CCAMLR

We present developments towards spatially explicit age-structured population dynamics operating models for Antarctic toothfish in the Ross Sea. The operating models consider both a coarse-scale and fine-scale spatial resolution and consider scenarios where abundance can be present over the entire Ross Sea region or constrained to areas where the fishery has operated. The models represent developments towards plausible operating model constructs that may be used for evaluating assessment biases or other factors to quantify risk and uncertainties for management of the Ross Sea Antarctic toothfish fishery. Further, we outline steps towards using the model as an estimation model to estimate movement and spatial distribution parameters as an aid in evaluating the operating models. The models are implemented in the generalised Bayesian population dynamics model, the Spatial Population Model (SPM). The SPM program allows implementation of an aggregate movement model for use with large numbers of areas as a discrete time-step state-space model that represents a cohort-based population age structure in a spatially explicit manner. Models can be parameterised by both population processes (i.e., ageing, recruitment, and mortality), as well as movement processes defined as the product of a set of preference functions that are based on known attributes of spatial location. The operating models considered were single sex age-structured models that categorised fish as immature, mature, or spawning. Observations can include spatially explicit commercial catch proportions-at-age, proportions mature, and tag-release and tag-recapture observations. Estimates of parameters when the operating models were used as estimation models with observations from the Ross Sea Antarctic toothfish fishery appeared to broadly reflect the hypothesised spatial distribution of Antarctic toothfish, suggesting that younger fish were found predominantly in the southern shelf areas and adult fish distributed along the slope and northern areas of the Ross Sea. Fits to the commercial catch proportions-at-age observations were generally good in most models, although fits to the plus group of the proportions-at-age catch data were less than ideal. Model estimates of proportions-mature appeared to be sensible, with a clear pattern that the proportions mature were a function of location and age. Tag release and recapture data were less well fitted by the models due, in part, to the conflict with assumptions of known abundance in the model and the abundance information inherent in the tag-recapture observations.

The Spatial Population Model (SPM) is a generalised spatially explicit age-structured population dynamics and movement model. SPM can model population dynamics and movement parameters for an age-structured population using a range of observations, including tagging, relative abundance, and age frequency data. SPM implements an age-structured population within an arbitrary shaped spatial structure, which can have user defined categories (e.g., immature, mature, male, female, etc.), and age range. This manual describes how to use SPM, including how to run SPM, how to set up an input configuration file. Further, we describe the population dynamics and estimation methods, and describe how to specify and interpret output.

The ICESCAPE Software Users Guide (Appendix 1) describes a parametric bootstrap (Davison and Hinkley 1997) model for implementing the general abundance estimator proposed by Southwell (2004) for Antarctic land-breeding predators. The software, which was developed as a suite of routines in the R language for statistical computing (R Development Core Team 2008), aims to adjust raw count data for availability and perception bias, and account for sampling fractions less than unity, in order to standardise estimates of breeding populations derived from count data collected under variable conditions. In Antarctica, adjustment for availability is particularly important, because difficulties in accessing breeding sites often lead to counts of land-breeding predators being taken at suboptimal times in a breeding cycle when the availability fraction is substantially different from unity. Adjustment for availability standardises counts to a common reference point of breeding chronology, and is achieved by applying an adjustment factor based on time series of availability throughout a breeding season. Such time series are typically collected at only a limited number of sites, so a search algorithm is used to determine surrogate availability information for a site when none exists. Importantly, standardisation in this way allows site-specific estimates to be aggregated to achieve region-scale population estimates. ICESCAPE uses a bootstrap resampling framework to propagate uncertainties associated with adjustments for availability bias, perception bias and sampling fraction through to final standardised population estimates.

The current assessment method for the mackerel icefish (Champsocephalus gunnari) in sub-area 48.3 employs the CMIX and GYM packages, deriving population numbers-at-age then projecting these numbers forward under the given harvest control rule to set a two-year TAC. One issue is the accurate identification of cohorts from the survey model, given the differential age structures observed in the population at Shag rocks and around South Georgia. In this paper we propose a length-based approach that removes this issue of determining cohorts, using stratified bootstrap techniques to estimate the length distribution of the population from the survey data which, in conjunction with the bootstrapped survey biomass data, yield an estimate of the population numbers-at-length. A length-based population projection is then used to estimate the two-year TAC, given the relevant harvest control rule.

In this paper a simple biomass dynamic population model is used as a paradigm for the dynamics of toothfish spp. in the CCAMLR area. This model underpins a management strategy evaluation framework where the process of stock assessment, TAC calculation and implementation and potential changes in future toothfish productivity (continuous and regime-based) are all simulated. The robustness of the current harvest control rule employed for assessed stocks of toothfish spp. in the convention area is explored with particular attention given to levels of biomass depletion (initial conditions), assessment precision, time-horizon, implementation error and future changes in productivity. An alternative control rule, using target and limit exploitation rates not biomass depletion levels and shorter time-horizons is explored and subjected to similar simulation testing.

The catch-at-age based CASAL model for toothfish in Subarea 48.3, which was presented in 2007, is developed and extended here to include survey data. Additional adjustments to the model suggested by WG-FSA-2007 are implemented. Fits to the survey length frequencies are good for recent surveys but in the past some cohorts have been missed by the survey. Fits to the survey CPUE are generally not very good. There is evidence for a size-based influence on post-tagging mortality, post-tagging growth retardation period and rate of tag loss. When these factors are included in the model the fits to the tag recaptures at size improve. However, there is still some evidence for a sex-based influence on the fits, particularly the poor fits observed from tags released in 2004 where significantly more males than females were recovered. A sex-disaggregated model was created, providing substantially similar fits and results to the sex-aggregated model, and no significant improvement to the tagging data fits. Estimates of cohort strength are made by the new model, but there is uncertainty about the estimates of recent cohorts made using short runs of catch-at-age data and the catchability being estimated by the survey. Estimates of cohort strength should be incorporated in calculations of TACs. A number of different mechanisms for doing this are proposed which maintain SSB above 50% B0 in both the short and long term.

An attempt to use CPUE data based on data of erussian observers in the assessment of Antarctic toothfish stock in division 58.4.1 is made. Unlike some previous results obtained for this division (e.g. WG-SAM-08/4), the assessment was intended to be independent of estimates of toothfich stock biomass in the Ross Sea. As a result of implementation of a dynamic production model, current biomass in this division is estimated as about 15000 t, while the estimate of virgin biomass B0 is found to be about 32000 t.

In order to clarify stock status and biological characteristics of Dissostichus spp. in Division 58.4.4 a & b, a total of 124 research hauls are allocated on 10-minute latitude x 20-minute longitude grid points (spaced approximately 10 nautical miles apart). Almost all of the research hauls are distributed between 500 and 2,000 m sea depths over each SSRU, except only 2 hauls which are distributed at shallow points of < 500 m in SSRU A in order to avoid catching small fishes. A Trot line system will be employed for 93 research hauls (three quarters of the total) in order to make an evaluation of the stock status and biological characteristics of Dissostichus spp. with the information collected by the Trot line system in the same Division in 2007/08. The applied methodology is also considered to maintain the consistency of the data from the two serial survey seasons, 2008/09 and 2009/10, and lead the efficient utilities of the data and reducing uncertainties. In 31 hauls (a quarter of total sets), the experimental gear, which consists of three segments of Trotline system and Spanish line system respectively within one fishing line, will be used in order to standardize the CPUE of Trot line more accurately in this Division. The proposed sample size is 140,000 kg for Dissostichus spp. It is calculated taking into account the need for completion of the proposed survey (124 research hauls) and impacts on the stock. Concentrated tagging, 3 fish/ton, will be made to apply the mark-and-recapture studies.

A survey was conducted in order to collect information for the stock status and various biological information on toothfish in Division 58.4.4a & b by using a commercial bottom longline vessel, Shinsei Maru No. 3, from July 23 to September 27, 2008. The survey was undertaken in two Phases (Phase 1 and Phase 2). Shinsei Maru Trot line system was used as a fishing gear. Toothfish contributed 67-70 % in numerical number and 90-96 % in weight to the total samples in SSRUs. Mean CPUE of toothfish in number by SSRU ranged from 7.2 to 15.1 indiv. / 1,000 hooks and from 10.3 to 18.7 indiv. / 1,000 hooks for Phase 1 and Phase 2, respectively. Mean CPUE of toothfish in weight by SSRU ranged from 64.0 to 80.7 kg / 1,000 hooks and from 120.8 to 308.6 kg / 1,000 hooks for Phase 1 and Phase 2, respectively. Using a grid-based (10-minute latitude x 20-minute longitude) standard stratification method, mean CPUE for the Division was estimated to be 11.0 indiv. / 1,000 hooks and 108.8 kg / 1,000 hooks. With the possibly overestimated conversion factor between CPUEs of Trotline and Spanish line systems, the results of this survey suggest the recovery of the stock since 2000/01.

The integrated assessment of Patagonian toothfish, Dissostichus eleginoides, for the Heard and McDonald Islands (Division 58.5.2) was updated by replacing catch-at-length proportions from commercial catches with catch-at-age proportions using age length keys (ALKs) where the ALK for each combination of fishery and year had available at least 50 aged fish. For the trawl fisheries that were divided into periods within each year, the same ALK for the year was applied to the length frequency (LF) samples for each fishing period within that year. For years where insufficient fish were aged the catch-at-length proportions were retained but for a given fishery the same age-specific selectivity function and parameter values were logically applied to both types of data. For 2006 and 2007 random stratified trawl surveys, there were sufficient aged fish to convert abundance-at-length to abundance-at-age data. Effective sample sizes for the commercial catch-at-age proportions, assuming a multinomial distribution, and the coefficient of variation (CV) for the abundance-at-age, assuming a lognormal distribution, each took into account uncertainty due to haul-level variability in catch-at-length proportions, ALK sampling error (sampling fraction of the LF samples that were aged ranged from 0.8% to 18%) and random ageing error. CASAL allows a single ageing error matrix to be defined and applies this matrix to predictions of numbers-at-age and proportions-at-age. In other work, this matrix was found to depend on the readability score of the otoliths used for ageing, and sensitivity of the assessment results to the assumed readability score was investigated for readability scores of moderate (3), good (4), and excellent (5). The median score for all aged fish in the assessment was 3 but some fishery-by-year combinations had a higher value of 4. The output from the integrated assessment of most interest in this study is the CV of the estimated historical recruitment series, since this parameter strongly influences the effect of the depletion rule on the allowable catch. Compared to the assessment that did not incorporate catch-at-age or abundance-at-age data, the aged-based assessment dramatically lowered the CV for the recruitment series, from around 1.5 to 1.8 down to approximately 0.3 to 0.4, if a readability score 5 was assumed or if for a score of 4 the most stable subset (1986-2000) of the full historical series (1984-2006) was used to estimate the CV. There was no reduction in CV for either series if a score of 3 was assumed. The difference between a readability of 3 and 4 in ageing error is that zero ageing errors are relatively less prevalent (e.g. for age 8 the percentage of errors that were zero was estimated from previous work at 40% for score 3 and 48% for score 4, the corresponding +/- 1 yr errors had prevalence of 46% and 45%, respectively). A +/- 1 year error may seem minor relative to the complete age range modelled of 1 to 35 yr, however most fish caught are in a more restricted, younger age range. For example, the upper age of fish in the main survey that have an upper selectivity greater than 0.2 was approximately 12 yr while the corresponding values for the trawl and longline fisheries were 15 and 20 yr, respectively. The results presented suggest that future ageing work would give a greater improvement to the integrated assessment if otoliths with readability score of at least 4 can be obtained in sufficient numbers to allow ALKs to be constructed using only the ages obtained from these otoliths.