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

Oregon nearshore commercial fishery logbook index

\Glsentrylong{odfw} has required nearshore, permitted vessels and open access vessels within the commercial fishery to submit fishing logbooks since 2004. Compliance is generally high, averaging around 80\%. Early compliance was lower, e.g., 65\% in 2007; while recent compliance has reached greater than 90\%. Furthermore, although fishers are required to provide all requested information in the logbook per fishing gear set, there has been substantial variation in the quantity and quality of information reported. In theory, the database contains information per gear set on catch by species (retained and released fish), effort (hook hours, number of hooks used multiplied by the number of hours fished), sample location (port), date, vessel, fishing depth, fishing gear, fishing permit, number of fishers, and harvest trip limits. For this analysis, multi-set trips were aggregated to the trip level, and the full data set included data from 38,350 trips from 2004--2020. Weight of released fish (numbers) was converted converted to weight by multiplying numbers by median catch weight prior to calculating \gls{cpue}.

Data were filtered to increase their consistency (Table \@ref(tab:cpue-commercialfixed-gear-filter)). Filtering helps to attain the best possible consistent and representative data set for the time period available such that trends in relative abundance reflect stock biomass. In general, encounter rates for r spp were relatively high, and thus, filtering was minimal compared to that done for some other species. Records were eliminated if they contained missing or unrealistic values. Trips were reduced to only those from permitted vessels using hook-and-line jig gear from Port Orford, Gold Beach, and Brookings ports and vessels that fished in at least three years since the implementation of logbooks. Filtering did not account for vessel operator and was tied only to the vessel name. Gear types other than hook-and-line were excluded because hook-and-line gear accounted for approximately 78.5\% of all documented trips for r spp. Data were also filtered for trips that were not abnormally deep. Some of these filters are new relative to the previous assessment [@haltuch2019lingcod] and were developed for the recent cabezon assessment [@cope2019cabezon].

\Glsentrylong{cpue} was modeled using a Bayesian delta-\gls{glm} model [@rstanarm2020]. Probability of occurrence, the binomial component, was modeled using a logit link. Log of positive catch, the rate component, was modeled using a Gaussian distribution with an identity link function. Covariates were selected using \gls{aic} and \gls{bic}, retaining only those that improved model fit. The full model included covariates for year, month, vessel, port, depth, and people. Depth, a continuous variable, was included to account for general differences in bathymetry and restrictions where fishing can take place. People was included to control for the potential oversaturation of hooks at a given fishing location and increased fishing efficiency with more fishers onboard. Accounting for vessel also accounts for variability in fishing capacity. Period, two-month interval was also considered but was excluded due to multicollinearity with month.

A quantile-quantile plot of the simulated [@hartig2021] residuals suggested the binomial model provided a reasonable approximation of the data (Figure \@ref(fig:cpue-commercialfixed-gearqqbin)). However, the fits were not as good for the rate model (Figure \@ref(fig:cpue-commercialfixed-gearqqpos)). Despite filtering, there were a wide range of positive \gls{cpue} values, resulting in a heavy left skew. The binomial model closely matched the distribution from replicate datasets and the positive model matched less closely. Multiple alternative distributions and more parsimonious positive models were explored, as well as aggressive filtering for outliers. But, none of the efforts led to improved diagnostics. There was insufficient time to fully resolve these diagnostic issues and alternative approaches may be need for future assessments. The final time series indicated an increase in 2010 that was in line with other data sets (Figure \@ref(fig:index-fits-all-fleets)).

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