Fish and cephalopods monitoring on the Bay of Biscay and Celtic 1 Sea continental shelves

10 The demersal fish and cephalopod communities of the continental shelves of the Bay of Biscay and the Celtic Sea 11 have been monitored for more than 30 years by the EVHOE series of fisheries surveys. Since 1987, a total of 4247 12 stations have been sampled in the fall with a GOV bottom trawl in a depth range of 15 to 600m. The main objective 13 of these surveys is to monitor 22 benthic fish stocks and 10 cephalopods but also to provide a description of the 14 distribution of a total of 250 fish and 50 commercial invertebrate taxa. The dataset 15 (https://doi.org/10.17882/80041) provides abundance and biomass information by station for all observed taxa. 16 Size distributions for a selection of species are also available. These data are part of a larger set of standardized 17 European surveys that provide essential information for monitoring demersal communities in the Northeast 18 Atlantic. We propose here a critical analysis of the dataset especially in terms of the evolution of the sampling 19 effort and strategy as well as the taxonomic precision.


Components
8.ab -From the initial and contractual stock assessment objectives, a more diversified data collection has been put in 48 place to progressively monitor the entire marine ecosystem. In addition to the fish and cephalopods species 49 historically observed, the entire benthic invertebrate community ("benthos") as captured by the trawl has now been 50 recorded since 2008. It provides information on regional biodiversity, improves our understanding of the structure

59
In the present paper we provide details of the long term dataset for fish and cephalopods collected on the 60 continental shelves of the Bay of Biscay and the Celtic Sea during the EVHOE survey. Non-commercial 61 invertebrate ("Benthos") data are not included in this first dataset; they will be the subject of a later addition. a few years were missing from the data series for autumn sampling (1991,1993,1996,2017) ; the absence of a 77 survey in these years were usually due to technical problems with the R/V. The studied area was limited to the 78 Bay of Biscay, between the latitudes 43°40' N and 48°30'N, from 1987N and 48°30'N, from to 1989N and 48°30'N, from . In 1990    The net is fitted with a 20 mm codend liner. During the trawling, the gear parameters were monitored by "Scanmar" 93 system (Table 2) and in recent years by "Marport" system. The parameters that are monitored are the door spread, 94 the wing spread, the headline height and the height of ground rope. They allow appreciating the behavior of the 95 gear during fishing operations but also to assess the area or even the volume sampled. The accuracy of trawling 96 parameters has therefore evolved over time and the data from the trawl geometry sensors were not recorded until  swept area (about 0.076 km 2 for a standard 30' tow) was then calculated from the distance covered by the trawl.

100
The file also provides the duration of the haul, which is a useful standard effort value when combining data from 101 different surveys using a similar fishing gear. Moreover, the trawl is equipped with a CTD probe allowing for each 102 station to record temperature, depth and salinity profiles (the latter only for stations less than 300m deep).

103
Additionally, a number of navigational parameters or meteorological variables were also monitored but are not 104 included into the published dataset.

105
The sampling scheme defined a geographic stratification that separates the Bay of Biscay in 2 areas and the Celtic 106 Sea into 3 areas and seven depth strata from 20 m to 600 m ( Fig. 1 and Table 3). From 1987 to 2015, the sampling 107 strategy followed a stratified random strategy (Fig. 1). A Neyman allocation on numbers variance averaged on the 108 4 most important commercial species (hake, the two species of monkfish and northern megrim) was utilized to set

115
From 2016, the sampling strategy was changed to a fixed sampling strategy. The reason for this change was that 116 the spatial coverage of some large strata was too highly variable from one year to another. Thus, depending on the 117 random selection of points, areas of significant size were left unobserved. The stabilization of sampling points also 118 facilitated analyses that aimed at studying the spatial structures of species or communities and their evolution over

123
(e.g. some rarely sampled points into enclosed bays). In the central-eastern part of the Celtic sea, we added 4 124 additional to complete strata coverage.

125
Sampling was carried out with straight tows during the daylight, lasting 30 minutes at the bottom (a minimum of 126 20 minutes accepted in the protocols to validate a haul) at a constant speed of 4 knots. Some tows were stopped 127 before the end of the total trawl time when excessively high tensions were detected (a sign of large catches) or 128 more recently (from 2018 onward) when a strong pelagic acoustic signal was observed from the on-board sounders.

129
These tows were considered valid and included in the dataset when they lasted at least 20 minutes and that the 130 fishing gear has not suffered any damage. They represent less than 10% of the tows (about 2 to 14 tows per year) 131 with higher proportions in recent years due to the improved control of the trawl variables described above.

134
Wherever possible, the entire catch was sorted, with fish and commercial shellfish, crustaceans and 135 cephalopods species identified to the lowest taxonomic level. On the other hand, when the total catch in the trawl 136 was too large (e.g. several tons of small pelagic fish), only a fraction of the total catch was fully processed (mostly 137 1/2 to 1/4 and exceptionally >1/5 of the total catch weight). For the partially sorted part, individuals of rare or 138 particularly large species were still extracted and processed. On average for the recent years (from 2014), those 139 partially processed tows represented 11 to 18% of the total number of stations. Due to a lack of data, this proportion 140 could not be properly assessed for surveys prior to 2014. It can be assumed that this proportion may have been 141 higher in the past, particularly at Thalassa I, due to less efficient sorting facilities.

142
Individuals lengths were recorded for most fish species and some commercially important cephalopods class and by sex. However, these data required significant revision and were not included in the submitted version 151 of the dataset. They will be the subject of an additional publication.

152
Data entry on board was initially carried out on paper forms that were then copied to computer databases. Starting

153
in 1997, on the R/V Thalassa 2, a computer system for recording catches ("pupitri") allowed for the automated 154 banking of species and their total weight, with individuals informations (sex, counts, size measurements, maturity) 155 still being entered on paper forms. These data were then transferred to an internal database under "Microsoft

156
Access" software (database specific to the EVHOE campaign, not standardized with others IFREMER databases).

159
From 2016 onwards, the lengths were also measured using an electronic ichthyometer directly connected to the 160 data management system. Only the sizes of the largest individuals (> 85cm) and the weight data of the sub-samples 161 and individuals fish were still entered manually. In addition, a set of automated data control and correction tools 162 were put in place in recent years (both within the "Allegro" software and from separate dedicated tools greatly 163 improved in 2020). These tools have been applied a posteriori on the whole data series; the EVHOE dataset 164 proposed here (Laffargue et al. 2021) has therefore been significantly corrected.

165
The dataset consists of 3 tables in a ".csv" file format (Table 4)  In addition to a description of the data, we propose a short critical analysis by comparing in particular the evolution

261
We can also note the interest of the data produced to identify certain elements of the remarkable diversity that is 262 all the more appreciable with the improvement in the quality of species determination. In particular, the detection 263 of rare or new species in the study area are valuable data for characterizing regional biodiversity and judging the 264 evolution of continental shelf communities.

265
The strength of this series also lies in the additional data (hydrological, other biological compartments) acquired 266 simultaneously and offering an increasingly complete panorama of the ecosystems of the continental shelves in 267 the fall period. These additional observations, which are sometimes relatively recent, are processed independently 268 of this dataset and will be the subject of subsequent publications.

Raja undulata, Raja microocellata
Raja microocellata more rare before 2000 due to probable confusion

Raja montagui Raja bravhyura
The criteria are difficult (even for specialists) and errors are possible especially before 2010 but still likely in recent years.

Scyliorhinus stellaris
Sporadically identified at the beginning of the series, S.stellaris appears more frequent from 2010 which may show confusion between the 2 species. The size of the individuals and the difficulties of identification make certain determinations unreliable, particularly of P. minutus (e.g. only 1 species of the genus Pomatoshistus before 2002).

Scorpaena
Difficulties in identification lead to frequent errors. The species Scorpaena elongata is most probably mistakenly identified and has not been described in the Bay of Biscay from others studies.

S. typhle
Identifications are difficult and errors are likely to occur during the data series; greater attention paid to these species after 2017.

Trachurus trachurus T. mediterraneus, T. picturatus
The sometimes very high abundance of horse mackerel in the catches and a consecutive important sub-sampling make the detection of closely resembling but rarer species more difficult.