamics, and Uncertainty

    (1992)
  • Inter-American Tropical Tuna Commission, 2005. Tunas and billfishes in the Eastern Pacific Ocean in 2004. Fishery...
  • International Scientific Committee for Tuna and Tuna-like Species in the North Pacific Ocean, 2008. Report of the...
  • International Scientific Committee for Tuna and Tuna-like Species in the North Pacific Ocean, 2009. Report of the ninth...
  • T. Itoh et al.

    Migration patterns of young Pacific bluefin tuna (Thunnus orientalis) determined with archival tags

    Fish. Bull.

    (2003)
  • T. Itoh et al.

    Swimming depth, ambient water temperature preference, and feeding frequency of young Pacific bluefin tuna (Thunnus orientalis) determined with archival tags

    Fish. Bull.

    (2003)
  • Cited by (31)

    • Times are changing, but has natural mortality? Estimation of mortality rates for tropical tunas in the western and central Pacific Ocean

      2022, Fisheries Research
      Citation Excerpt :

      Mark-recapture data can be analysed by various models depending on the frequency of tag release events and other modelling assumptions. Size and age influences on natural mortality can be estimated when tagging data are combined with other sources of data, e.g. growth (Hampton, 2000), length (Hillary and Eveson, 2015), catch (Krause et al., 2020), or multiple tag types (Kurota et al., 2009; Whitlock et al., 2012). Tag attrition models (Kleiber et al., 1987; Hampton, 2000) are a general framework for analysing tagging data that can estimate mortality parameters, which in the simplest case are assumed constant over time.

    • Length-based approaches to estimating natural mortality using tagging and fisheries data: The example of the eastern Aleutian Islands, Alaska golden king crab (Lithodes aequispinus)

      2022, Fisheries Research
      Citation Excerpt :

      Tag-recapture data have been used to estimate various population parameters, such as abundance, mortality, and movement (Seber, 1982) and provide a promising avenue for M estimation. Multinomial likelihood is widely used in tag-recapture data analysis (e.g., Seber, 1982; Pollock et al., 2002) and some variants, such as Poisson distribution (e.g., Hilborn, 1990) and negative binomial distribution (e.g., Whitlock et al., 2012), are used to estimate mortality when movement is an integral part of the analysis. Methods such as simple linear regression (Balsiger, 1974), least squares (Siddeek et al., 2002), likelihoods (Pollock et al., 2002), and Bayesian mark-recapture models (Whitlock et al., 2012) have been used with tag-recapture data to estimate various population dynamics parameters (including M) in various fish and crustacean populations.

    • An Introduction to Modelling Abundance and Life History Parameters in Shark Populations

      2017, Advances in Marine Biology
      Citation Excerpt :

      Researchers are often unable to estimate more complex parameters, like M, directly. As stated by Whitlock et al. (2012), “estimation of M using catch curve analysis (Ricker, 1987; Vetter, 1988) requires data from an unexploited or very lightly exploited population and when catch-at-age data become available, exploitation rates are typically nonnegligible. In contrast, tagging experiments with sufficiently rigorous tagging design and high tag recovery rates are often not feasible in many instances (Walters and Martell, 2004)”.

    • Bayesian analyses of Pacific swordfish (Xiphias gladius L.) genetic differentiation using multilocus single nucleotide polymorphism (SNP) data

      2016, Journal of Experimental Marine Biology and Ecology
      Citation Excerpt :

      Yet, those samples were identified as members of the temperate cluster, in spite of the enormous distances (≈ 11,000–18,000 km) between them. Whereas long transoceanic migrations have been documented in other billfish, tunas, and sea turtles (Kobayashi et al., 2008; Ortiz et al., 2003; Whitlock et al., 2012; Fonteneau and Hallier, 2015), PSAT tagging studies in swordfish have failed to document trans-Pacific crossings. While this may be explained in part by the relative short duration of most deployments (< 1 year), several tracks suggests swordfish were starting to return to the release area towards the end of the time period monitored.

    View all citing articles on Scopus
    View full text