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

The 2017 STAR report noted four unresolved problems and three major sources of uncertainty as well as provided three specific recommendations for future research with respect to r spp [@haltuch2019lingcod]. In addition, the Panel made four recommendations for stock assessments in general.

All four unresolved problems were addressed during this assessment.

1. The models did not use the available age data sampled from the fishing fleets due to concerns that unsexed fish had been assigned equally to the sexes without regard to length and because of evidence there had been non-random subsampling of fish for age-reading.

Age data included in the northern model were conditioned on length for sexed r spp. Unsexed age samples were not included in either model. Length distributions of unsexed fish were represented as independent vectors in addition to the vectors of female and male compositions.

1. The available age-readings had been done by at least two laboratories. It was unclear that age-reading protocols had been employed consistently.

Analyses on ageing error did not show differences between readers, labs, or cohorts (see Section \@ref(ageing-precision-and-bias)). The ages available for the north model were fit reasonably well while those available in the south from sources other than \gls{s-wcgbt} showed a lack of fit that was unlikely attributable to ageing uncertainty alone.

1. In the northern model the STAT fixed the parameter for female length at age 14 years because when this parameter was freely estimated the model estimated asymptotic selection for the trawl fishery and greatly altered the estimates of spawning biomass. It was unclear what data sources were responsible for this result.

The addition of more years of age data from the \gls{s-wcgbt} along with the inclusion of ages from the fisheries resulted in a total of 72,121 age samples represented in the north model which was more than adequate to provide reasonable estimates of growth.

1. Sensitivity analyses for draft versions of both models indicated they were sensitive to underlying structural assumptions such as the starting year for recruitment deviations and which indices were included. Although the revisions to the models developed during the STAR may have lessened the sensitivity of the models by removing sources of tension and keeping the more reliable data, there was not sufficient time during the review to explore the sensitivity of the final base models.

A number of sensitivities were explored that excluded data sources and attempted to reconcile data conflicts. Additional explorations made during the STAR panel review led to changing the timing of the main recruitment period in the southern model given removal of the early length data that had small sample sizes and were only representative of a small portion of the coast. Sensitivity to underlying structural assumptions is a general issue of stock assessment modeling and not unique to r spp.

All three major uncertainties were explored during this assessment.

1. Stock structure: Aspects of the length- and age-compositions evident in the NWFSC survey data strongly indicate spatial patterns that probably cannot be well mimicked with separate, independent models for the north and the south.

Research on r spp stock structure that was completed since the last stock assessment provided evidence of a stock break in central California [@longo2020strong]. Details were described in the Stock delineation section. Although this may not resolve evidence of wide scale spatial patters, major Uncertainty 1 can now be rexamined under additional information on stock structure. The data from the \gls{s-wcgbt} was, in general, well fit by both models.

1. Key productivity parameters: Neither the northern model nor the southern model were able to estimate the steepness or the female natural mortality parameters given the available data. Values for these key parameters had to be fixed but there is very little knowledge to inform the choice of those values. As such this is a source of considerable uncertainty. During review of this report the STAT suggested that including the age-composition data in the northern base model (data had been removed during the STAR) would allow estimation of M and h.

Stock-recruit steepness and natural mortality (for both females and males) were estimated in both north and south models and provided plausible estimates. The representation of the uncertainty in these estimates provides a better treatment of the uncertainty in quantities of interest, such as the management reference points.

1. Habitat area, north versus south: The northern and southern base models estimate appreciable differences in the unfished spawning biomass of r spp (37,974 mt in the north versus 20,462 mt in the south). It is unknown whether such a difference is consistent with the habitat areas suitable to support r spp in the north versus the south.

Using simple assumptions on available habitat, by assuming the relative difference in the amount of area in depths between 55 m and 300 m, roughly approximates the relative difference in habitat area between modeled regions.

Next, we provide responses to each of the three specific recommendations from the 2017 STAR panel.

1. There should be a study to cross-validate age-readings of r spp among the different laboratories contributing age data to the assessment. It may be necessary to develop laboratory-specific (and possibly year-specific) ageing-error vectors.

Additional ageing comparisons were conducted and the ageing uncertainty was re-estimated as described in the Ageing Error section of this report. However, the new estimates showed similar uncertainty in ageing to the previous assessment. The differences among age distributions within the same length bins among sources available for the southern model are large enough that they are unlikely due to uncertainty in ageing alone. Removing all ages in the southern model other than those from the \gls{s-wcgbt} resolved this issue.

1. Available information on r spp catches, abundance trends, and age-compositions should be acquired from Canadian and Mexican authorities to take an initial step towards a more spatially-comprehensive view of r spp population trends and dynamics.

Additional information on r spp in Mexico, Canada, and Alaska is provided in the Foreign Fisheries section of this report.

1. The next iteration of this assessment could be an update assessment. If a full assessment is done it should explore developing a spatial model that encompasses the northern and southern areas rather than again treating them as independent stocks, as in the current and previous assessments.

New information on genetic differences between the modeled areas supports the treatment of these areas as independent models (see Section \@ref(stock-delineation)). A spatially-explicit model encompassing data from both stocks was not investigated for this assessment because the widespread use of such approaches are not yet standard and remain operational for only a few select data-rich stocks, largely in a simulation context. Limitations still exist because of the model complexity associated with increasing spatiotemporal dimensions, including underlying data limitations, confronting the expanding number of decisions or assumptions that need to be made, and the amount of analyst time required to develop, test, and vet spatial procedures given production stock assessment timelines [@10.1093/icesjms/fsaa203; @10.1016/j.fishres.2019.01.014]. Scientifically, there is growing appreciation and application of spatial stock assessment methods using simulations and the results of this research underlines the importance of acknowledging spatial processes (e.g., connectivity dynamics between unique segments of a stock; biological characteristics that change across environmental gradients; and regulations that impose local changes in fishing patterns) across the management domain [@10.1016/j.fishres.2020.105608; @10.1093/icesjms/fsaa203]. But, a production-level spatially-explicit stock assessment framework was not available for this assessment.

The four general recommendations on stock assessments were as follows:

1. Modify the software used to develop length- and age-compositions from PacFIN data so that unsexed fish are flagged rather than including them in compositions after the automatic application of an assumed sex-ratio (e.g., 50:50). If the analysts preparing the composition data need to develop sex-ratio coefficients to accommodate unsexed fish (e.g., by length-bin), the assessment documents should clearly state the methods and data used for this purpose and the resulting sex-ratio coefficients.

2. If assessments use marginal age-compositions the STATs should evaluate whether the raw data are consistent with random sub-sampling from the available lengths. If the ages appear to have been subsampled non-randomly (e.g., no more than 5 fish from any length-bin), the age data should be suitably expanded to reflect the variable sampling fraction.

3. A standard approach for combining conditional age-at-length sample data into annual CAAL compositions should be developed and reviewed. If age data are not selected in proportion to the available lengths, simple aggregation of the ages by length-bin may provide biased views of the overall age-composition and year-class strength.

4. Comprehensively evaluate whether the Triennial survey should be split into early and late segments and the basis for making the decision. The r spp assessment split the Triennial survey into separate early and late surveys, whereas there was a single Triennial survey in the draft assessment for Pacific ocean perch brought to this STAR.

Ongoing work by the NWFSC staff has addressed recommendations 1, 2, 3, and explorations by the r spp STAT can inform recommendation 4, though not comprehensively. The r spp STAT chose to represent ages as conditioned on length rather than apply an expansion to reverse the effect of non-random sampling.

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