Common name | Species | Max length (cm) | Age class | Depth range (m) |
|---|---|---|---|---|
Arrowtooth flounder | Atheresthes stomias | 22 | 1 | 50-470 |
Darkblotched rockfish | Sebastes crameri | 15 | 0-1 | 80-240 |
Dover sole | Microstomus pacificus | 17 | 1-2 | 50-465 |
English solea | Parophrys vetulus | 16 | 1 | 50-140 |
Lingcod | Ophiodon elongatus | 25 | 0 | 50-240 |
Longspine thornyhead | Sebastolobus altivelis | 7 | <5 | 385-1245 |
Pacific grenadier | Coryphaenoides acrolepis | 3 | ~1 | 490-1275 |
Pacific hake | Merluccius productus | 15 | 0-1 | 50-700 |
Pacific sanddab | Citharichthys sordidus | 13 | 0-1 | 50-245 |
Petrale sole | Eopsetta jordani | 21 | 1-2 | 50-200 |
Sablefish | Anoplopoma fimbria | 29 | 0 | 50-475 |
Shortspine thornyhead | Sebastolobus alascanus | 8 | <5 | 160-625 |
Splitnose rockfish | Sebastes diploproa | 10 | 0-1 | 65-460 |
27 Juvenile Groundfish Abundance
Description Yearly indices of the abundances of juvenile sablefish, Dover sole, shortspine thornyhead, and longspine thornyhead along the West Coast were calculated using species distribution models. Strong year classes can determine age structure and set stock size for marine fishes, and may also indicate favorable environmental conditions, increased future catches, and impending potential bycatch issues. Here, we provide estimates of juvenile abundance for 13 species of West Coast groundfishes, including four from DTS assemblage (Dover sole, thornyheads, and sablefish) as a potential leading indicator of incoming strong year classes. The DTS assemblage is a valuable fishery, and bycatch of some species, like small sablefish, can impact other fisheries such as the at-sea hake fishery.
Indicator Category Ecological Integrity
Data Steward N. Tolimieri, NMFS/NWFSC (nick.tolimieri@noaa.gov)
Additional Information Data are collected and analyzed independently by N. Tolimieri, who submits unpublished figures and plots to the CCIEA editorial team.
Data sources Data for indicators come from the West Coast Groundfish Bottom Trawl Survey (WCGBTS) (Keller et al. 2017) for 2003-2021. There were no data for 2020 because the WCGBTS was canceled due to the COVID-19 pandemic. The survey data includes estimates of age, length, and biomass for subsamples of each haul, and occasionally for the entire haul when catch is low.
Data extraction Data were downloaded from the Fishery Resource Analysis and Monitoring data Warehouse (https://www.webapps.nwfsc.noaa.gov/data/map)
Data analysis We used species distribution models to calculate indices of abundance for juvenile groundfish. The approach follows the general approach of Tolimieri, Wallace, and Haltuch (2020) but uses the ‘sdmTMB’ package (Anderson et al. 2022) for R instead of the ‘VAST’ package (Thorson 2019). VAST was was reviewed by the SSC-ES in September 2021. The sdmTMB approach is used by many West Coast groundfish stock assessment biologists to assimilate survey data and was reviewed favorably by the SSC-Groundfish Subcommittee in summer of 2022 (PFMC 2022c).
The analyses estimate the biomass for each species by using length-age and length-weight relationships to expand the trawl data. Length is measured (cm total length) for all individuals in the subsample, but many individual fishes lack weight or age data due to time constraints in the field and ageing lab. To expand the subsample,
Missing weights for individuals in the subsample were obtained by first estimating the length-weight relationship from existing data and using this relationship to estimate the missing weights from known lengths. For Dover sole, sablefish and longspine thornyhead, male and female length-weight relationships were estimated separately and the average of these relationships used to determine weights for individuals where sex was not known. For shortspine thornyhead, we used a single length-weight relationship.
Individual fish were then allocated to age classes following Tolimieri, Wallace, and Haltuch (2020) by using length-age relationships from the WCGBTS data to determine age-class maximum lengths. See Tolimieri, Wallace, and Haltuch (2020) for more detail. The maximum lengths used here (Table 27.1) were taken from Tolimieri, Wallace, and Haltuch (2020) (Table 27.1).
The proportional biomass of juveniles in each subsample was calculated and used to estimate the total biomass of juvenile fishes in the full trawl.
Trawl biomass was then used in the following sdmTMB species distribution models.
Coastwide juvenile groundfish abundances were estimated using a spatially explicit, species distribution model evaluated with the sdmTMB package in R. The response variable was CPUE quantified as kg of juveniles per km2. The models included one common intercept across years, and spatial and spatiotemporal random fields, with anisotropy to account for different rates of autocorrelation with latitude versus longitude (~ depth). The common intercept prevents the model from forcing biomass to increase or decrease coastwide in a given year (thereby potentially overestimating recruitment in some areas) as would be the case for yearly intercept. Normalized depth was included to account for differences in density across depths.
To avoid projecting to areas with zero biomass, the depth range of the data used for each species in the analysis was restricted based on the distribution of positive biomass observations (Table K.1). Again, the values used here follow Tolimieri et al. (2020) with the exception that the lower depth limit for sablefish was set to 250 m, which encompasses more than 99% of their observed juvenile biomass. Pass was included as a fixed factor (as a proxy for time of year; the WCGBTS conducts two coastwide passes each year, in May-July and August-October). Models were fit with a delta-gamma distribution to account for the prevalence of zeros in the data, and the mesh was set to 10 km, resulting in 650-800 knots depending upon species. Model fits were then extrapolated to a 2x2 km grid of the West Coast to estimate total abundance in kg for juveniles in a given year. For some species, it was necessary to combine age or size classes to obtain enough data for models to converge. The resulting biomass estimate was converted to an index scaled between 0-1 by dividing all values by the maximum upper 95% confidence limit in the time series.
To address previous suggestions by the SSC-ES, we also evaluated models with year included as a fixed factor or allowed year to have a random intercept. When included as a fixed factor, models failed to converge likely due to identifiability problems due to also including the spatiotemporal random field. For sablefish and Dover sole, inclusion of year with a random intercept also created fit problems leading to very large standard errors for some estimated parameters. Therefore, we excluded the term from the final models.
2025-01 Update
Previously models for all species included a delta-poisson-link-gamma model/error structure and normalized depth as a linear variable. In 2025 models were updated to tailor model structure and the form of depth (linear, quadratic, smoothed) to individuals species. We fit models with Tweedie, delta-lognormal, delta-poisson-link-gamma, and delta-gamma model/errors. Depth was normalized and included as a linear variabe, as a quadratic, smoothed (GAM), or not included. The best model was chosen for each species based on a comparison of AIC values, QQ residual plots, and sanity (model fit) output (Table 27.2). This approach produced better residuals and tended to dampen some of the higher biomass estimates (see sablefish in the main report).
Species | Distribution | Depth |
|---|---|---|
Arrowtooth flounder | Tweedie | GAM smooth |
Darkblotch rockfish | Tweedie | GAM smooth |
Dover sole | delta-gamma-poisson-link | quadratic |
English sole | delta-gamma | quadratic |
Lingcod | Tweedie | quadratic |
Longspine thornyhead | delta-lognormal | GAM smooth |
Pacific grenadier | delta-gamma | GAM smooth |
Pacific hake | Tweedie | GAM smooth |
Pacific sandlance | Tweedie | quadratic |
Petrale sole | delta-lognormal | quadratic |
Sablefish | delta-lognormal | quadratic |
Shortspine thornyhead | Tweedie | quadratic |
Splitnose rockfish | Tweedie | GAM smooth |
References
Anderson, Sean C, Eric J Ward, Philina A English, and Lewis AK Barnett. 2022. “sdmTMB: an R package for fast, flexible, and user-friendly generalized linear mixed effects models with spatial and spatiotemporal random fields.” bioRxiv, 2022–03.
Thorson, James T. 2019. “Guidance for decisions using the Vector Autoregressive Spatio-Temporal (VAST) package in stock, ecosystem, habitat and climate assessments.” Fisheries Research 210: 143–61.
Tolimieri, Nick, John Wallace, and Melissa Haltuch. 2020. “Spatio-temporal patterns in juvenile habitat for 13 groundfishes in the California Current Ecosystem.” PloS One 15 (8): e0237996.