Paul A. Breen

A length-based Bayesian stock assessment model for the New Zealand abalone Haliotis iris

We describe a length-based Bayesian model for stock assessment of the New Zealand abalone Haliotis iris (paua). We fitted the model to five data sets: catch-per-unit-effort (CPUE) and a fishery-independent survey index, proportions-at-length from both commercial catch sampling and population surveys, and tag–recapture data. We estimated a common component of error and used iterative re-weighting of the data sets to balance the residuals, removing the arbitrary data set weightings used in previous assessments. Estimates at the mode of the joint posterior distribution were used to explore sensitivity of the results to model assumptions and input data; the assessment itself was based on marginal posterior distributions estimated from Markov chain–Monte Carlo simulation. Assessments are presented for two stocks in the south of New Zealand. One may be recovering after recent catch reductions; the other is over-exploited and likely to decline further. Assessment for the first stock was robust; assessment for the second stock was sensitive to the CPUE data and may be too optimistic. We discuss future directions and potential problems with this approach.

A multi‐stock, length‐based assessment model for New Zealand rock lobster (Jasus Edwardsii)

Abstract We describe a multi‐stock, length‐based Bayesian assessment model for New Zealand spiny lobster (Jasus edwardsii) fisheries. This model allows simultaneous modelling of two or more stocks with a mixture of common and stock‐specific parameters: recruitment is always stock‐specific, but any other parameter can be specified as either common or stock‐specific. Common parameters are estimated from a wider base than they would be in a single‐stock model. The models time step is flexible and can be changed during the period being modelled to accommodate better data quality in recent data. Other options include the capacity to estimate movements among stocks, allow density‐dependent growth, and choose among likelihood functions for the various data sets, between finite and instantaneous fishing dynamics and between two forms of selectivity curve.

A voluntary reduction in the commercial catch of rock lobster (Jasus edwardsii) in a New Zealand fishery

Abstract We describe the development and application of a management procedure (decision rule) that resulted in a voluntary reduction in the commercial catch of spiny rock lobster (Jasus edwardsii) in the lower east coast of North Island of New Zealand. The management procedure was developed from an accepted assessment of the CRA 4 (Wellington‐Hawkes Bay) fishery, which used an integrated length‐based assessment model fitted to commercial fishery catch‐per‐unit‐effort (CPUE) biomass indices, commercial length‐frequency data, and tag‐recapture data. The assessment model had been Bayesian, and used the joint posterior distribution of parameters to predict the effect of 384 alternative harvest control rules on the future size of the CRA 4 stock. The harvest control rules all used CPUE as their input, and generated annual changes in catch, which were then simulated by the population dynamics of the operating model. Uncertainty was added to evaluations through observation error, added to the simulated CPUE observations, and stochastic serial auto‐correlation variation in recruitment. We describe how this management procedure was used to effect a voluntary reduction in catch to address the problem of a rapidly declining population.

Can additional abundance indices improve harvest control rules for New Zealand rock lobster (Jasus edwardsii) fisheries

Abstract Although New Zealand rock lobster (Jasus edwardsii) fisheries can be assessed with a sophisticated Bayesian length‐based model, these assessments are expensive and time consuming; they cannot be conducted for each area every year. Harvest control rules are increasingly important management tools in New Zealand rock lobster fisheries. Recent work has developed and evaluated procedures for rebuilding or maintaining lobster stocks based on criteria agreed by stakeholders. Most management procedures depend on a single abundance index, often catch per unit of effort (CPUE). When management procedures react slowly to changes in vulnerable biomass, allowable catches get out of phase with the stock, causing large oscillations in both catches and CPUE. Lags between data and management actions and “latent years” are features of rules that reduce responsiveness. This study explores ways to improve the responsiveness of harvest control rules by using additional data to predict changes in vulnerable biomass. Four data sets are examined: CPUE trends, pre‐recruit indices, puerulus settlement indices, and size frequencies. Only pre‐recruit indices, which were explored with a simple delay‐difference model based on parameter estimates from recent assessments, appeared to have immediate potential for use in improving management procedures.

A voluntary harvest control rule for a New Zealand rock lobster (Jasus edwardsii) stock

Abstract This paper describes a study to develop and test a simple harvest control (decision) rule that could be used by the commercial rock lobster industry in the CRA 5 quota management area in New Zealand. Time contraints prevented a full stock assessment, so alternatively the history of annual production was estimated from catch and catch‐per‐unit‐of‐effort (CPUE) data, and two versions of an operating model were based on the results: one used a fitted relation between biomass and production, and the other used constant production. Production was simulated with autocorrelated stochastic variation, and CPUE was simulated with observation error. Using a base case and two sensitivity trials for each model gave six operating models. Six candidate harvest control rules were developed, and sets of 1000 simulations were made for each combination of rule and model. The rules were then compared on the basis of fishery and stock indicators. Although choosing a specific rule is difficult, having a voluntary harvest control rule that reduces catches when CPUE is low maintains a higher mean biomass and prevents low biomass episodes.

Comment on sea lion population viability analysis

Chilvers (2011) used an age-structured simulation model to project future population trajectories of Hooker’s or New Zealand sea lions (Phocarctos hookeri) with and without incidental fisheries mortality and epizootic events. A core conclusion was that ‘‘sustained fisheries bycatch at current estimated levels... could result in a population decline and possible functional extinction over the modelled time period’’ (runs were made for 100 years). The modelling contained a fundamental error that invalidates this conclusion. Much of the breeding of these sea lions occurs around the Auckland Islands, and a fishery for arrow squid (Nototodarus sloanii) in adjacent waters results in incidental fisheries mortality when sea lions drown inside the trawl nets (Wilkinson et al. 2003). This problem has driven much of the population and analytical work on this species over the past two decades. A major input to the Chilvers simulation model was sexspecific schedules of annual mortality-at-age. These and their uncertainty estimates were attributed to Chilvers and MacKenzie (2010), who estimated survival rates by analysing re-sightings data from animals that had been tagged as pups at the Sandy Bay rookery. The tagged population was the subject to fisheries mortality: for instance, Roe (2006) reports Department of Conservation tags on four animals that had been caught by the squid fishery and returned for autopsy. Chilvers and MacKenzie (2010) did not address fisheries mortality estimates in their analysis. Consequently, their survival estimates are the survival from both natural causes and incidental fisheries mortality. Mortality rates derived from these survival rates are total mortality estimates: they include natural causes and incidental fisheries mortality. However, Chilvers (2011) treated these mortality estimates as natural mortality estimates. Thus, her model 1 was meant to represent the ‘‘baseline’’ scenario with neither fisheries mortality nor epizootics, and she added fisheries mortality in models 5–13. Because the survival estimates of Chilvers and MacKenzie (2010) already included the effects of fisheries mortality, this amounted to ‘‘doublecounting’’ this mortality: the fisheries mortality was removed twice in models 5–13. Under the assumptions of the Chilvers model, the actual effect of fisheries mortality is seen in model 1 (without epizootics) and in models 2–4 (with epizootics). None of these four models showed a probability of quasi-extinction greater than zero (see Chilvers’s Table 2). (As an aside, the double-counting problem does not exist for epizootic mortality on pups, because Chilvers excluded the ‘‘epizootic years’’ from pup mortality estimates (see her Table 1), but the problem does exist for adult mortality. If the epizootics affected adult survival, the mortality estimates used by Chilvers already included it, so models 4, 10, 12 and 13 double counted the adult mortality from epizootics.) This is a very basic and surprising error. The main conclusion of the paper, that continuing fisheries mortality threatens decline and possible extinction, is invalidated by This is a comment on Chilvers (2011), doi:10.1007/s00300-011-1143-6.