Plankton Imager zooplankton data, acoustic fish biomass estimates, chlorophyll measurements and CTD data from the Celtic Sea and Western English Channel in October 2018

notPlanned

All data were recorded as part of the annual PELTIC survey (PELagic ecosystem in the western English Channel and eastern CelTIC Sea). During the day, multifrequency acoustic data are acquired along the transects and combined with *ad hoc* pelagic trawls to map and quantify the small pelagic fish community. At night, zooplankton and chlorophyll (as a proxy for phytoplankton) information was gathered at 23 stations, along the acoustic transects. Discrete samples were collected for chlorophyll analysis, while the Plankton Imager (PI) was used to obtain zooplankton taxonomic and size data. The structure of the water column at all these stations and at 83 additional stations in the study area was determined using a SAIV SD204 conductivity-temperature-depth sensor (SAIV A/S Environmental Sensors & Systems, Norway). Acoustic data were collected along transects during the day, using a Simrad EK60 scientific echosounder, with the split-beam transducers mounted on the vessel’s drop keel at a depth of 3.2 m below the vessel’s hull or 8.2 m sub surface. Three operating frequencies were used during the survey (38, 120 and 200 kHz) for trace recognition purposes, with 38 kHz data used to generate the abundance estimate for clupeids (and other fish with swimbladder) and 200 kHz for mackerel (Van Der Kooij et al., 2016). All frequencies were calibrated at the start of the survey. A single pelagic midwater trawl with a vertical opening of c. 12 m was used to collect information on species and size composition and provide biological samples, and was fitted with a 20 mm codend liner to ensure the retention of small and juvenile fish. Trawl monitoring, trawl door type and dimensions and rigging are provided in ICES (2015). As the trawls were deployed to obtain qualitative rather than quantitative information, no fixed trawl duration was employed during the survey although deployment was generally 30 minutes. All components of the catch from the trawl hauls were sorted and weighed; fish and other taxa were identified to species level. Length frequency and length-weight data were collected for all species of the catch. Length measurements of sprat *Sprattus sprattus*, sardine *Sardina pilchardus*, anchovy *Engraulis encrasicolus*, boarfish *Capros acer* and herring *Clupea harengus* were to be taken to the nearest 0.5 cm below, horse mackerel *Trachurus trachurus* and blue whiting *Micromesistius poutassou* were measured to the whole cm below total length. Where possible the total catch component of the haul per species was measured. When this was not, a suitable sub sample was selected to provide a true (length) representation of the species. Biomass estimates for pelagic fish species followed routine methods (ICES, 2015). The acoustic recordings of Nautical Area Scattering Coefficient (NASC, m2nmi-2) for each nautical mile along the transects were partitioned by species based on school characteristics and trawl catches. To determine the underlying spatial distribution of pelagic fish species, statistical models (Generalized Additive Models, GAMs) were employed using physical covariates as predictors (i.e. latitude/longitude/depth/distance from coast). Analysis of covariance between predictors using the Variance Inflation Factor indicated that depth and distance from coast were strongly correlated (VIF > 2). Similarly, depth and longitude where strongly correlated so models were built for fish species using latitude/longitude/distance from coast only (all VIF < 2). The acoustic data for fish in 2018 demonstrated a high percentage of zero densities in the along-track relative biomass estimates (NASC) ranging from 16% for *Sardina pilchardus*, 23% for *Trachurus trachurus* and 26% for *Engraulis encrasicolus* to much higher levels for *Sprattus sprattus* (62%), *Clupea harengus* (75%), *Capros acer (9*1%) and *Micromesistius poutassou* (97%). Only *Sardina pilchardus*, *Trachurus trachurus, Engraulis encrasicolus,**Sprattus sprattus *and *Clupea harengus* were considered for further analysis. NASC data are continuous so commonly used zero-inflated models such as the negative binomial GAM are unsuitable in this instance. Instead, NASC data were fourth-root transformed prior to analysis to normalise the data and a gaussian GAM model fitted using a 3D tensor smooth. The distribution of species was then projected across the survey area and the predicted values compared to the zooplankton size and abundance/biomass data from the PI. Zooplankton densities and size (length) were extracted from the images collected by the PI at 23 stations selected as part of the wider survey aims. The PI is a high-speed colour line scan-based imaging instrument connected to the ship water supply and can take images of all zooplankton of size above 100 µm passing through a cell at a 22 l.min-1 flow rate. The flow cell is 25 mm brass tube that has two quartz optical windows halfway along its length. The flow cell at the windows is rectangular, with the same cross-sectional area as the 25 mm inlet. A Basler 2048-70kc camera, sampling at 70K lines per second, images the water running through the flow cell. The flow rate is monitored by a Bell electro-magnetic flow meter. Colour images are captured using an EPIX E8 frame store. RGB composite images are constructed by joining consecutive lines together, thresholding and extracting a region of interest ROI, or vignette, that is saved to hard drive as a TIF file. Each TIF image is time-stamped and named in the Zooscan convention of date+imageID.tif. Raw images are stored to maximize dynamic range of the captured particles. These are converted to 8-bit resolution through a process of scaling and conversion from 12 to 8-bit resolution, for viewing and for subsequent processing. The PI was in operation continuously throughout the survey and thus acquiring images of all particles passing through the flow-cell. Due to operational requirements (continuous image processing while maintaining manageable file-size), only those particles within the mesozooplankton range 200µm – 2cm were processed and saved. In order to reproduce the sub-sampling procedure used for physical samples (as collected with a ring net), random images were extracted until each sample contained a minimum of 200 mesozooplankton), analysed with the PI for taxonomic data and then identified and validated by an expert taxonomist manually (also providing a new source of training data for future use by the machine learning PI image recognition algorithm). Lengths were measured for all individuals in each subsample, as the largest distance between 2 pixels. Biomass values were estimated for copepods only, because these are the most critical food for zooplanktivorous fish. A length-mass regression was applied to the PI length measurements in order to derive average individual wet weight values for each taxonomic group (μg WW ind-1), as well as a wet weight for each individual copepod measured in the subsample of images analysed for taxonomic data. To derive biomass values (mg WW m-3) using individual image sizes, the total observed copepod wet weight for each station was summed, and then scaled by the number of images analysed and the water volume sampled at each station. Using the average size of each copepod group, the mean wet weight of the group was multiplied by the abundance of that group in ind. m-3 . For the length-mass regression, we used literature information on copepod prosome and urosome sizes. Individual volumes for copepod species were then calculated by approximating the body shape of a copepod as an ellipsoid for the prosome and a cylinder for the urosome. Volumes were then converted to wet weights by application of a specific gravity factor of 1.025. These calculated species-specific wet weights were plotted against their associated total lengths (i.e. prosome + urosome lengths) measured by the PI and the regression with best fit model was applied to derive the follwoing equation: Wet Weight = 0.0299 * Total Length^2.8348.