Changing distribution and abundance of small pelagics may drive changes in predator distributions, affecting predator availability to fisheries and surveys. However, small pelagic fish are difficult to survey directly, so we developed a novel method of assessing small pelagic fish aggregate abundance via predator diet data. We used piscivore diet data collected from multiple bottom trawl surveys within a Vector Autoregressive Spatio-Temporal (VAST) model to assess trends of small pelagics on the Northeast US shelf. The goal was to develop a spatial “forage index” to inform survey and/or fishery availability in the bluefish (Pomatomus saltatrix) stock assessment. Using spring and fall surveys from 1973-2020, 20 small pelagic groups were identified as major bluefish prey using the diet data. Then, predators were grouped by diet similarity to identify 19 piscivore species with the most similar diet to bluefish in the region. Diets from all 20 piscivores were combined for the 20 prey groups at each surveyed location, and the total weight of small pelagic prey per predator stomach at each location was input into a Poisson-link delta model to estimate expected prey mass per predator stomach. Best fit models included spatial and spatio-temporal random effects, with predator mean length, number of predator species, and sea surface temperature as catchability covariates. Spring and fall prey indices were split into inshore and offshore areas to reflect changing prey availability over time in areas available to the recreational fishery and the bottom trawl survey, and also to contribute to regional ecosystem reporting

Using NEFSC bottom trawl survey diet data from 1973-2021, 20 small pelagic groups were identified as major bluefish prey with 10 or more observations (in descending order of observations): Longfin squids (Doryteuthis formerly Loligo sp.), Anchovy family (Engraulidae), bay anchovy (Anchoa mitchilli), Atlantic butterfish, (Peprilus triachanthus), Cephalopoda, (Anchoa hepsetus), red eye round herring (Etrumeus teres), Sandlance (Ammodytes sp.), scup (Stenotomus chrysops), silver hake (Merluccius bilinearis), shortfin squids (Illex sp.), Atlantic herring (Clupea harengus), Herring family (Clupeidae), Bluefish (Pomatomus saltatrix), silver anchovy (Engraulis eurystole), longfin inshore squid (Doryteuthis pealeii), Atlantic mackerel (Scomber scombrus), flatfish (Pleuronectiformes), weakfish (Cynoscion regalis), and Atlantic menhaden (Brevoortia tyrannus).

Prey categories such as fish unidentified, Osteichthyes, and unidentified animal remains were not included in the prey list. Although unidentified fish and Osteichthyes can comprise a significant portion of bluefish stomach contents, we cannot assume that unidentified fish in other predator stomachs represent unidentified fish in bluefish stomachs.

Image credits: Striped and bay anchovy photo--Robert Aguilar, Smithsonian Environmental Research Center; redeye round herring photo--https://diveary.com ; sandlance photo--Virginia Institute of Marine Science; all others NOAA Fisheries.

Bottom trawls are not designed for efficient sampling of small pelagics. While we can estimate aggregate planktivore biomass for different ecoregions on the Northeast US shelf, estimates have high observation error. Black lines = NEFSC bottom trawl, Red lines = NEAMAP inshore bottom trawl. Orange line on GB Fall indicates significant increasing long term trend.

Spatial and spatio-temporal correlation decay with increasing distance estimated as $\kappa$ in a Matern function with fixed smoothness and geometric anisotropy (directional correlation, optionally estimated by the model).

Initial model selection consistently supported the inclusion of spatial and spatio-temporal random effects and anisotropy across all datasets: fall, spring, and annual.

The figure shows bluefish diet collection stations for fall surveys, 1985-2019.

NEAMAP survey stations with diet collections for piscivores (n = 3838) had a higher proportion with our defined bluefish prey (n = 2418, 63.0015633%).

All size classes of 50 fish predators captured in the NEFSC bottom trawl survey were grouped by diet similarity to identify the size classes of piscivore species with the most similar diet to bluefish in the region. Diet similarity analysis was completed using the Schoener similarity index (@schoener_nonsynchronous_1970; B. Smith, pers. comm.), and is available available via this link on the NEFSC food habits shiny app. The working group evaluated several clustering methods to develop the predator list (see this link with detailed cluster results).

Predators with highest diet similarity to Bluefish from the NEFSC diet database (1973-2020) include Atlantic cod, Atlantic halibut, buckler dory, cusk, fourspot flounder, goosefish, longfin squid, shortfin squid, pollock, red hake, sea raven, silver hake, spiny dogfish, spotted hake, striped bass, summer flounder, thorny skate, weakfish, and white hake. The NEAMAP survey operates closer to shore than the current NEFSC survey. The NEAMAP dataset includes predators sampled by the NEFSC survey and adds two species, Spanish mackerel and spotted sea trout, not captured by the NEFSC survey offshore but included based on working group expert judgement of prey similarity to bluefish. Predator size classes included are listed in Table 2 of the forage fish index working paper at this link.

Image credits: Weakfish and Spanish mackerel-- https://marinefishesofgeorgia.org ; spotted seatrout-- https://fishinginmiami.com ; Sea Raven photo 2/11/2019 11:07:56 AM, Photographer: Andrew J. Martinez, Location: Massachusetts, Stellwagen Bank NMS; all others NOAA Fisheries.

Diets from all 22 piscivores (including bluefish) were combined for the 20 forage fish (bluefish prey) groups at each surveyed location, and the mean weight of forage fish per predator stomach at each location was calculated. Data for each station included station ID, year, season, date, latitude, longitude, vessel, mean bluefish prey weight (g), mean piscivore length (cm), number of piscivore species, and sea surface temperature (degrees C). Because approximately 10% of survey stations were missing in-situ sea water temperature measurements, National Oceanic and Atmospheric Administration Optimum Interpolation Sea Surface Temperature (NOAA OI SST) V2 High Resolution Dataset [@reynolds_daily_2007] data provided by the NOAA PSL, Boulder, Colorado, USA, from their website at https://psl.noaa.gov were used to fill gaps. For survey stations with in-situ temperature measurements, the in-situ measurement was retained. For survey stations with missing temperature data, OI SST was substituted for input into VAST models.

Models were developed combining all data for the year ("Annual") and with separate data for "Spring" (collection months January - June) and "Fall" (collection months July-December) to align with assumptions used in the bluefish stock assessment. Modeled years included 1985-2021 to align with other data inputs in the bluefish stock assessment.

SST is also likely to affect prey distribution, but differently for each prey species. Therefore, SST is not modeled as a density covariate for aggregate small pelagics.

NEFSC survey strata definitions are built into the VAST northwest-atlantic extrapolation grid already. We defined additional new strata to address the recreational inshore-offshore 3 mile boundary. The area within and outside 3 miles of shore was defined using the sf R package as a 3 nautical mile (approximated as 5.556 km) buffer from a high resolution coastline from thernaturalearth R package. This buffer was then intersected with the current FishStatsUtils::northwest_atlantic_grid built into VAST and saved using code here. Then, the new State and Federal waters strata were used to split NEFSC survey strata where applicable, and the new full set of strata were used along with a modified function from FishStatsUtils::Prepare_NWA_Extrapolation_Data_Fn to build a custom extrapolation grid for VAST as described in detail here.

aggregate forage important in this system with many forage species: Jason Link horrendogram

potential trends in aggregate forage to be compared with trends in zooplankton and environmental drivers

WHAM is a state space stock assessment model framework: https://timjmiller.github.io/wham/

The Bigelow index fit with the fall forage fish index did not improve the model fit (AIC), was slightly worse fit and gave identical results The Albatross index fit with the fall forage fish index did not converge or hessian was not positive definite for any of the models (even when how = 0 for some of them). The MRIP index fit with the annual forage fish index did not converge or hessian was not positive definite for any of the models