Sampling gear and protocol

M ac rofauna were collected from 193 station s in the Estuary and thc northern Gulf of St. Lawrence (EGSL) during the an nuai summer gro undfish survey made by th c Canadi an Department of Fishcries and Oceans (Q uebec Region), aboard the CCGS Te/eost rcsearch trawler from August 1 st to 31 st, 2006 . The sa mpi ing strategy used cons isted of a stratified random sa mpling followin g predetermined strata bascd on depth (Doubleda y, 198 1) Samp les were co ll ccted with a four- sided shrimp bottom trawl (Campe/en 1800 type). It was rigged with variable net mesh sizes appropri ate for each part of the trawl: 80 mm (« center Imo t » to « center knot ») for the wings, 60 mm for the first bell y and the square, and 44 mm for the second and third bellies. The codend and th e lengthening picce are a lso 44 mm stretched mesh size and are equipped w ith a 12. 7 mm knotless nylon lining. T raw l is titted with a Rockhopper footgear (M cCa llum and W alsh, 2002) (A ppendi x 1). T hc standard tow duration was 15 minutes on the bottom , being shortcr dcpending on the roughness of the substrata. Scanmar™ hydroaco ustic sensors monitor trawl characteristics configuration (e.g. distance between do ors and w ings, vert ica l nct opcnin g and bottom depth).

Fish and invertebrates were sorted, identifi ed, counted, and weighed. A li were identified to th e lowest taxonomic level poss ibl e for which identification is ccrtain (see list Appendix 2). Invertebrate spec ies that were not easi ly identifi able were preserved in 70% ethanol or fro zen for later identification in laboratory. Taxonomic names we re verified on the Integrated Taxonomic Information System on lin e (www.ITIS. gov) . Density estimates for ail identified macrofauna taxa were obtained by dividing the number or mass by the total area swept by the trawl, which de fines the effort. Catch per unit effort (CPUE) was used as standardized abundance indi ces. Biomass density indices in kg/km2 were used for many taxa that were too abunda nt to be counted. Co lonial organisms such as Bryozoa and Hydrozoa could not be enumerated and were expressed in kg/km2 ln the database. The taxa Cnidari a and Ech inodermata , predominantl y composed of Brisaster fragilis , Ctenodisicus crispatus , Gorgonocephalus sp. and Strongylocentrotus sp. , were also included in thi s database. The other taxa were expressed as ind/km2 (abundance database) . This, resu lt in two different databases; one expressed as biomass and the other one expressed in abundance, these two databases do not consider the same species . However a third one using occurrence (PIA) data contained ail th e surveyed spec ies.

Stati stica l ana lyses

Data were analysed by multivariate approach uSlll g the VS PRIMER Analytica l Package (C larke and Warwick, 1994). The multi variate procedure on specl es assemblages was based on the Bray-Curti s di ss imil arity on sq uare-rool (-V) data transformed and presence/absence (P/A) community data . This in termed iate transformation was chosen in order to provide the best balance between a  » narrow view » of community structure based on abundance of few dominant taxa and a « wide view » based on ail species, g iving too much weight on rarest taxa (Clarke and Warwick, 1994). Prior to analyses, taxa which appeared once or were associated with o nly one station were excluded from the dissimilarity matI’ix, as suggestcd by C larkc and Warwick (1994), Differences in the structure of epibenthic macrofaunal assemb lages along thc Estuary and the Gu lf of St. Lawrence were examined using non-metric multidimen sional scaling (nMDS) ordination technique. Ordinations were based on the Bray-Curti s dissimilarity measure (Bray and C urti s, 1957). The SIMPER routine (s imil arity percentage analysis) was used to determin e which species were predominantl y responsible for the dissimilarity within groups, and to characteri ze each assemblage (Clarke, 1993). Multivariate ordination techniques were also applied to analyse th e spatial variation in the species abundance data sets, usi ng CANOCO 4.5 program (ter Braak and Smilaeur, 2002).

Canonical correspondence analyses (CCA) were calculated on: 1) biomass database, 2) abundance database a nd on 3) a database included both biomass and abundance data, to explore the relationship between the observed Illacrofauna assemblages and their environment (CANOCO, ter Braak and Smilaeur, 2002). A preliminary detrented correspondence analysis (DCA with detrending by segments) was applied to estimate the gradient length in standard deviation (SD) units. Gradient va lues exceeding 3 SD were obtained in each case, and thcn subsequent numerical analyses involved techniques were based on underlying unimodal species-response model (Jongman et al., 1995; Leps and Smilauer, 2003). Statistical associations between species macrofaunal assemblage patterns and env ironmental parameters were further quantified by canonical correspondence anal ys is (CCA), a non-lincar cigenvector ordination technique related to CA, but where the axes correspond ing to the direction s of the greatest data set va ri a bility can be ex plained by the cnvironmental va ri ab lcs (tcr 8raak, 1986). .

Macrobenthi c cOl11muni ty structure C lus te r ana lysis, based on squ are root trans form ed da ta, d iv ided the b iomass fa un a database (kg/kIl12 ) into seven di stinc t communities. Most stati ons we rc gro uped ( fi ve gro ups) at 70% of di ssimil arity a nd two oth ers at 55% . Multi variate ll1ac ro in ve rtebra te assembl ages showed two ma in di stin ct gro ups with a pa rti c ul ar spec ies di stributi on. A distinct cOll1l11unity (gro up V: .J) was spati ally located w ith maj ori ty o f statio ns in th e Es tua ry . G roup IV ( .& ) also showed a distinct w ide spectrum o f assembl age sca lte red over the Laurentian and Esquiman channels. Community of Groups l, II, III, V I and V II , sma ll er than Groups V and IV, were punctually distributed and closer spati a ll y, over the study area (F ig. 3). G roup l (0 ) indicated a specifie di stributi on di vided in two ma in spots, in the Strait o f Bell e Isle and near Mingan Isla nd . Thc greatest di ss ill1il arity between gro ups (S IMP ER analys is) was observed w ith Gro up V rela ti ve to ail other groups, a nd was c haracteri zed by the hi gh contributi o n of Acti ni a ri a, Ctenodiscus crispatus and Brisaster frag ilis. Scyphozoa and fllex sp. were th e macrofauna taxa characte ri z ing g roup IV ( À ) (Tab le 1). C lu ster ana lys is based on the abundan ce database (ind/kJn2 ), sugges ted th e ftrst ftve groups a t 95% of dissimilarity with another two g roups (II: • a nd III: T ) di stinct at 90%. Groups r (0 ) and II ( . ) were the small est, based on two stati ons. Gro up 1 was located 0 11 th e south wes tern pa rt of th e Newfo undland w he reas Group Il was rou nd a t the head of the Laurentian Cha11l1el. Grou p V (0 ) was di stributed from the Estuary throu gh the Ca bot Strait (F ig . 4).

SIMPER analysis revealed that a hi g h co ntributi on of similarity frol11 a particular species was fo und in each group. For example, Hippasteria phrygianna contributed 63,06% in Group IV , while Syscenus in[efix was 73 ,77% in Group V (Tab le 2). W ith rega rds to the occurrence database (PIA), the ft ve groups we re we ll-di spersed over the study a rea (Fig. 5). We a lso observed the sa me cOl11munity compos iti on (Group IV: ), using the observed abundance database, ra ng ing from the Lau rentian C hann el out to the Atla ntic Ocean beginning at Cabot Strait (Fig. 5). The same aggregated group near the Strait of Belle Isle and th e Minga n Islands was also founded (Gro up I: 0 ), but stations of western part of Newfoundland were also ineluded in thi s group . The results of SIM PER analys is showed a simil ar average fo r th e ri ve clusters, with a sim ilar contribution among th c spec ies (Tabl e 3). Fu thermore, th c three nMO S pl ots (occurence, biomass and de nsity) showed similar community distri buti on over th e study area. These are: ( 1) th e same aggregated group was fo und nea r Minga n Island and SU’a it of Belle-Isle; (2) the sa mc communi ty compos ition ra nging from the Laurentian Chann el out to the Atlanti c Ocea n beginning at Cabot Strait, and fin ally; (3) a mi x of community assembl ages was fo un d at two spec ific loca ti ons at the northeas lern and th e north wes lern end of Anticosti Island.

Pattern analysis

To identify environmental parameters respo nsible for macrobenthi c community distribution from set of eco logica l data, the first step was to identify pattcrn s of thc spec ies abundance. As preliminary steps, gradients lengths were explored at constrained canonical analys is to estimate th e standard deviatio n. Detrended correspondence analys is (DCA), detrending by segments, gives values between 3.28 to 7.24 SD , and therefore unimodal methods of ordination were used for subsequent analyses. DCA was appli ed to macrobenthic databases (abundance, biomass and occurence) and a lso to SIMPER spccies ab undance database which are taxa that account for 15% and more to the similari ty w ithin the groups. Incidenta lly, canonical correlati on (CCA), combined w ith numerical analysis, can revea l the eco log ica l preference of spec ies coloni z ing th e study habitat. Cano nica l correspondence anal yses (CCA) with Monte Carlo permutation and the fo rward se lection opti on were performed to test the signifi ca nce of the re lationships of the sampl es and spec ies to the available environmenta l parameters. The inter-samp le and Hill’s sca ling were choosen for the CCA on the abundancc data sets. Hill ‘s sca ling is more appropriated to a strong unimodal response for long gradi ent. In summary, species which occurred at the station li e around the sample’s point in the plot (Leps and Smilauer, 2003). When exa mll1l11g the assemblages of the biomass abundance database, a CCA showed s ignifi cant relationship (p < 0.01) with cight environmental parameters.

T hese are: oxygen sa turatio n, depth , maximal bottom current, and five sediment types (pelite, very sa nd y-pe lite, g ravel-shell , gravelly-pelite-sand and grave lly-sand y-pelite). First princ ipa l CCA axis accounted for 39.60% and 28.30% for th e second axis, these together accounted for 67.90% of the relationship between species and environmental parameters (Table 4a). The presence of gravel and pelite appeared an important factor on the first axis, with hi gh correlation va lues (0.42) with gravelly-pelite-sand and gravel-shell . On the second axis, very sa ndy pelite substratum was the highest positively correlated facto r (0.36), whereas fiv e of eight environmental variables were negatively correlated . The group V (0) was primaril y influ enced by three factors: depth , presence of pe lite (P), and very sandy pelite (VSP) sediment (Fig.6). Group l (0 ), di stributed on the upper right corner of th e CCA gra ph , appeared to be impacted by hig h oxygen saturation and presence of grave l (gra ve l-shell (GSh) and gravelly-pe lite-sa nd (GPS)). Group IV ( A. ), w ith the greatest number of stations, was influenced principall Y by medium to hi gh oxygen saturation value. This group was also impacted by bottom current (F ig. 6) . Groups l (0 ) and IV ( .. ), found on the ri ght side of the ordination, indicated distinctive mac robenthos composition in comparison with Group V (0) principally located on the left side.