James N. Ianelli

AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models

Many criteria for statistical parameter estimation, such as maximum likelihood, are formulated as a nonlinear optimization problem. Automatic Differentiation Model Builder (ADMB) is a programming framework based on automatic differentiation, aimed at highly nonlinear models with a large number of parameters. The benefits of using AD are computational efficiency and high numerical accuracy, both crucial in many practical problems. We describe the basic components and the underlying philosophy of ADMB, with an emphasis on functionality found in no other statistical software. One example of such a feature is the generic implementation of Laplace approximation of high-dimensional integrals for use in latent variable models. We also review the literature in which ADMB has been used, and discuss future development of ADMB as an open source project. Overall, the main advantages of ADMB are flexibility, speed, precision, stability and built-in methods to quantify uncertainty.

Spatial Distribution, Energetic Status, and Food Habits of Eastern Bering Sea Age-0 Walleye Pollock

Abstract Age-0 walleye pollock Theragra chalcogramma were collected from the eastern Bering Sea during years when ocean temperatures were anomalously warm (2004-2005) and cool (2006-2007). Variability in spatial distribution, food habits, energetic condition, and recruitment to the age-1 life stage was investigated in relation to thermal regime. Age-0 walleye pollock were large in size and widely distributed during warm years, whereas they were small and largely confined to the middle domain of the eastern Bering Sea during cool years. Energy density was positively correlated with body weight and was greater during years when conditions were anomalously cool. The proportion of smaller age-0 walleye pollock in the diet of larger individuals was high when conditions were warm (21.9% of diet by weight), and euphausiids were the most important prey (36.5% of diet by weight) when conditions were cool. Eastern Bering Sea age-0 walleye pollock were abundant and broadly distributed from Bristol Bay to offshore an...

A STATISTICAL MODEL FOR STOCK ASSESSMENT OF SOUTHERN BLUEFIN TUNA WITH TEMPORAL CHANGES IN SELECTIVITY

Assessment of the status of southern bluefin tuna (SBT) by Australia and Japan has used a method (ADAPT) that imposes a number of structural restrictions, and is similar to methods used for a number of stocks world-wide. A flexible method for assessment of the SBT population is presented that is much less restrictive and has potentially wide applicability. The three key features are: (1) all fitting to data is within the context of maximum likelihood, (2) catch-at-age data are not assumed to be without error (as in existing methods), but rather to be random variables, while age-specific selectivity is allowed to change over time within the bounds of specific structure, and (3) autocorrelation in recruitment processes is considered within the likelihood framework of the model. While the results suggest the stock has been depleted considerably from its virgin biomass, and are generally consistent with previous assessments, they also indicate that it is not as much below the biomass that will produce maximum suitable yield as previously estimated and that the extent of stock rebuilding necessary may not be as large as has been argued. The available data are shown to provide little information on the steepness parameter of the stock-recruitment function, and hence on sustainable catch levels for the stock.

Simulation Analyses Testing the Robustness of Productivity Determinations from West Coast Pacific Ocean Perch Stock Assessment Data

Abstract The estimated stock size and recruitment data from assessment models provide a way to evaluate the history of a fishery relative to the expected response given a target fishing mortality rate. For example, if recruitment estimates are below the fishing mortality replacement line, then further reductions in spawning biomass can be expected. However, recruitment variability, serial correlation, and variable historical harvest rates can create misleading perceptions of sustainable harvest rates when they are derived from these simple replacement-line curves. Careful analysis of the stock–recruitment relationship given the available data is required in order to evaluate stock productivity in age-structured modeling. In this study, simulation–estimation procedures were developed to evaluate stock productivity estimates given the level of information and historical fishing mortality currently reported for the Pacific ocean perch Sebastes alutus stock off the coast of Washington, Oregon, and California....

Decision analysis of alternative harvest policies for the Gulf of Alaska Pacific Ocean Perch Fishery

By the mid-1970s the estimated biomass of Pacific ocean perch (POP), Sebastes alutus, in the Gulf of Alaska (GOA) had been reduced to about 10% of the estimated level present during the early 1960s. Although fishing mortality was greatly reduced, the stock has shown only modest increases in biomass; the current estimate of spawner biomass is between 15% and 20% of the estimated peak levels. This decline has raised concern that past management measures may have been inadequate to allow growth of the POP stock in the GOA. Consequently, we performed an analysis to: (a) identify optimal fishing rates for rockfish species such as Pacific ocean perch, (b) identify the biomass level that would produce an optimum yield, and (c) evaluate alternative fishing policies designed to improve the condition of the POP resource. The selection of an optimal fishing rate for POP in the GOA was based on a maximin criterion. Recruitment variability was shown to play an important role in determining the effects of different harvest policies. Based on estimates of optimal biomass and fishing mortality rates, four alternative harvest policies were developed and evaluated using a stochastic simulation model. These policies ranged from harvests based on rates under the status quo policy to increasingly restrictive measures. Policy outcomes were measured in terms of yield in weight and dollar-value, female spawner biomass, and risks. We presented the results in the form of Bayesian decision tables. The ability to predict future stock levels with a high degree of certainty is poor. In no case is there a guarantee that rebuilding will occur.

Spatial and Temporal Variability in Atka Mackerel Female Maturity at Length and Age

Abstract Spatial and temporal variability in maturity at length and maturity at age was examined for female Atka mackerel Pleurogrammus monopterygius in the eastern and central Aleutian Islands of Alaska. Postovulatory follicles (POFs) were evaluated as a potential character to distinguish spent mature fish from immature fish. We estimated that POFs remain in the ovaries for approximately 1 month after spawning, therefore allowing us to include some collections of recently spent fish for maturity estimates. Maturity-at-length data varied significantly among geographic areas and years, whereas maturity-at-age data failed to indicate differences. These results are probably attributable to different growth rates between the areas. An age-based model was developed to account for differences in year-class strengths, and modeling results showed that length at maturity could vary by as much as 4 cm over time because of variable recruitment. Maturity in Atka mackerel seems to be determined more by age than by somatic body size.

Reducing retrospective patterns in stock assessment and impacts on management performance

Reducing retrospective patterns in stock assessment and impacts on management performance Cody S. Szuwalski,* James N. Ianelli, and André E. Punt Bren School of Environmental Science and Management, Marine Science Institute, University of California, Santa Barbara, CA 93106, USA National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Alaska Fisheries Science Centre, 7600 Sand point Way NE, Seattle, WA 98115, USA School of Aquatic and Fishery Sciences, University of Washington, 1122 Boat Street, Seattle, WA 98195, USA

Integrating data from multiple surveys and accounting for spatio-temporal correlation to index the abundance of juvenile Pacific halibut in Alaska

Integrating data from multiple surveys and accounting for spatio-temporal correlation to index the abundance of juvenile Pacific halibut in Alaska Kotaro Ono*, James N. Ianelli, Carey R. McGilliard, and André E. Punt Centre for Ecological and Evolutionary Synthesis (CEES), Department of Biosciences, University of Oslo, P.O. Box 1066 Blindern. NO-0316 Oslo, Norway Centre for Coastal Research (CCR), University of Agder, 4604 Kristiansand, Norway School of Aquatic and Fishery Sciences, University of Washington, PO Box 355020, Seattle, WA 98195-5020, USA Alaska Fisheries Science Center, National Marine Fisheries Service, 7600 Sand Point Way NE, Seattle, WA 98115, USA *Corresponding author: tel: þ47 22 85 44 00; fax: þ47 22 85 40 01; e-mail: kotaro.ono@ibv.uio.no

Feasibility of Tagging Walleye Pollock Captured with Hook and Line using External Tags

Abstract We evaluated methods of minimizing mechanical injury to Walleye Pollock Gadus chalcogrammus when tagging them with external identification tags. Walleye Pollock (20–62 cm FL) were captured with hook and line near Auke Bay, Southeast Alaska, and were tagged with either T-bar anchor tags or lock-on tags, which were anticipated to be used for tagging studies in the Gulf of Alaska and Bering Sea. The tested handling procedures included transferring the tagged fish between live tanks either by using a dip net (dipnetted group) or with wet, bare hands (non-dipnetted group). Sixty percent of the dipnetted fish (63 of 105) died, whereas 12% of the non-dipnetted fish (17 of 138) died. Overall, 50% of the deaths occurred within 7 d after capture, and 89% of the deaths occurred within 10 d after capture. Of the dipnetted fish that died, 68% (n = 43) died from dermal infection due to scale loss, whereas 30% of the non-dipnetted fish mortalities were from dermal infection. Additionally, injuries that were recorded as potential mortality factors included fin loss, torn jaws, internal dysfunction, and unknown. All of the fish that received lock-on tags and 93% of the fish that were anchor-tagged developed an infection at the point of tag insertion. Tag retention rates were 99.5% for lock-on tags and 93.7% for anchor tags, and tag type did not affect survival. Use of hook-and-line capture is an effective method for reducing mechanical injury and mortality in Walleye Pollock. Recommended procedures for capturing and tagging Walleye Pollock include the use of hook and line and the use of wet, bare hands (or a similar low-abrasion approach) when handling the fish.