Biogeographical patterns of meso- and bathypelagic ﬁsh along a Northeastern Atlantic transect

bathypelagic Northeastern The inﬂuence of oceanographic variables on assemblages of meso- and bathypelagic ﬁsh was investigated along a Northeastern Atlantic Ocean transect (Cape Verde to the Bay of Biscay) during May 2019. Fish were collected using a mrozooplankton trawl during daylight hours at ten stations. Along the transect, 17 hydrographic stations were also performed with a CTD (Conductivity, Temperature, and Depth). A total of 130 ﬁsh taxa were identiﬁed. The dominant family was Gonostomatidae, with four species ( Cyclothone braueri , Cyclothone microdon , Cyclothone pseudopallida , and Cyclothone pallida ) being responsible of more than 78% of the total density. The most frequent species that appeared to be ubiquitous were C. braueri and C. pseudopallida , while Myctophidae was the most diverse family. Multivariate analyses revealed two clusters related with the latitudinal gradient. The ﬁsh community in the southern stations (25–37 (cid:2) N) was more diverse than in the northern stations (42–48 (cid:2) N). Temperature from 300 to 700 m depth explained 65% of variation in terms of density and 58% in terms of biomass, both statistically signiﬁcant. The investigated variation in the deep-pelagic ecosystems on a large spatial scale gives essential information to ecosystem management approhes and marine spatial planning.


Introduction
The deep-pelagic ocean (waters deeper than 200 m) is the largest habitat by volume on Earth (Webb et al., 2010). In general, this dark environment is primarily cold and well oxygenated, but wide environmental variations do occur, for example in the oxygen minimum zones (OMZs) and the hydrothermal vent plumes (Thurber et al., 2014). The conditions of the deep-sea have resulted in a specialized fauna, often dominated by species that are rare or absent in shallower waters, such as myctophids, stomiiforms, or mrourids (Drazen and Sutton, 2017). The deep-sea provides crucial ecosystem services, including the support of fisheries, the provision of energy, and mineral resources, as well as the regulation of the climate and nutrient cycling (Thurber et al., 2014). Because of this and its vulnerability to threats such as overfishing and climate change (Webb et al., 2010), it is necessary to increase our knowledge of the deep-pelagic ocean.
Fishes are an important component of the deep-sea ecosystems and, as intermediate trophic levels as well as top predators, a critical part of food webs (Drazen and Sutton, 2017). Both, the mesopelagic (water masses between 200 and 1000 m depth) and the bathypelagic zone (from 1000 m depth to ca. 100 m from the seafloor) belong to the deep-sea ecosystems (Sutton, 2013). Although the biomass of mesopelagic fish is still in question due to the uncertainty of the oustic estimations (Proud et al., 2019), it probably dominates the global fish biomass (Irigoien et al., 2014). The recent global estimates based on oustic methods ($10 10 tons) are one order of magnitude larger than historical estimates based on net sampling (Gjøsaeter and Kawaguchi, 1980;Irigoien et al., 2014). Diel vertical migration, i.e. the tive migration to the epipelagic zone (0-200 m depth) at night to feed, is a general feature of the species inhabiting the mesopelagic zone (Robinson et al., 2010;Sutton, 2013). In this process, migrants tively transport carbon by respiration, excretion, and defecation to depth as well as being prey of carnivorous predators (Robinson et al., 2010;Davison et al., 2013). In addition, the efficiency of the energy transfer from phytoplankton to mesopelagic fish is higher than typically assumed, with mesopelagic fish respiring an estimated 10% of the primary production in deep waters (Irigoien et al., 2014). Hence, due to their ubiquity, high biomass and that they may be respiring $10% of the primary production in deep waters (Irigoien et al., 2014), mesopelagic fish play an important role in the biological pump (Drazen and Sutton, 2017). Furthermore, trophic intertion of deep-pelagic fish and demersal fish on continental slopes bypassing the detrital flux and transferring carbon to deep long-term storage plays a key role in the ocean carbon cycle (Drazen and Sutton, 2017).
The Northeastern Atlantic is an interesting region to investigate meso-and bathypelagic fishes because of its varying geographic and oceanographic properties. The main ocean circulation pattern is the northern North Atlantic subpolar gyre (SPG) and the southern subtropical gyre (STG). The study area was mainly affected by the southern gyre, but in Bay of Biscay there was an intergyre zone with weak circulation, enclosed by the two gyres (Pollard et al., 1996). This region is also influenced by the North Atlantic Drift, a northern branch of the Gulf Stream ( Figure 1).
The subtropical gyre (STG) is a wind-induced anticyclone gyre rehing down to 700-800 m depth in the eastern boundary (e.g. Valdés and Déniz-Gonzalez, 2015). It comprises of the eastward Azores Current in the north, which is a southern branch of the Gulf Stream, the southward Canary Current and the westward North Equatorial Current in the south at 15-20 o N (e.g. Valdés and Déniz-Gonzalez, 2015; Figure 1). The northeasterly winds along the African coastline cause upwelling of colder water masses along the coast. The main water mass in the eastern STG is the North Atlantic Central Water (NW), which is the principal water mass in the upper layer of the North Atlantic (e.g. Bashmhnikov et al., 2015). At intermediate levels, Mediterranean Overflow Waters (MOW) flow out through the Strait of Gibraltar and occupy the 700-1500 m depth. It is charterized by high salinities and temperatures (Carredo et al., 2016) and low oxygen and nutrient contents (Bashmhnikov et al., 2015). Northwest of the Iberian region below 500 m depth, NW gradually mixes with the MOW (Ríos et al., 1992). Within this region, wind-induced upwelling events are common and the main source of primary production for most of the year (Tenore et al., 1995).
Another principal water mass at intermediate level is Antarctic Intermediate Water (AAIW) that spreads northward from the Antarctic Circumpolar Current and along the African northwest coast to 30-32 N (Machín and Pelegrí, 2009). AAIW can be found at depth between 500 and 1200 m with a salinity minimum (Stramma and England, 1999). Originally, AAIW is oxygen rich, but the oxygen concentration decreases as it flows northward. In the central STG, high-salinity waters exist in the surfe layer (Subtropical Under Water), and near 20 N and 35 W, the salinity exceeds 37.0 psu, which is the highest in open ocean (O'Connor et al., 2005).
There is a global pattern of mesopelagic fish species richness, with highest values in equatorial waters, followed by subtropical, temperate, and high-latitude waters (Backus et al., 1977). The biomass of mesopelagic fish increases near large topographic structures like continental shelf breaks, seamounts, mid-ocean ridges, or volcanic islands (Hulley and Lutjeharms, 1989;Porteiro and Sutton, 2007;Sutton et al., 2008;Wienerroither et al., 2009). Recent studies on mesopelagic diversity in the northeastern central Atlantic (or in general) are mostly on a regional scale (e.g. Wienerroither et al. 2009;Ariza et al. 2016;García-Seoane et al. 2020), and variable sampling methods and gaps of information do not allow for statistical comparisons on a larger geographical scale. In spite of this limitation, Sutton et al. (2017) managed to present a global biogeographic classification of the mesopelagic zone reflecting regional variation of biodiversity and function. In our study, fish sampling covers two of these ecoregions: the North Atlantic drift region and the Central North Atlantic ecoregion. The North Atlantic drift ecoregion is a westerly wind biome, with large seasonal changes in mixed layer depth due to high westerly wind stress in winter . It is located in an intergyre zone enclosed by the STG and SPG (Pollard et al., 1996), charterized by large spring algal blooms . The North Atlantic drift ecoregion is an eddy field area, expanding eastwards as the continuation of the Gulf Stream . On the other hand, the Central North Atlantic ecoregion is a trade wind biome with low productivity, weak seasonality, and a persistent deep maximum in chlorophyll . Small amplitude responses to trade with the variability of the wind are typical for this type of biome .
In terms of marine biodiversity, the deep-pelagic ocean is chronically underrepresented in the global databases (Webb et al., Figure 1. Map of the study area showing locations of the CTD stations (circles) and mroplankton trawls (filled circles), and with SST ( C) averaged over the period 2-22 May 2019. The CTD stations are numbered from 120 to 136, and the trawl stations (in parentheses) from 4603 to 4618. Only stations used in the analysis are numbered. Some currents, referred to in the text, are schematically included in the map. The SST data are daily means from satellite data and product of the operational Sea Surfe Temperature and Ice Analysis (OSTIA) system run by the UK Met Office (Donlon et al., 2012). 2010) and basic information on species composition, their distribution and the ftors that affect the community diversity is rare (Mengerink et al., 2014). Considering the importance of deep-sea fish in the biological pump, knowledge of these species is needed for further investigations of the carbon flux. Hence, the aim of this work was to assess the influence of latitude and the oceanography on the meso-and bathypelagic fish assemblages along the Northeastern Atlantic Ocean.

Field sampling
A multidisciplinary cruise was conducted on board the R.V. "Kronprins Haakon" from 2 to 22 May 2019 along the eastern North Atlantic (from Cape Verde to northern France waters) ( Figure 1). Meso-and bathypelagic fish were caught using a mrozooplankton trawl (Krafft et al., 2010;Heino et al., 2011) with a theoretical mouth opening of 6 Â 6 m and a total length of 67 m. The gear consisted of a net of 8-mm stretched mesh and 3 Â 3 mm light-opening from the start to the end. This prevented size-biased escapement from the trawl through differently sized mesh and avoided herding of organisms by larger meshes in the front. Hauls were conducted during daylight hours at ten stations. The net was towed obliquely at a speed around two knots, from the surfe to 1200 m depth, net geometry was recorded continuously. The sorting, subsampling, identification, and length and weight measurements were conducted on board. The total catch was weighed and, in case of a very high number of individuals, subsampled. Prior to subsampling, all large specimens were removed and the whole catch was checked for individuals of uncommon species (i.e. about 1 in 250 individuals). The subsample was randomly selected, weighed, and sorted. Subsample weights were obtained for eh taxon using a motion compensated balance, and these weights were extrapolated to the total catch by multiplying with the raising ftor (calculated by dividing the total catch by the subsample weight). For selecting the subsample size, we targeted having at least 100 individuals per species. Fishes were identified to the highest taxonomic category possible (Whitehead et al., 1984(Whitehead et al., -1986Carpenter and De Angelis, 2016;Sutton et al., 2020). Most specimens were identified to species, however, some small/juvenile or badly damaged specimens could only be identified to genus or family level. Length measurements were mainly to standard length, (some species were measured to total length following Mjanger et al. (2017)), and only undamaged specimens were measured. Validity of scientific names was checked with Fricke et al. (2020) and systematic order is cording to .
Along the transect, 17 hydrographic stations were performed (Figure 1). At eh station, a Conductivity-Temperature-Depth (CTD) profiler cast was conducted down to 1200 m depth using a probe with an SBE 911plus CTD (Sea-Bird Electronics, WA, USA) and 12 water bottles rosette atthed. The CTD carried extra sensors for measuring dissolved oxygen concentration (SBE 43) and fluorescence (Wet Labs ECO-FL). The CTD-data were analysed and visualized with MATLAB software (MathWorks Inc.). At eh station, vertical averages were computed over 0-200, 300-700, and 700-1200 m depths, representing surfe, upper, and intermediate layers, for comparison with fish densities and biomasses.

Data analyses
Fish data were standardized using the filtered volumes (for density: individuals 10 3 m À3 ; for biomass: g wet weight 10 3 m À3 ) to compare the trawls. The volume of water filtered by the trawl was calculated using the vertical and horizontal opening of the trawl net as well as the trawl speed through water. The vertical opening was the distance between depth recorders pled on the headline and the footrope (Starmon TD; Star-Oddi, Gardabaer, Iceland), the horizontal opening was measured with oustic trawl instruments (SCANMAR AS, Å sgårdstand, Norway). A trawl-mounted oustic Doppler current profiler (ADCP; Signature500 in deepwater housing; Nortek AS, Rud, Norway) was used to measure the water flow in 50-cm bins ross the trawl's cross-section. The ADCP was mounted inside the trawl on the footrope and the water flow was averaged over all bins between the trawl netting to calculate the speed of the trawl through the water.
To calculate the sampled volume per minute, the average area of the opening of the trawl net was multiplied by the distance the trawl moved (60Â speed of the trawl). Vessel propeller turbulence limited the use of the ADCP close to the vessel and therefore the sample volumes were calculated when the headline rehed 20 m. The sampled volumes per minute along a single trawl trk were then summed to estimate the total sampled volume of the haul.
Diversity was assessed based on Species Richness (S) calculated for eh haul. Taxa that are not typically meso-or bathypelagic but epipelagic or epibenthic were not excluded from the analyses because we cannot be sure where they were caught (surfe waters or below 200 m depth). The barplot graphs were constructed with R statistical software (R Development Core Team, 2019).
Patterns of the meso-and bathypelagic fish community structure were investigated using PRIMER 7 with PERMANOVA þ Clarke and Gorley, 2015). In these analyses, both density and biomass data were fourth root transformed before analysis to reduce the weighting of dominant species (Clarke et al., 2014). Rare species (i.e. those representing <0.1% of density or <0.5% of biomass) were excluded from the multivariate analyses. Therefore, a total 38 of taxa were employed in the construction of the data matrix for density and 37 in the data matrix for biomass. Bray-Curtis similarities were calculated for eh pair of hauls to produce a similarity matrix. This matrix was classified by non-metric multidimensional scaling (nMDS) with the group average linking method and by cluster analyses. To investigate the latitudinal differences, the SIMPROF procedure (with a significance level of 1% and 9999 permutations) was used to identify significant groups of samples. The SIMPER procedure was applied to identify the key species.
Data exploration was conducted to identify outliers and collinearity following Zuur et al. (2010). One haul with one outlier in fluorescence was discarded from the analyses. The ten environmental variables were checked for collinearity: average temperature, average salinity, and average oxygen (from 0 to 200 m depth, from 300 to 700 m depth, and from 700 to 1200 depth) and average fluorescence in the upper 200 m (Supplementary Figure S1). Collinear variables were removed after inspection of correlation between variables and the variation inflation ftor (VIF > 3) (Zuur et al., 2010). The variable temperature at 0-200 m depth was perfectly collinear with salinity at 0-200 m depth, and with average fluorescence at 0-200 m depth, as well as the temperature at 700-1200 m depth with the salinity at 700-1200 m depth. Thus, these two salinity variables were excluded before conducting VIF analyses. The distance-based linear models (DistLM) and distance-based redundancy analysis (dbRDA) (Legendre and Anderson, 1999;McArdle and Anderson, 2001) were used to study the relationship between the fish assemblages Biogeographical patterns of meso-and bathypelagic fish and the environmental variables selected (temperature at 300-700 m, oxygen at 300-700 m, and oxygen at 700-1200 m, Supplementary Figure S2) and to build the multivariate statistical model. To rank the resulting models, the Akaike information criterion (AICc) was used, with the Step-wise selection criteria and 9999 permutations. AICc is a modification of the AIC and handles better situations where the ratio of samples to predictor variables is relatively small .

Oceanography ross the North Atlantic
The sea surfe temperature (SST) along the cruise trk ranged from 21 C at the southernmost station to 12 C at the northernmost station in the Bay of Biscay (Figure 1). A lower SST ($18 C) along the northwestern coast of Africa, due to coastal upwelling of colder water masses, was clearly observed. The general trend along the transect northwards in the surfe layer (0-200 m) is a decline in both temperature and salinity and an increase in oxygen ( Figure 2). From station 120 to 136, the temperature decreased by 8 C (i.e. from 20 C to 12 C). Highest salinity in the surfe layer (36.8 psu) was observed at station 120. The fluorescence in the surfe layer had a stable lower level south of 40 N (stations 120-131; 0.2 mg/m 3 or less). At stations 132-135, the fluorescence increased to 0.3-0.5 mg/m 3 , and at the northernmost station 136, it was substantially higher (1.2 mg/m 3 ) compared to the other stations ( Figure 2d). In the upper layer, 300-700 m depth, the temperature and salinity varied little along the transect while the oxygen varied most at this depth (3.3 ml/l at station 121 in the south to 5.4 ml/l at station 136 in the north). In the intermediate layer (700-1200 m depth), all three variables temperature, salinity, and oxygen increased with latitude. The plots of temperature and salinity show clearly that the northern part of the section had a weaker vertical stratification compared to the southern part. In the northern part, temperature and salinity showed little differences between the three layers (e.g. the temperatures varied between 10 C and 13 C at stations 133-136) indicating a nearly homogenous water column. This is in contrast to the southern part where the temperature and salinity decreased heavily with depth (e.g. from about 20 C in the upper layer to about 7 C at 700-1200 m depth at station 120).

Fish diversity and distribution ross the North Atlantic
A total of 18 orders and 36 families were recorded for at least 130 taxa (Table 1). The dominant family was Gonostomatidae. Four species of this family (Cyclothone braueri, Cyclothone microdon, Cyclothone pseudopallida, and Cyclothone pallida) were responsible for more than 78% of the total density. Due to the small size of Cyclothone spp., these four species together counted for only 28% to the biomass, with C. microdon dominating the biomass (almost 16% of the total biomass and C. braueri representing only 7%). Large growing specimens like Eurypharynx pelecanoides (ranging from 90-770 mm, Table 1), and Serrivomer beanii (ranging from 115 to 640 mm, Table 1), represent 12 and 5% of the biomass. The most species-rich family was Myctophidae (43 species) followed by the families Stomiidae (15) and Gonostomatidae (11). Among the myctophids, the most abundant species in terms of number were Benthosema glaciale and Lobianchia dofleini, while by weight they were Lampanyctus cuprarius, B. glaciale, and Lampanyctus ater. Valenciennellus tripunctulatus, Argyropelecus hemigymnus, and Sternoptyx spp. were the most numerous Sternoptychidae, whereas in terms of biomass the most important was Maurolicus muelleri. The most frequent species were C. braueri and C. pseudopallida, which appeared in all the hauls. Cyclothone microdon, L. ater, and S. beanii appeared in nine of the ten hauls analysed.
Several species, like A. hemigymnus, C. braueri, C. microdon, C. pseudopallida, Notolychnus valdiviae, and S. beanii, were caught at almost all stations, and thus appear to be ubiquitous (Supplementary Figure S3). Other species, such as E. pelecanoides, the myctophids Benthosema suborbitale, Hygophum hygomii, Hygophum taaningi, L. dofleini, Notoscopelus resplendens, the stomid Chauliodus danae and the sternoptychids V. tripunctulatus and Sternoptyx spp., showed a southern distribution from the coast of Africa to off the Cape S. Vicente, while the myctophid B. glaciale and the sternoptychid M. muelleri where only registered along the Galician coast and the Bay of Biscay (Supplementary Figure S3). Most of the species did not show a clear latitudinal size distribution (Supplementary Figure S4). However, A. hemigymnus, C. pseudopallida, Diogenichthys atlanticus, and N. valdiviae showed increasing sizes in northern and colder waters.
The highest species richness (63 taxa) was found at station 4610 (Figure 3a), which was off the coast of Morocco. The stations along the coast of Africa showed similar species richness, ranging from 48 to 63 taxa. However, the total number of taxa was considerably lower (between 17 and 20) in the northern stations (i.e. Bay of Biscay and off the coast of Galicia, northwestern Spain). Station 4607 showed low-density values (2.8 individuals/ 1000 m 3 ), which is particularly lower than at stations close-by (ranging from 7.1 to 13.1). The overall density from 0 to 1200 m depth ranged from 2.3 (in the Bay of Biscay) to 13.1 (the more coastal station) individuals/1000 m 3 (Figure 3b), whereas the biomass varied from 1.1 (in the Bay of Biscay) to 4.7 (north of the Canary Islands) g wet weight/1000 m 3 (Figure 3c). Both overall density and biomass were variable among stations, but lower biomasses were recorded from station 4613 northwards.  Table 1. Mean (number/1 000 m 3 ) and relative density (in % to the total community), mean (g weight/1 000 m 3 ), and relative biomass (in % to the total community), frequency of occurrence (%FO) and size range of the fish taxa caught at the ten stations (0-1 200 m

Fish community structure
The nMDS plots (both density and biomass) reveal two clusters related to the latitudinal gradient ( Figure 5). Samples are more scattered in the density nMDS and more clustered in the biomass nMDS plot. In terms of density, the most scattered stations in the southern group were the most tropical (4603) and the more inshore station (4608) (in the right and left side of the group, respectively). Stress values were low for both nMDS ( 0.01), being useful two-dimensional ordinations, with no prospect of misinterpretation (Clarke et al., 2014). Cluster analyses (using SIMPROF) separated the samples into two clusters in terms of density (Figure 6a), as well as in terms of biomass (Figure 6b). Again, all these clusters followed the north progression of the stations in the transect, confirming the clusters of the nMDS. In terms of density, there were no significant differences between stations from cluster 1 and the most different station was the most tropical station (4603). The most inshore station (4608) was also separate from the other southern stations. However, in terms of biomass, 4603 and 4605 are similar, and 4607 and 4608 as well.
The size of the individuals collected in the mrozooplankton trawl ranged between ten (Argyropelecus olfersii) and 770 mm (E. pelecanoides) (Table 1). No difference with respect to size range or modal size classes was evident between northern and southern areas for C. braueri and C. microdon. In contrast, some differences were found between north and south areas for C. pseudopallida, with bimodal distributions, but larger individuals being detected in the north (Figure 4). Densities and biomass per area for all taxa are given in detail in Supplementary Table S1.
The SIMPER routine identified the principal species in eh cluster regarding the taxa contribution to the average of similarity of eh cluster. In terms of density, the key species in cluster 1 (stations between 25-37 N) were C. braueri, C. pseudopallida, and C. microdon (counting for 22%) with other species such as V. tripunctulatus, A. hemigymnus, Sternoptyx spp., and C. danae contributing another 13% (Table 2). In cluster 2 (stations from 42-48 N), the rank was different, the key species being C. microdon (counting for 26%), B. glaciale, C. braueri, and C. pseudopallida, contributing another 48%. In terms of biomass, the key species were E. pelecanoides, C. braueri, and C. danae in cluster 1 (counting for 21% of the average similarity), while C. microdon, B. glaciale, and Lampanyctus crocodilus were key in cluster 2 (Table 3). Though the southern cluster 1 showed a higher number of charteristic taxa for both density and biomass data, the taxa contributed less to the similarity. The northern cluster 2 had less taxa, but they contributed more. However, cluster 2 included some species only charteristic of that cluster, such as B. glaciale, M. muelleri, and Xenodermichthys copei.

The relationship between meso-and bathypelagic fish community and environmental parameters
The DistLM marginal test showed that only temperature from 300 to 700 m depth was statistically significant (p < 0.05) and it  Biogeographical patterns of meso-and bathypelagic fish explained 65% of variation in terms of density (Table 4) and 58% in terms of biomass (Table 5). Oxygen at 300 to 700 m explained 39 and 35% in terms of density and biomass, but it is not significant. The model with smaller AICc values included only one variable (the temperature at 300-700 m depth) for both density and biomass data. For both density and biomass data, the dbRDA showed similar patterns (Figure 7). There was discrimination among the latitudes, with the southern stations (25-37 N) associated with higher mean temperatures at 300-700 m and northern stations (42-48 N) associated with lower mean temperatures. The dispersion along the dbRDA2 axis 2, which is related to oxygen at 700-1200 depth, was present in both clusters and larger in terms of density than in biomass.

Fish communities ross the transect
We have identified 42 of the 80 myctophid species reported in the Northeast Atlantic, 11 of the 18 gonostomatid species, and 15 species of the 108 stomiid species reported for the same area (Badcock, 1984;Gibbs, 1984;Hulley, 1984;Sutton et al., 2020). Several studies (Sutton et al., 2010;Olivar et al., 2017) have also reported Myctophidae as the most speciose family of meso-and bathypelagic fish, followed by Stomiidae and Gonostomatidae, pointing to a general pattern in the North Atlantic Ocean. The gonostomatid fishes C. braueri, C. microdon, C. pseudopallida, and C. pallida dominate our samples despite midwater fish trawls not being optimal for the sampling of these smaller sized individuals (Gartner et al., 1989;Olivar et al., 2017). Cyclothone spp. are the dominant taxon in meso-and bathypelagic ecosystems, both in the North Atlantic (Opdal et al., 2008;Sutton et al., 2010;Olivar et al., 2017;Kenchington et al., 2020) and oceans worldwide (Miya and Nemoto, 1987;Collins et al., 2012;Davison et al., 2015). The rank of the Cyclothone species changes with the latitudinal gradient, with the southern area dominated by C. braueri and the northern area by C. microdon, a species charteristic for the northern North Atlantic (Kenchington et al., 2020). In the tropical and equatorial Atlantic, the occurrence and abundance of Cyclothone parapallida have been correlated with zones where AAIW occupies the mesopelagic layers, whereas Cyclothone livida increases in abundance when the AAIW disappears from the mesopelagic layers (Olivar et al., 2017).
A crucial first step towards the effective and sustainable ecosystem management is understanding the distribution of marine biodiversity (Webb et al., 2010). In this study, diversity (expressed as species richness) was lowest at the northern stations. There is no clear lower abundance (in terms of density) in the northern stations, but lower biomasses were recorded from 35 N and northwards. Sutton et al. (2010) also reported higher diversity (expressed as both species richness and Shannon index) in the southern Sargasso Sea, as well as higher abundance. In the area of the Charlie-Gibbs Frture Zone, Cook et al. (2013) found a decrease in diversity (expressed as Shannon index) from southeast to northwest. This corresponds with a similar gradient in SST and underlying water masses, suggesting that water masses can play an important role in the configuration of spatial assemblages of mesopelagic fish (Olivar et al., 2017). In addition, the present eddies could be the mechanism that enhance diversity in the southeast area (Cook et al., 2013), highlighting the importance of hydrographic structures in fish spatial distribution.
The southern stations were dominated by C. braueri, C. microdon, and C. pseudopallida in terms of density and by E. pelecanoides in terms of biomass. The density and biomass of northern stations were dominated by C. microdon, C. braueri, and B. glaciale (Supplementary Figure S3). Community structure in highly productive regions is in general charterized by a few largely dominant species (Andersen et al., 1997). The higher fluorescence values in comparison with the southern stations reflect the higher productivity in the northern area. It is important to note that at the time of the cruise, the high densities of salps observed in the northern region point to a salp bloom in the area, which could be an explanation for lower densities and biomass of fish registered. This pattern of lower diversity in cooler and more productive stations and an increase in diversity in warmer conditions was also reported in the mesopelagic community of the Scotia Sea, Southern Ocean (Collins et al., 2012). This suggests that the latitudinal diversity gradient described for pelagic species (i.e. species richness peaking around the tropic warm and oligotrophic waters and decreasing gradually towards the cold and productive polar waters) (Longhurst et al., 1995;Reygondeau and Dunn, 2019), also applies for meso-and bathypelagic species. Saunders and Tarling (2018) found that Bergmann's rule, i.e. that body size increases with decreasing temperature and increasing latitude, can be applied to the majority of biomass-dominant myctophid species in the Southern Ocean. The transect of our study covered more latitudes and the temperature gradient was lower, but still

Oceanography
The transect covers a range of different water masses and there is a clear distinction between the southern and northern part of the section, i.e. south and north of $38 N. The upper layers, 0-200 and 300-700 m, in the southern part are charterized by warmer, saltier water, and lower oxygen concentrations compared to the northern part. This region is influenced by the Azores current that transports warmer and more saline Atlantic Water (i.e. NW) to the waters off Northwest Africa. The observed low chlorophyll concentration at the surface layer (0-200 m) south of 38 N confirms the known low productivity within the STG (Longhurst et al., 1995). The higher chlorophyll concentration observed in the northern part, with a clear maximum at station 136 lies in a region with a wind-induced deep mixed layer during winter that exhibits large spring blooms . The observed weak vertical stratification with narrow ranged temperature and salinity values in this region is in contrast with the more stratified waters in the southern part that had much larger spans in temperature and salinity that also increased southward. The SST shows the upwelling effect close to the Northwest African continent, but the known upwelling effect on the production (e.g. Longhurst et al. 1995) is not observed in chlorophyll concentrations at the stations. This is most likely because the southern stations are too far away from the upwelling region.
The influence of the North Atlantic OMZ in the region is observed at 300-700 and 700-1200 m depth in the southern part of the section, particular south of 28 N (including fish station 4603). These concentrations are low enough to function as hypoxic conditions for several fish taxa (Vaquer-Sunyer and Duarte, 2008). The presence of the MOW is observed in the high salinity at 700-1200 m depth from station 126 (32 N) and northwards. MOW is also charterized by relatively high oxygen values and low nutrient concentrations, due to its origin in the oligotrophic Mediterranean Sea (Howe et al., 1974).

Biogeography in relation to oceanography
The two distinct clusters found in this study are largely in agreement with the two ecoregions defined by Sutton et al. (2017) in the same area. The two northernmost stations of the northern cluster are located in the North Atlantic drift ecoregion while the third is in the Central North Atlantic ecoregion. Note that ecoregion boundaries must be Table 2. Average similarity obtained with SIMPER analyses for eh cluster from fourth root density data, with indication of taxa that contributed for a 95% cut-off. Clusters identified by the SIMPROF routine in PRIMER (see Figure 6a). Clusters identified by the SIMPROF routine in PRIMER (see Figure 6b).
regarded as transitional zones and not as abrupt borders, and seasonal changes are not considered in the classification of Sutton et al. (2017). Altimeter data showed that the positions and number of branches of the North Atlantic Drift tend to shift from time to time as a consequence of long-lasting mesoscale eddies (Read et al., 2010). The North Atlantic drift ecoregion is a transition between more boreal (e.g. B. glaciale, Lampanyctus macdonaldi, and Protomyctophum arcticum) and more subtropical species (e.g. Paralepis brevirostris and Paralepis elongata). All the stations belonging to the southern area correspond with the Central North Atlantic ecoregion . Along the northwest of Africa, high chlorophyll concentrations were expected due to upwelling effects. However, the observed chlorophyll concentrations were low in this area. This was most likely because the cruise transect was too far away from the African continent as discussed earlier.
In this work, the relationship between the fish assemblages and the temperature from 300 to 700 m depth was statistically significant. The latitudinal gradient in diversity is related to the temperature, which plays a key role in faunal distribution patterns . Each fish species has a thermal preference that optimizes physiological processes (Selleslagh and Amara, 2008). However, other environmental variables correlated with temperature at 300-700 m might be responsible for the changes in the fish community structure along the latitudinal gradient as well. Such correlated variables were not included in the models. If collinearity is ignored, it could result in misleading statistics without any significance and where dropping one covariate can make others significant or even change the sign of estimated parameters (Zuur et al., 2010). For example, the fluorescence, which showed high collinearity with temperature at 300 to 700 m, is a proxy of primary production in surface waters, and primary production is a demonstrated driver of mesopelagic community biogeography .

Methodological aspects
The macrozooplankton trawl used in this study showed bias against capturing large fish and the lower size range of some species. For example, small species such as C. braueri, H. hygomii, or N. valdiviae differed only little from the maximum sizes reported in the literature, respectively, 38, 68, and 25 mm (Sutton et al.   Figure 7. Distance-based redundancy analyses (dbRDA) to visualize the distance-based linear models (DISTLMs) in terms of (a) density and (b) biomass of the environmental variables in the study area. T300_700 m ¼ average temperature from 300 to 700 m depth. O300_700m ¼ average oxygen from 300 to 700 m. O700_1200m ¼ average oxygen from 700 to 1200 m. 2020), whereas larger species showed a bias towards the lower end of their size range, e.g. Chauliodus sloani (350 mm), L. crocodilus (172 mm), or Sigmops elongatus (275 mm) (Sutton et al., 2020). The reasons of that bias could be avoidance of the large and fast swimming fish when trawling at low speeds (1 ms À1 ; Gartner et al., 1989;Kaartvedt et al., 2012) and the relatively small mouth area of the net ($35 m 2 ). Few individuals in the lower size range of Cyclothone spp. were caught with the 3 Â 3 light-opening meshes, which agrees with previous studies (Gartner et al., 1989;Olivar et al., 2017). Gartner et al. (1989) showed that midwater trawls with mesh size of >2 mm underestimate the lower size range (fish smaller than 30mm SL). The fact that Olivar et al. (2017) reported significantly higher sampling efficiency with plankton nets than with a midwater trawl with graded-mesh from 30 mm in the trawl opening to 4 mm in the cod-end could be explained by the larger meshes at the front of that trawl. The gear used in this study had the same mesh size throughout the trawl and therefore the selectivity of the trawl was constant.
In conclusion, the structure of fish assemblages in the Northeastern Atlantic changes along the latitudinal gradient. Fish community in the southern area was more diverse than in the northern region and temperature seems to be an important factor in species distribution. Although many species are present in both areas, abundance, rank and, though on a small scale, sizes changed. Overall, this study improves our understanding on the variation in deep-pelagic ecosystems on a large spatial scale and given its value in the provision of ecosystem services, this information is essential to the ecosystem management approach and marine spatial planning.

Supplementary data
Supplementary material is available at the ICESJMS online version of the manuscript.

Data availability
Data are available on request.