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

\Glsentrylong{s-wcgbt} lengths

The length compositions of survey catches in each year was summarized using length bins in 2 cm increments from 10 to 130 cm (Figure \@ref(fig:wcgbts-lenComp)). The first bin includes all observations less than 10 cm, and the last bin includes all fish larger than 130 cm. The observed length compositions were expanded to account for subsampling of tows, and the expansion was stratified by depth. Depth strata of 55 - 183 m and 183 - 400 m were selected, based on the sampling design of the survey (@keller2017). Depth strata were capped at 400 m because catches of r spp in the \Gls{s-wcgbt} occur infrequently beyond 400 m (Figure \ref{fig:wcgbts-presAbs}). Samples were often sexed, so only male and female length frequencies were used. The few unsexed individuals were assigned as male or female according to the sex ratio of the respective length bin. An assumed sex ratio of 0.5 was applied for unsexed fish in length bin less than 40 cm, as sex of smaller sized r spp is harder to differentiate. A bin of 40 cm was chosen as this is the length bin at which the length-weight relationship starts to diverge for males and females, and therefore, equal assignment is not influenced by sex-specific size differences.

The input sample sizes (Table \@ref(tab:sample-size-length)) for length and marginal age-composition data for all fishery-independent surveys were calculated according to Stewart and Hamel [-@stewart_bootstrapping_2014], which determined that the approximate realized sample size for species in the "others" category (which included r spp) was $2.38*N_{\text{tow}}$.

\acrlong{s-tri} lengths

Post-processing of length data for the \gls{s-tri} (Table \ref{tab:sample-size-length}; Figure \ref{fig:tri-lenComp}) followed the same methods as those used for the \gls{s-wcgbt} but depth strata were bracketed using depths of 55--183 m and 183--350 m. Strata were split at 183 m because sampling became less intense in depths deeper than 183 m and raw \gls{cpue} was more variable deeper than 183 m than it was in shallower waters (Figure \ref{fig:tri-depthSplit}). A maximum depth of 350 m was used because r spp were infrequently caught at depths greater than 350 m (Figure \ref{fig:tri-depthSplit}).

```{asis, opts.label = 'south', echo = TRUE}

\Glsentrylong{s-hkl} lengths

The length composition of survey catches in each year was summarized using length bins in 2 cm increments from 10 to 130 cm (Figures \ref{fig:hkl-lenComp-sexed} and \ref{fig:hkl-lenComp-unsexed}). The first bin includes all observations less than 10 cm, and the last bin includes all fish larger than 130 cm. Length compositions from this survey were used as numbers of fish, all fish were measured, and were not expanded (Table \ref{tab:sample-size-length}). As such, composition data are available for male, female, and unsexed r spp.

#### Lam research lengths

In collaboration,
the \gls{nwfsc} and Moss Landing Marine Labs
sampled `r spp` in nearshore and offshore rocky reef habitats between
January 2016 and January 2017 via hook and line on chartered \glspl{cpfv}
(@lam2019geographic).
Sixteen latitudinal distinct sampling sites, or ports,
were chosen for sampling from northern Washington to southern California.
At each port,
85--120 individuals were caught using methods identical to those used by the
onboard recreational `r spp` fishery,
except that individuals smaller than the legal-size limit of 22 inches were retained and
areas closed to recreational harvest were occasionally utilized
(CDFW Permit #SC-6477, ODFW Permit #20237, WDFW Permit ID Samhouri 16-138). 
Deviating from the onboard-sampling methods was necessary to
ensure an even distribution of size and age classes from each port,
such that estimates of von Bertalanffy growth curves could be region specific and
easily comparable.
Of the total fish samples
(n = 1,784, 922 Males, 862 Females; Table \ref{tab:sample-size-length}),
32 were removed because they
were sampled on \gls{s-ccfrp} and would have led to double counting.
Four additional samples were also excluded because they had no year associated with them.
Length compositions from this survey were used as numbers of fish
and were not expanded (Figure \ref{fig:lam-lenComp}).
Length measurements were taken using total length,
which were converted to fork length following conversions from
Laidig et al. [-@laidig_lingcod_conversions_1997].
Total and fork lengths for `r spp` are generally similar for `r spp` given their tail shape.

#### \Glsentrylong{s-wcgbt} ages

Age-composition data from the \Gls{s-wcgbt} (Figure \ref{fig:wcgbts-ageComp}) were 
included in the model as sex-specific \gls{caal} distributions by year
(Figures \ref{fig:comp-condAALdat-bubflt7mkt0-page1}-\ref{fig:comp-condAALdat-bubflt7mkt0-page`r ifelse(params$area == "North", 2, 3)`}).
These data were not expanded, which is standard for \gls{caal} data, and thus,
numbers of fish were used for the input sample size without any adjustment.
Each age bin included just one year and summarized ages ranging from 0--20
with a plus group for all fished aged older than 20 years.

Individual length and age observations can be thought
of as entries in an age-length key (matrix).
Age-length keys are typically structured with
age across the columns and length down the rows.
The \gls{caal} approach consists of tabulating the sums within rows as the
standard length-composition distribution and, instead of also tabulating the sums to 
the age margin, the distribution of ages in each row of the age-length key is 
treated as a separate observation, conditioned on the row (length) from which it came. 
Using \gls{caal} instead of marginal ages has several benefits.
First, age structures are generally collected as a subset of the fish 
that have been measured. If the ages are to be used to create an external age-length 
key to transform the lengths to ages, then the uncertainty due to sampling and missing 
data in the key are not included in the resulting age-compositions used in the stock 
assessment. If the marginal age compositions are used with the length compositions 
in the assessment, the information content on sex-ratio and year class strength is 
largely double-counted as the same fish are contributing to likelihood components 
that are assumed to be independent. Using \gls{caal} distributions for each 
length bin allows only the additional information provided by the limited age data 
(relative to the generally far more numerous length observations) to be captured, 
without creating a ‘double-counting’ of the data in the total likelihood.
Second, in addition to being able to estimate the basic growth parameters inside
the assessment model, the distribution of lengths at a given age, governed by two 
parameters for the standard deviation of length at a young age and the standard 
deviation at an older age, is also quite reliably estimated. This information could 
only be derived from marginal age-composition observations where very strong and 
well-separated cohorts existed and where they were quite accurately aged and measured; 
rare conditions at best. By fully estimating the growth specifications within the 
stock assessment model, this major source of uncertainty is included in the assessment 
results, and bias in the observation of length-at-age is avoided. 

\Gls{caal} compositions were only implemented for male and female `r spp` and
no sex ratio was applied to unsexed fish
(Table \ref{tab:sample-size-age}).
This resulted in `r ifelse(params$area == "North", 60, 504)`
unsexed fish being excluded from the \gls{caal} distributions, or approximately
`r ifelse(params$area == "North", "one", "fifteen")` percent of the aged fish.
Sensitivities to using \gls{caal} data instead of marginal age data
were explored by replacing \gls{caal} compositions with marginal-age compositions.

#### \acrlong{s-tri} ages

The preparation of age data from \gls{s-tri}
(Table \ref{tab:sample-size-age}; Figure \ref{fig:tri-ageComp})
followed the same methods as the \gls{s-wcgbt}.

```{asis, opts.label = 'south', echo = TRUE}
#### \Glsentrylong{s-hkl} lengths

The preparation of age data from \gls{s-hkl}
(Tables \ref{tab:sample-size-age}; Figure \ref{fig:hkl-ageComp})
followed the same methods as the \gls{s-wcgbt}.

Lam research ages

A random stratified subsample by size and sex was selected per region for ageing and genetic analysis. The age-composition data were post-processed as \gls{caal} data similar to the other surveys (Table \ref{tab:sample-size-age}; Figure \ref{fig:lam-ageComp})

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