In iantaylor-NOAA/Lingcod_2021: Estimate the Status of Lingcod off the US West Coast in 2021

Model

Previous assessments and reviews

History of modeling approaches

Since 1986, there have been nine assessments of r spp covering at least part of the U.S. west coast. The first assessment was performed by Adams [-@pfmclingcod1986] and consisted of a yield per recruit analysis. Each of the subsequent assessments utilized an age-structured model, though the structure and form has changed with time.

The second assessment, conducted by Jagielo [-@jagielo1994assessment], was conducted using an earlier version of Stock Synthesis [@methot_stock_2013] and focused on the northern area. Data were limited to the region between Cape Falcon in Northern Oregon and 49\textdegree 00\textquotesingle N latitude (off of southwest Vancouver Island in British Columbia). This included \Gls{psmfc} areas 3A, 3B, and 3C, inclusive of Canada. Two fleets were modeled, trawl and recreational, from 1979-1993, two \gls{cpue} time series were included, \gls{s-tri} and trawl, and length- and age-composition data were used to estimate selectivity. The final spawning output was estimated to be approximately 20\% of pristine levels, and catch recommendations ranged between 2500 and 3000 mt based on fishing mortality levels of 20\% and 40\%.

The third assessment was conducted in 1997 [@jagielo1997assessment] and expanded the area included south to Cape Blanco (42\textdegree 50\textquotesingle N latitude). The estimated fraction unfished was below 0.10, which led to the Council implementing substantial coastwide reductions in harvest and recommending that the southern portion of the stock also be assessed. Two years later, Adams et al. [-@pfmclingcod1999] conducted a length-based, age-structured population model implemented in \gls{admb} [@fournier1996] for the southern area, including Eureka, Monterey, and Conception \gls{inpfc} areas. Since then, r spp assessments have covered the entire U.S. west coast but been broken into two assessments using various spatial delineations.

In 2000, Jagielo et al. [-@jagielo2000assessment] fit data to \gls{admb} models using the same spatial delineation but removing data from Canadian waters. Both stocks showed increasing trends in their abundance and new ageing methods suggested the stocks were younger and more productive. The populations were assessed again in 2003, this time using Coleraine [@pfmclingcod2003]. In 2005, Jagielo et al. [-@pfmclingcod2005] switched back to Stock Synthesis, which was then at version 2 of the code base. Estimates suggested that the northern stock had recovered substantially from a low point in the 1990s and was at 0.87 fraction unfished, while the southern area was not estimated to have recovered and was at 0.24 fraction unfished, with a 0.64 coastwide fraction unfished.

The 2009 stock assessment, which used Stock Synthesis 2, divided the northern (Washington and Oregon) and southern (California) stocks at the Oregon-California state line [@hamel2009lingcod]. The northern and southern models were made as equivalent as possible by keeping fixed and estimated parameters largely the same for the two assessments. Natural mortality was fixed at 0.18 for females and 0.32 for males in both assessments, while steepness was fixed at 0.8. Age data were removed because of issues with outliers and possible aging bias. The estimate of fraction unfished at the start of 2009 was 0.619 for the northern area and 0.737 for the southern area, indicating the stocks had recovered. For the northern stock, the axis of uncertainty for the decision table provided to managers was $M$. The low, base, and high values of $M_{female}$ were 0.16, 0.18, and 0.20. The low, base, and high values of $M_{male}$ were 0.285, 0.320, and 0.355. For the southern stock, the lower axis of uncertainty was for the he exclusion of the dockside recreational \gls{cpue} index and the higher axis of uncertainty was the inclusion of age data.

The 2017 assessment [@haltuch2019lingcod] was conducted using Stock Synthesis and matched the spatial structure used for 2009 [@hamel2009lingcod]. The assessment updated the existing data sources, as per normal protocol, and also included the following changes:

• expansion of the time period of the assessments back to 1889,
• separation of commercial fleets into TW and FG,
• separation of recreational fleets in the northern stock to state-specific fleets,
• addition of numerous fishery-dependent \gls{cpue} indices and the \gls{s-hkl} index
• updated prior on natural mortality
• updated biological parameters for maturity and length-weight relationship
• updated ageing uncertainty
• included some age data as marginal ages instead of ages conditioned on length

The estimated scale of the population was sensitive to the inclusion of age data for both stocks. These estimates of status remained the same regardless of whether or not the model was fit to age data. The northern stock was estimated to be above the target reference point of 0.40 fraction unfished. The southern stock was estimated to be below the target reference point of 0.40 fraction unfished at 0.321.

In 2019, a catch-only projection was conducted. This update assessment included observed catches for 2017 and 2018 in the forecast calculations. Expected catches for 2019 and 2020 were also included in the forecast calculations. The results provided updated catch limits for management.

Most recent STAR panel and SSC recommendations{#STAR}

sa4ss::read_child(file.path("STARpanelRecs.Rmd"))

Response to Groundfish Subcommittee recommendations

Model structure, assumptions, and results

sa4ss::read_child("model-changes.Rmd")

General model specifications

sa4ss::read_child("model-specifications.Rmd")

sa4ss::read_child("model-selectivity.Rmd")

Model parameters {#sec-model-parameters}

sa4ss::read_child("model-parameters.Rmd")

Base model selection {#sec-model-selection}

sa4ss::read_child("model-selection.Rmd")

Base model results

sa4ss::read_child("model-results.Rmd")

Model diagnostics

Convergence

sa4ss::read_child(file.path(paste0(params$model, "_jitter_0.05"), "model-results-jitter.md"))

Sensitivity analyses

sa4ss::read_child("model-sensitivities.Rmd")

Likelihood profiles

sa4ss::read_child("model-profile.Rmd")

Retrospective analysis

A five-year retrospective analysis was conducted by successively removing years of data starting with the most recent year. For the removal of the most recent year of data there were no changes to the model trajectory or estimates of fraction unfished (Table \ref{tab:RetroMohnsrho}; Figures \@ref(fig:RetroSsb) and \@ref(fig:RetroFractionunfished)). ```{asis, opts.label = "north"} Peeling off more than one year of data led to a retrospective pattern with respect to the estimated scale of the population but not current status.

```{asis, opts.label = "south"}
Removing five years of data led to a different historical time series for the stock,
one that indicated much higher productivity.
Regardless, the estimate of the current spawning stock biomass remained largely
consistent with the estimates from the base model.

Given time constraints, each of these models were not tuned, i.e., data weighting was not performed in an iterative fashion and bias adjustment was not conducted, which could impact the model results.

Evaluation of uncertainty

sa4ss::read_child("scientific-uncertainty.Rmd")

Reference points

``{asis, opts.label = "north"} The estimate of 2021 spawning biomass relative to unfished equilibrium biomass (fraction unfished =r Bratio_2021`) was well above the management target of 40\% (Table \@ref(tab:table-dq-base) and Figure \@ref(fig:SPR4-phase)). Even the lower 95\% interval was estimated to be above the target (Table \@ref(tab:table-dq-base)). Only a few years since the start of the modeled period were estimated to be below the target (blue points to the left of the vertical line at 0.4 in Figure \@ref(fig:SPR4-phase)).

Equilibrium yield given the estimate of steepness shows disparity between maximum sustainable yield (blue) and the target biomass (red; Figure \@ref(fig:yield2-yield-curve-with-refpoints)). Whereas, the target biomass is in line with the SPR target (green; Figure \@ref(fig:yield2-yield-curve-with-refpoints)). The time series of fishing intensity was estimated to have been below the management harvest rate limit for all years included in the model (Figure \ref{fig:SPR2-minusSPRseries}).

```{asis, opts.label = "south"}
The estimate of 2021 spawning biomass relative to unfished equilibrium biomass
(fraction unfished = `r Bratio_2021`)
was estimated to be close to the management target of 40\%
(Table \@ref(tab:table-dq-base); Figure \@ref(fig:SPR4-phase)).
The uncertainty in this estimate spans above and well below the target,
suggesting the current status of the stock is uncertain
(Figure \@ref(fig:SPR4-phase)).

Equilibrium yield given the estimate of steepness shows disparity between
maximum sustainable yield (blue) and the SPR target (green;
Figure \@ref(fig:yield2-yield-curve-with-refpoints)).
Whereas, the target biomass (red) is in line with the maximum sustainable yield and the current biomass
(Figure \@ref(fig:yield2-yield-curve-with-refpoints)).
The time series of fishing intensity was estimated to have been below the
management harvest rate limit for the early years of the fishery and the
most recent years included in the model, i.e.,
since the early 2000s
(Figure \ref{fig:SPR2-minusSPRseries}).

Decision table

The forecast of stock abundance and yield was developed using the base model (Table \@ref(tab:table-projections-base)). The total catches for the first two years of the forecast period were based on values provided by the \glsentrylong{gmt}. These assumed removals are likely higher than what the true removals will be for this year and next year but their influence on the assessment of stock status and future removals are limited. ```{asis, opts.label = "south"} The projections, including the first two years, assume a 40:60 split between the trawl and fixed-gear commercial fleets based on guidance from the \glsentrylong{gmt}.

```{asis, opts.label = "north"}
The states of nature in the decision table (Table \@ref(tab:decision)) 
do not fall on a single axis of uncertainty. Instead, two alternative model 
structures were chosen from the set of sensitivity analyses to represent
alternative states of nature. This represents the consensus among the 
participants in the STAR panel as a better representation of uncertainty
than any axis could. The low state of nature included sex-specific selectivity
while the high state excluded the fishery-dependent age data.

```{asis, opts.label = "south"} The axes of uncertainty in the decision table (Table \@ref(tab:decision)) are based on uncertainty in female natural mortality.

Three alternative catch streams were created for the decision table (Table \@ref(tab:decision)).
The first option uses recent average catch as provided by the \glsentrylong{gmt},
the second option uses a $P^*$ of 0.40, and
the third option uses a $P^*$ of 0.45.
These $P^*$ values are combined with the category 2 default $\sigma$ = 1.0
in calculating the buffer between \gls{ofl} and \gls{abc}.

### Unresolved problems and major uncertainties

```r
sa4ss::read_child("unresolved-problems.Rmd")

Research and data needs

sa4ss::read_child("research-and-data-needs.Rmd")

Comparison of north and south models

sa4ss::read_child("model-north-vs-south.Rmd")

iantaylor-NOAA/Lingcod_2021 documentation built on Oct. 30, 2024, 6:42 p.m.