Dave Fournier¹, James Ianelli^1,2 and Steve Martell^1,3
1ADMB Foundation; ²Alaska Fisheries Science Center; ³International Pacific Halibut Commission

April 21, 2015 9:00 (PST): FSH 203

A proposal for rehabilitating the multinomial likelihood for compositional data

The multinomial likelihood is widely used in many stock assessment models for evaluating compositional data. However, this simple model does not account for over-dispersion and does not allow for correlation among adjacent categories. In response to recent publications proposing alternatives for the multinomial, we propose to extend and modify the ideas put forward by Hranfnkelsson and Stefánsson (2004) to better accommodate over-dispersion and correlation in compositional data. Specifically, to develop a self-scaling multinomial type estimation procedure and optionally incorporate positively autocorrelated errors. Models that incorporate composition data need to take into account over-dispersion and correlation if uncertainty in model parameters, and the risks associated with management actions, is desired.

Dr. Melissa Haltuch¹, Z. Teresa A’mar², Nicholas A. Bond³, and Juan L. Valero^4
1Northwest Fisheries Science Center, ²Alaska Fisheries Science Center, ³University of Washington, and ⁴Center for the Advancement of Population Assessment Methodology

March 10, 2015 9:00 (PST): FSH 213

Assessing the effects of climate change on U.S. west coast sablefish productivity and on the performance of alternative management strategies.

The U.S. west coast sablefish fishery is a valuable commercially targeted species, making assessing and understanding the interaction between climate change and fishing a priority for (1) forecasting future stock productivity and (2) for testing the robustness management strategies to climate variability and change. The horizontal-advection bottom-up forcing paradigm describes large-scale climate forcing that drives regional changes in alongshore and cross-shelf ocean transport, directly impacting the transport of nutrients, mass, and organisms. This concept provides a mechanistic framework through which climate variability and change alter sea surface height (SSH), zooplankton community structure, and sablefish recruitment, all of which are regionally correlated. This study assesses future trends in sablefish productivity as well as the robustness of harvest control rules to climate driven changes in recruitment by conducting a management strategy evaluation of the currently implemented harvest control rule as well as an alternative. We use 50 year ensemble forecasts of sablefish productivity under a suite of future climate variability and change scenarios. Future recruitments are generated under two scenarios (1) the fit of a Beverton-Holt stock-recruitment curve based on historical data and (2) recruitments driven by a SSH-recruitment relationship that is treated as an age-0 survey of abundance with associated uncertainty. Multi-decadal forecasts of sablefish productivity could provide long term strategic advice to allow fishers and managers to plan for and respond to shifts in productivity.

Hiromichi Ueno, Ph.D.^1
1Faculty of Fisheries Science, Hokkaido University
March 04, 2015 11:00 (PST)

Temporal and spatial variation in growth factors of Pacific salmon.

Dr. Ueno is an Associate Professor at Hokkaido University in Hokkaido, Japan. He has a PhD (2003) and Masters (2000) in Oceanography from the University of Tokyo, where he also later performed postdoctoral research. His research focuses on four main topics: 1) mesoscale eddies, physical characteristics and its impact on marine ecosystem; 2) pacific salmon and the physical environment affecting their growth; 3) temperature inversions and haloclines in the subarctic North Pacific; and 4) the physical environment and its impact on marine ecosystems in the Chukchi Sea.

This talk will start with the introduction of international intensive course for sustainable fishery resource use at Hokkaido University toward building a joint program in the North Pacific on board T/V Oshoro-maru. I will also talk about temporal and spatial variation in growth factors of Pacific salmon from the view point of bioenergetics. The prey-density function for consumption is used as the growth factor. For the estimation of the prey density, we used zooplankton density obtained from an ecosystem model, NEMURO embedded in a 3-D physical model. In this study we focus on the three species of Pacific salmon (chum, pink and sockeye), zooplankton feeders whose biomass exceeds 90% of Pacific salmon. Empirical Orthogonal Function (EOF) analysis is conducted for the growth factor in the subarctic North Pacific during 1948–2007. The growth factor of Pacific salmon varies in a decadal timescale. The variation of the growth factor in the Bering Sea, the western subarctic gyre and the eastern subarctic North Pacific are consistent with the variation of total catch of chum, pink and sockeye salmon.

Ole Shelton^1
1Northwest Fisheries Science Center
February 24, 2015 09:00 (PST)

Inferring marine communities from environmental DNA: merging traditional and new molecular techniques in the sampling of the oceans.

Environmental DNA (eDNA)—genetic material recovered from an environmental medium such as soil, water, or feces—reflects the membership of the ecological community present in the sampled environment. As such, eDNA is a potentially vast source of data for basic ecology, conservation, and management, because it offers the prospect of quantitatively sampling ecological communities from easily-obtained and non-lethal samples. However, like all sampling methods, eDNA sequencing is subject to methodological limitations that bias the resulting estimates of abundance and community membership. I will describe a framework for thinking about current methodological challenges for eDNA, discuss how these challenges are largely analogous to biases in more traditional survey methods, and provide detailed discussion of methods for an important step in the eDNA process – estimating relative abundance of DNA from marine community.

André Punt¹, Ian Taylor², and Owen Hamel^2
1School of Aquatic and Fishery Science
²Northwest Fisheries Science Center
February 11, 2015 14:00 (PST)

Data reweighting (and the Francis Method)

Weighting of data is fundamental to modern stock assessment science based on ‘integrated’ methods of assessment. There is considerable experience with weighting index data, and the methods for doing this have largely stabilized. This is not however the case for composition data even though it is clear that inappropriate weighting of composition data can lead to markedly biased estimates. This Think Tank focuses on two methods for calculating the effective sample size for composition data (McAllister-Ianelli and Francis) as well as how estimates of effective sample size should be combined over years. A key focus for the Think Tank and future work is how to apply the “Francis method” to conditional age-at-length data. With luck, approaches for addressing this question will arise during the Think Tank.

Chris Legault¹ and Michael Palmer^1
1Northeast Fisheries Science Center
January 27, 2015

What direction should the fishing mortality target change when natural mortality increases within an assessment?

Traditionally, the natural mortality rate (M) in a stock assessment has been assumed to be constant over years and ages. When M increases within an assessment, as has occurred in a number of Canadian cod stocks, the United States Gulf of Maine cod and Atlantic herring stocks, the question arises how to change the fishing mortality rate target (Ftarget). Yield per recruit considerations lead to an increase in the Ftarget, while maximum sustainable yield considerations often lead to a decrease in the Ftarget. Both approaches are examined using results from recent assessments of Gulf of Maine cod and Georges Bank yellowtail flounder. Problems are found with both the yield per recruit and maximum sustainable yield approaches, leading us to recommend either not allowing M to change within an assessment model or if strong empirical evidence supports a changing M to base the Ftarget on the natural mortality rate considered most appropriate based on the life history traits of the species of interest.

Chantel Wetzel^1
1School of Aquatic and Fishery Science
January 13, 2015

Are we there yet? The impact of reduced composition data on the ability to monitor rebuilding for overfished stocks.

An overfished declaration leads to changes in harvest limits, often dramatic, for the overfished stock. A reduction in catch while the stock is overfished, can lead to a reduction in length and age samples from the fishery. This can be especially problematic when the stock is rarely encountered by fishery independent surveys, with the primary source of composition data coming from the fishery. On the U.S. west coast, several overfished rockfish stocks fall into this category (e.g. yelloweye, canary). The dramatic reductions in catch will inevitably lead to rebuilding of the stock; however, the reduction in available composition data may lead to increased uncertainty about the estimated biomass status. A Management Strategy Evaluation was performed to address the long-term impact of reduced data on the ability to monitor a stock during rebuilding. This work simulates an overfished flatfish and rockfish stock where harvest and the collection of new data are restricted to address two questions; 1) does the increased uncertainty due to limited data impact the ability to correctly detect when an overfished stock is rebuilt, and 2) is there a degradation of ability to detect status correctly with limited data and if so how does this change as the amount of data increases?

Curry Cunningham^1
1School of Aquatic and Fishery Science
December 09, 2014

Methods for Forecasting of Bristol Bay, Alaska Sockeye Salmon Abundance

Bristol Bay, Alaska is home to the largest wild sockeye salmon (Oncorhynchus nerka) fishery in the world, with a total harvest value of $165 million USD in 2010 and supporting an estimated equivalent to 10,000 year-around American jobs. However, returns of sockeye to Bristol Bay are highly variable, with observed returns ranging between less than 10 million and greater than 65 million since 1960. The profitability of the commercial fishery and the ability of the commercial fishing fleet to allocate effort amongst the 5 terminal fishing districts in Bristol Bay depends to some extent on accurate predictions for the number of returning sockeye prior to, and during, the summer fishing season. The Alaska Salmon Program at UW provides inseason forecasts for Bristol Bay sockeye salmon abundance, which are updated on a semi-daily basis throughout the duration of the commercial fishery as new data become available. These inseason forecasts incorporate a wide range of data sources: 1) catch, 2) escapement, 3) age composition, 3) genetic composition of catch, and 4) inseason test fishery catch indices, to inform estimates of run size for Bristol Bay as a whole and for the major fishing districts. I will describe our current forecasting methods for inseason assessment of Bristol Bay sockeye salmon abundance, a retrospective analysis of the efficacy of these forecasting methods, and elicit new ideas for improving these methods in the future.

Dr. André E. Punt¹ and Dr. Jason Cope^2
1School of Aquatic and Fishery Science
²Northwest Fisheries Science Center
November 18, 2014

Simple Stock Synthesis (SSS) and Depletion-Based Stock Reduction Analysis (DB-SRA) are two methods for calculating overfishing limits for data-poor fish stocks. Both methods are Bayesian in that priors are imposed on the parameters of the underlying population dynamics model (although there is no information to update the priors). DSB-SRA is based on a generalized production function which allows independent priors to be imposed on BMSY/B0 and FMSY/B0. However, SSS is based on the Beverton-Holt stock-recruitment relationship so there is a functional relationship between BMSY/B0 and FMSY/B0 as expressed in the steepness parameter (h). We outline two stock-recruitment relationships (the Shepherd and Generalized Ricker) which if incorporated in Stock Synthesis may allow the goal of independent priors to be imposed on BMSY/B0 and FMSY/B0 to be achieved. We show for some example species how well these stock-recruitment relationships achieve this goal and compare DB-SRA and SSS when they are applied to some example rockfish stocks.

Owen Hamel¹, Melissa Haltuch¹, and Jason Cope^1
1Northwest Fisheries Science Center
October 28, 2014

Because growth patterns of many west coast groundfishes limit the informational value of length data, fish ages are very important inputs to our age-structured stock assessments. At the Northwest Fisheries Science Center, we conduct a variety of ageing studies, and we review three recent avenues of research: age validation, alternative ageing methods, and understanding ageing error. Age validation: We developed the first bomb radiocarbon reference chronology for the California Current, using known-age petrale sole (Eopsetta jordani). Petrale sole spend a substantial portion of their first year of life in areas subject to variable upwelling. This variable environment illustrates the importance of using reference curves for age validation that are region- and species-specific, whenever possible. Alternative ageing methods: Traditional age-reading methods are time-consuming for long-lived species. Based on initial success with two groundfishes, we explore the use of otolith weights for rapid age determination of long-lived groundfishes. We also explore the consistency of these relationships over time and space. Ageing error: The Pacific hake (Merluccius productus) stock is characterized by infrequent, strong year-classes, surrounded by average and below-average cohorts. Ageing is conducted annually, such that readers routinely know the year of collection. Ageing error is typically assumed to be largely consistent across years in stock assessments, however, we hypothesized that readers are more likely to assign uncertain hake reads to predominant ages. We conducted a double-blind study wherein previously read otoliths from many years were reread without readers knowing the collection year. Results confirmed that strong year classes experienced less effective ageing error in the regular course of ageing otoliths. Accounting for this tendency improved model fits to age data. Each of these research avenues has improved our understanding and is enabling us to develop more reliable population models and management guidance.