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

Oregon recreational index

Trip-level \gls{cpue} data from \gls{orbs} dockside sampling program was analyzed by \gls{odfw} and included in the assessment as a relative index of abundance for the Oregon recreational fleet. The dataset represents the longest time series available for this fleet and contains information on catch by species (number of retained fish), effort (angler hours), sample location (port where data were collected), date, bag limits, boat type (charter or private), and trip type (e.g., bottom-associated fish). \Gls{cpue}, expressed in terms of fish per angler-hour. Angler-hour is the number of anglers multiplied by the number of hours fished minus by travel time. Travel time is twice the distance between the port of origin and the fished reef multiplied by 13 mph for charter boat trips and 18 mph for private boat trips.

The unfiltered data set contained 411,528 trips. Multiple filters were applied to the trip-level data to increase the number of positive trips for r spp (Table \@ref(tab:cpue-recreationalOregon-filter)). These filters removed trips from several ports with extremely small sample sizes; trips that met criteria for implausible effort values or extreme catch rates; trips with incorrect interview times, which impact calculation of effort; unreasonably long or short trips; catches exceeding bag limits, which would impact catch rates; private vessel trips; and trips that did not target bottom fish. Filters were also utilized to account for temporal or spatial closures impacting fishing. Stephens-MacCall filtering was also investigated to identify trips that were likely to catch r spp given the occurrence of other species in the catch [@stephens2004]. Stephens-MacCall filtering did not change the diagnostics of the resulting model, and thus, it was not utilized to filter the data.

A Bayesian delta-\gls{glm} was used to model \gls{cpue} [@rstanarm2020]. Investigated covariates included month, mega-region (north and south coasts divided at the port of Florence on the central Oregon coast), port, season (summer and winter), r spp bag limit, and depths available to the recreational fleet for fishing (Figure \@ref(fig:reccpueor-orbsdatasummary)). Season and mega-region were collinear with month and port, respectively, and not investigated further.

The binomial model component for positive catches was modeled using a logit link function. Based on \gls{aic} and \gls{bic}, the best binomial model included year, port, month, and depths available to fishing. Models with interactions between year and port or month did not converge due to missing trips in some strata. A quantile-quantile plot of the simulated residuals [@hartig2021] suggests that the binomial component of the delta-model is a reasonable approximation of the data (Figure \@ref(fig:cpue-recreationalOregonqqbin)). Diagnostics were substantially improved by excluding interactions between year and open depths and the removal of private boat trips.

The rate component for log of positive \gls{cpue} was modeled with a Gaussian distribution and an identity link function. The positive model included year, port, month and open depths available to fishing. There were issues in the DHARMa simulated residuals, including curvature and significant outliers (Figure \@ref(fig:cpue-recreationalOregonqqpos)). However, these diagnostics were improved by exclusion of any interaction terms and the removal of the private vessel trips. There was insufficient time to fully resolve these diagnostic issues and alternative approaches may need to be explored for future assessments. For example, an area-weighted model was not utilized for r spp, as has been used for other nearshore species in recent assessments, such as cabezon [@cope2019cabezon] or blue rockfish [@dick2018bluedeacon].

The final index (Figure \@ref(fig:index-fits-all-fleets)) matched major trends in other \gls{cpue} fits with an increase in 2010 and a current decrease in abundance.

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