Boreogadus Saida Et Arctogadus Glacialis Vie Larvaire Et Juvénile De Deux Gadidés Se Partageant L’Océan Arctique

Home , Arctogadus, Boreogadus saida, Navaga

Boreogadus saida et Arctogadus glacialis Vie larvaire et juvénile de deux gadidés se partageant l’océan Arctique

Thèse

Caroline Bouchard

Doctorat interuniversitaire en océanographie Philosophiae doctor (Ph.D.)

Québec, Canada

Résumé

Le très abondant Boreogadus saida occupe au sein de l’écosystème marin arctique une place prépondérante, ce qui lui vaut une attention croissante des scientifiques. Arctogadus glacialis, commun dans toutes les mers arctiques, est beaucoup moins étudié. Les deux espèces et leurs jeunes stades cohabitent mais ces derniers sont pratiquement impossibles à différencier. Seuls des outils génétiques, ou une méthode utilisant la taille du noyau de l’otolithe développée dans cette thèse, peuvent distinguer les deux espèces. Ces méthodes d’identification ont permis d’étudier pour la première fois l’écologie des jeunes stades d’Arctogadus et d’estimer la proportion de cette espèce dans des échantillons de gadidés arctiques. À la lumière des observations faites en mer de Beaufort, il apparait que les jeunes Arctogadus ont une abondance environ vingt fois moindre, une taille à l’éclosion supérieure, un taux de croissance similaire, et un taux de mortalité inférieur aux jeunes Boreogadus. Pour Boreogadus, l’hypothèse selon laquelle certaines larves éclosent en hiver près des panaches des fleuves, a été testé, d’abord en comparant la saison d’éclosion dans six régions de l’océan Arctique caractérisées par différents apports d’eau douce. Conformément à cette hypothèse, l’éclosion commence en hiver dans les mers recevant de forts apports fluviaux alors que l’éclosion débute au printemps dans les régions aux apports d’eau douce limités. Les larves qui éclosent en hiver profitent d’une longue saison de croissance leur permettant d’atteindre des tailles pré-hivernales largement supérieures aux larves qui éclosent en été, ce qui favoriserait leur survie. Cette même hypothèse a ensuite été testée en comparant la composition chimique des otolithes de Boreogadus provenant de ces six régions, et les différences observées semblent appuyer l’hypothèse. Les tendances actuelles au devancement de la débâcle, au réchauffement des eaux de surface et à l’augmentation du débit des fleuves pourraient favoriser le recrutement de Boreogadus, et possiblement aussi celui d’Arctogadus. Découle de cette thèse une connaissance accrue de l’écologie de gadidés habitant un océan confronté à une pléthore de changements.

iii

Abstract

The very abundant polar cod (Boreogadus saida) plays a preponderant role in the Arctic marine ecosystem and consequently has received significant attention in recent years. The ice cod (Arctogadus glacialis), a common species in all Arctic seas, is much less studied. Both species co-occur on Arctic continental shelves and their early life stages are often found together in ichthyoplanktonic collections. However, larvae and juveniles of polar cod and ice cod are almost impossible to differentiate. Only genetic tools, or a method using the size of the otolith nucleus developed in this thesis, can distinguish the two species. These identification methods allowed to study for the first time ice cod early life stage ecology and estimate the proportion of this species in Arctic gadids samples. In light of observations made in the Beaufort Sea, it seems that young ice cod are about twenty times less abundant, hatch at a larger size, grow at the same rate, and have a mortality rate inferior to young polar cod. For polar cod, the hypothesis that some larvae hatch in winter near river plumes, was tested, first by comparing the hatching season in six regions of the Arctic characterized by different freshwater inputs. Consistent with this hypothesis, hatching starts in winter in seas receiving large river discharge while hatching starts in spring in regions with limited freshwater inputs. The larvae hatched in winter benefit from a long growth season allowing them to reach larger pre-winter size than larvae hacth in summer, a condition that likely favors their survival. This same hypothesis was further tested by comparing the otolith chemistry of polar cod juveniles from those six regions, and the differences observed seem to support the hypothesis. On-going trends of earlier ice break-up, warmer surface layer, and increased river discharge could favor polar cod, and possibly also ice cod, recruitment. Arise from this thesis an increased knowledge of the ecology of gadids living in an Ocean facing a plethora of changes.

Table des matières

Résumé ...... iii

Abstract ...... v

Table des matières ...... vii

Liste des tableaux ...... xi

Liste des figures ...... xiii

Remerciements ...... xxiii

Avant-Propos ...... xxv

Chapitre 1 – Introduction générale ...... 1 1.1 Les écosystèmes marins arctiques en changement ...... 1 1.2 Boreogadus saida ...... 5 1.3 Arctogadus glacialis ...... 8 1.4 La survie larvaire : importance de l’eau douce ...... 11 1.5 Objectifs ...... 16 1.6 Approches utilisées ...... 17 1.6.1 Microstructure des otolithes...... 18 1.6.2 Chimie des otolithes...... 19 1.6.3 Génétique des populations ...... 22

Chapitre 2 – Circum-arctic comparison of the hatching season of polar cod Boreogadus saida: A test of the freshwater winter refuge hypothesis ...... 25 2.1 Résumé ...... 25 2.2 Abstract ...... 26 2.3 Introduction ...... 27 2.4 Materials and methods ...... 28 2.4.1 Study areas ...... 28 2.4.2 Sampling of fish larvae and juveniles ...... 32 2.4.3 Discriminating Boreogadus saida and Arctogadus glacialis larvae and juveniles ...... 34 2.4.4 Validation of the ageing of young polar cod ...... 36 2.4.5 Regional Hatch-date frequency distribution (HFD) of polar cod ...... 37 2.4.6 Regional long-term average surface salinity and temperature ...... 39 2.5 Results ...... 39

vii

2.5.1 Boreogadus saida versus Arctogadus glacialis ...... 39 2.5.2 Validation of the daily deposition of increments in the otoliths of polar cod ..... 40 2.5.3 Length-age relationships and growth rates...... 45 2.5.4 Interannual variability in regional hatch-date frequency distributions ...... 45 2.5.5 Average hatching season in relation to regional river input...... 48 2.5.6 Pre-winter size ...... 48 2.6 Discussion ...... 53 2.6.1 The true age of the true polar cod ...... 53 2.6.2 The winter thermal refuge hypothesis ...... 54 2.6.3 Winter hatching and the pre-winter size of polar cod juveniles ...... 55 2.6.4 The food of polar cod larvae under the ice in winter ...... 57 2.6.5 Climate change and the hatching season of polar cod ...... 58 2.7 Conclusion ...... 59

Chapitre 3 – Spatial segregation, dispersion and migration in early stages of polar cod Boreogadus saida revealed by otolith chemistry ...... 61 3.1 Résumé ...... 61 3.2 Abstract ...... 62 3.3 Introduction ...... 63 3.4 Materials and methods ...... 64 3.4.1 Study areas ...... 64 3.4.2 Sampling...... 65 3.4.3 Otolith preparation ...... 65 3.4.4 Otolith analysis ...... 69 3.4.5 Data analysis ...... 69 3.5 Results ...... 71 3.5.1 Relationships between elemental concentrations, salinity, temperature, hatch date and capture date ...... 71 3.5.2 Differences in otolith chemistry between regions with high and low freshwater input ...... 71 3.5.3 Differences in otolith chemistry among regions ...... 72 3.5.4 Differences in otolith chemistry among otolith zones ...... 72 3.6 Discussion ...... 83 3.6.1 Effects of salinity and temperature on element incorporation in polar cod otoliths ...... 83 3.6.2 Polar cod larval ecology: a further test of the freshwater winter refuge hypothesis ...... 83 3.6.3 Spatial segregation, dispersion and migration in polar cod early stages ...... 84 3.6.4 Is ontogenetic vertical migration of polar cod juveniles reflected in otolith chemistry? ...... 85 3.7 Conclusion ...... 87

viii

Chapitre 4 – The nucleus of the lapillar otolith discriminates the early life stages of Boreogadus saida and Arctogadus glacialis ...... 89 4.1 Résumé ...... 89 4.2 Abstract ...... 90 4.3 Introduction ...... 91 4.4 Materials and Methods ...... 91 4.4.1 Larval fish sampling ...... 91 4.4.2 Species determination based on genetic analyses ...... 92 4.4.3 Otolith Analysis ...... 95 4.5 Results ...... 95 4.5.1 Molecular identification of the two species ...... 95 4.5.2 Classification of genetically identified fish using the area of the lapillus nucleus ...... 96 4.6 Discussion ...... 101

Chapitre 5 – Compared early life history of sympatric polar cod Boreogadus saida and ice cod Arctogadus glacialis in southeastern Beaufort Sea ...... 105 5.1 Résumé ...... 105 5.2 Abstract ...... 106 5.3 Introduction ...... 107 5.4 Materials and Methods ...... 108 5.4.1 Study area ...... 108 5.4.2 Ichthyoplankton sampling...... 109 5.4.3 Morphometric measurements ...... 111 5.4.4 Species determination ...... 119 5.4.5 Age determination...... 119 5.4.6 Redistribution of length, species and age ...... 120 5.4.7 Hatch date frequency distributions ...... 121 5.4.8 Gut content analysis ...... 122 5.5 Results ...... 122 5.5.1 Compared occurrence, length, and growth of Arctogadus glacialis and Boreogadus saida ...... 122 5.5.2 Vertical distribution underice and in ice-free waters ...... 123 5.5.3 Monthly distribution of Arctogadus glacialis and Boreogadus saida in partially ice-covered southeast Beaufort Sea ...... 123 5.5.4 Hatchdate frequency distributions of Boreogadus saida and Arctogadus glacialis ...... 124 5.5.5 Prey and carbon intake ...... 125 5.6 Discussion ...... 135 5.6.1 Discriminating young Boreogadus saida and Arctogadus glacialis in large plankton collections ...... 135

5.6.2 The vertical distribution and feeding of Boreogadus saida and Arctogadus glacialis under the ice and in ice free waters ...... 136 5.6.3 Spatio-temporal sympatry of the planktonic stages of Boreogadus saida and Arctogadus glacialis in southeast Beaufort Sea ...... 138 5.6.4 Ecological divergences and niche separation during life in the plankton ...... 140 5.6.5 Conclusion ...... 141

Chapitre 6 – Conclusion générale ...... 143 6.1 Hypothèse du refuge thermique hivernal ...... 143 6.2 Le vrai âge du vrai Boreogadus saida ...... 144 6.3 Quel futur pour les gadidés arctiques? ...... 145 6.4 Perspectives de recherche ...... 147

Bibliographie ...... 149

Liste des tableaux

Table 2.1 Details of juvenile cod sampling by region and year, and sources of primary data...... 33

Table 2.2 Number of young polar cod captured, measured fresh, and aged by otolith analysis, as well as parameters of the regression of length on age by region. The slope of the length-age regression is an estimate of growth in mm d-1...... 35

Table 2.3 Number and percentage of Arctogadus glacialis in sub-sets of the young cod collected in different regions and years, based on the analysis of the microsatellite marker Gmo8 (Madsen et al. 2009). The number of fish analyzed and the number successfully amplified are given for each sub-set. All other fish in the sub-sets were positively identified as Boreogadus saida...... 41

Table 2.4 Statistics of the hatching season of polar cod by region and year. SD: standard deviation...... 50

Table 3.1 Region, year, sampling code, size range (standard length, SL), capture and hatch date range, age range, number of stations (stn) and number of otolith core (C), middle (M) and edge (E) zones successfully analysed by laser ablation ICP-MS...... 67

Table 3.2 Percentage of juveniles classified to each region by quadratic discriminant function analyses based on multi-elemental composition (Li, Mg, Mn, Sr, Ba) of otolith core, middle and edge zones for collection years 2005 and 2006 using jacknife leave-one-out crossvalidation. Correct classification percentage are in bold. See Table 3.1 for sampling codes...... 80

Table 4.1 Number, sampling period, mean standard length ± standard deviation (and range), and mean lapillus nucleus area index ± standard deviation (and range) of Boreogadus saida and Arctogadus glacialis collected from 2002 to 2008 in the Beaufort Sea and identified by molecular genetics...... 93

Table 4.2 Number of young gadids collected in 2004 and 2008 and classified to each species by the logistic regression (numerator) and by molecular identification (denominator), for the 295 fish analyzed for 18 microsatellites, the 228 fish analyzed for Gmo8 only, and for the total (pooled 523 fish). The corresponding percentages are given in parentheses...... 100

Table 5.1 Number of gadid collected by different sampling methods in 2004 and 2008, with average standard length L (± 1 standard deviation, SD) and percentage of Arctogadus glacialis...... 113

Table 5.2 Total number of gadids collected, number and percentage of gadids identified to species by genetics and/or otolithometry, by sampling month in 2004 and 2008 ...... 114

Table 5.3 Summary of references on relationships between carbon content (C, in mg), prosome (PL, in mm) or total length (L, in mm) for the main prey of Boreogadus saida and Arctogadus glacialis early life stages...... 117

Table 5.4 Number and standard length SL (mean ± 1 standard deviation, SD) of Boreogadus saida and Arctogadus glacialis collected by sampling months in 2004 and 2008...... 126

Table 5.5 Size at hatch and somatic growth rate of Boreogadus saida and Arctogadus glacialis collected in 2004 and 2008 as estimated respectively by the intercept and the slope of the regression of standard length against age for fish aged by otolithometry and identified to the species by genetics and/or otolithometry...... 127

xii

Liste des figures

Figure 1.1 Photos de Boreogadus saida et d’Arctogadus glacialis larvaires après conservation dans l’éthanol. Une pigmentation présentant des ramifications est notable sur la surface ventrolatérale de l’estomac chez Arctogadus. Photos Keita Suzuki...... 10

Figure 1.2 L’Arctique compte plusieurs des plus grands fleuves du monde. Cette carte montre les principaux systèmes fluviaux de l’Arctique et leur débit annuel en km3, ainsi que le bassin hydrographique de l’océan Arctique. Les systèmes illustrés sont ceux des fleuves Mackenzie, Yukon, Nelson, Kolyma, Indiguirka, Léna, Kotya, Ienisseï, Ob, Pechora et Dvina septentrionale. Source: CAFF 2001 (reproduit avec la permission de UNEP/GRID-Arendal)...... 15

Figure 2.1 Bathymetric map of the Arctic Ocean indicating the six regions studied (top). Panels a–f present the long-term average surface salinity during the hatching season of polar cod (months given) for regions characterized by strong (a–c) or weak river discharge (d–f). Surface salinities were extracted from the World Ocean Atlas. Symbols give the location and year of sampling of polar cod larvae and juveniles in each region: black: 1993, 2005; white: 2003, 2006; blue: 1998; and red: 2007...... 31

Figure 2.2 Regression of the number of increments counted between the oxytetracycline hydrochloride mark and the edge of the otolith against the number of days of life after the marking of the otolith (increments = 1.062 d – 0.445, r2 = 0.995, n = 11, p < 0.0001), for polar cod larvae and juveniles collected in the North Water and reared on board in 1999. The dashed line is the 1:1 line. Filled circles indicate two identical data points...... 42

Figure 2.3 (a) Composite photograph comparing the left lapillus of a 98-d old juvenile polar cod seen in scanning electron microscopy (left, taken at 400×) and the right lapillus of the same fish in light microscopy (right, taken at 400×). (b) Composite photograph comparing the nuclear region of the left lapillus of a 206-d old juvenile polar cod seen in scanning electron microscopy (left, taken at 1000×) and the right lapillus of the same fish in light microscopy (right, taken at 1000×)...... 43

Figure 2.4 Relationship between increment counts in the lapilli of polar cod analyzed under the light microscope (LM) in 2006 and 2009 and re-analyzed under the scanning electron microscope (SEM) in 2010 by the same operator (LM = 1.014 SEM – 0.027, r2 = 0.986, n = 13, p < 0.0001). The dashed line is the 1:1 line...... 44

Figure 2.5 Regression of young polar cod growth rate and long-term average surface temperature during the hatching season in a given region (growth = 0.032 temperature + 0.236, r2 = 0.720, p = 0.033, n = 6). Surface temperatures were extracted from the World Ocean Atlas for the months of polar cod hatching...... 46

xiii

Figure 2.6 Hatch date frequency distributions and mean hatch-dates (HD) of polar cod in four regions of the Arctic Ocean for which multiple years of data are available. The number of fish hatched in the same 7-d calendar interval in a given year is expressed as a percentage of the total number of fish sampled in that year...... 47

Figure 2.7 Average hatch date frequency distribution and mean hatch-date (HD) of polar cod in six regions of the Arctic Ocean ordered by decreasing freshwater input. The monthly salinities in the 0–10 m layer were extracted from the World Ocean Atlas. Total river discharge values are from Gordeev (2006) for the Laptev and Beaufort seas and from Déry et al. (2005) for Hudson Bay...... 49

Figure 2.8 Regression of standard length on 14 August (SL) on hatch-date (HD) for polar cod collected in late-summer and early fall in regions characterized by strong river discharge (full circles, Laptev Sea, Hudson Bay and Beaufort Sea) or weak river discharge (open circles, Northwest Passage and Baffin Bay). SL = 49.87 - 0.203 HD, r2 = 0.906, n = 1075, p < 0.0001...... 51

Figure 2.9 The frequency distribution of standard length on 14 August for polar cod sampled in late-summer/early fall in different regions of the Arctic Ocean. Standard length on 14 August was back-calculated from otolith microstructure. Mean standard length on 14 August differed significantly among groups A–D (Tukey–Kramer test). The growth rate of polar cod (GR: in mm d-1) in each region is also indicated...... 52

Figure 3.1 Bathymetric map of the Arctic Ocean indicating sampling locations of juvenile polar cod in 2005-2006. AG: Amundsen Gulf, BB: Baffin Bay, FB: Frobisher Bay, HB: Hudson Bay, LC: Lancaster Sound, LS: Laptev Sea...... 66

Figure 3.2 Photographs of polar cod otoliths seen in light microscopy (taken at 200×) after laser ablation. Yellow areas delimitate the ablation areas. (a) Small otoliths were analysed in two zones (core and edge). (b) Larger otoliths were analysed in three zones (core, middle and edge). The middle zone of the otolith shown in (b) has not been ablated yet...... 68

Figure 3.3 Elemental ratios in otolith edge in relation to salinity and temperature. Solid lines indicate significant regressions. Open symbols represent regions with weak freshwater input of Baffin Bay (cercles), Frobisher Bay (squares) and Lancaster Sound for sampling year 2005 (triangles) and 2006 (diamonds). Closed symbols represents regions with high freshwater input of Amundsen Gulf for sampling year 2005 (cercles) and 2006 (squares), Hudson Bay (triangles) and Laptev Sea (diamonds)...... 75

Figure 3.4 Elemental ratios in otolith core in relation to hatch date (left panels) and elemental ratios in otolith edge in relation to capture date (right panels). Open symbols represent regions with weak freshwater input of Baffin Bay (cercles), Frobisher Bay (squares) and Lancaster Sound for sampling year 2005 (triangles) and 2006 (diamonds). Closed symbols represents regions with high freshwater input of Amundsen Gulf for sampling year 2005 (cercles) and 2006 (squares), Hudson Bay (triangles) and Laptev Sea (diamonds). Lines indicate significant regressions for xiv

regions with weak freshwater input (long dash), regions with high freshwater input (short dash), and all regions grouped (solid lines)...... 77

Figure 3.5 Mean ± SE elemental ratio (µmol/mol) in core, middle and edge zones of otoliths from regions with low and high freshwater input. Asterisk above significantly higher values...... 78

Figure 3.6 Quadratic discriminant function analyses based on multiple elemental ratios (Li/Ca, Mg/Ca, Mn/Ca, Sr/Ca, Ba/Ca) in otolith core (top panels), middle (middle panels) and edge (bottom panels) zones of juvenile polar cod collected in 2005 (left panels) and 2006 (right panels) in six regions (see Table 3.1 for sampling codes). Ellipses indicate 95% confidence intervals. Element loadings onto the canonical variables are shown in the upper right of each panel...... 79

Figure 3.7 Elemental ratio in core, middle and edge zones of all otoliths analysed. Horizontal line within box: median value; top and bottom edges of box: 25th and 75th percentiles; whiskers: 10th and 90th percentiles; open circles: outliers beyond 90th percentile. Lettering in each panel indicates multiple comparisons results obtained from repeated-measures ANOVA or an equivalent non parametric test for longitudinal data. Groups with different letters are significantly different...... 81

Figure 3.8 Mean + SE elemental ratio in core, middle and edge zones of otoliths from different sampling locations (see Table 3.1 for sampling codes). Lettering above each panel indicates multiple comparisons results obtained from repeated-measures ANOVA or an equivalent non parametric test for longitudinal data. For each region, values with different letters are significantly different...... 82

Figure 4.1 Light micrograph (1000× magnification) of the nucleus region of the lapillus of Boreogadus saida (a) and Arctogadus glacialis (b) showing the shortest nucleus diameter (SN, full line) and the longest nucleus diameter (LN, dashed line) measurements used to calculate the index of nucleus area (SN×LN)...... 94

Figure 4.2 Projection of 250 Boreogadus saida (left cluster) and 45 Arctogadus glacialis (right cluster) in the first factorial plane of the correspondence analysis of 18 microsatellites (excluding Gmo8). Closed and open symbols are fish identified as B. saida and A. glacialis respectively by Gmo8 analysis...... 97

Figure 4.3 Regressions of SN×LN on standard length (SL) for genetically identified Boreogadus saida (closed symbols, SN×LN = 2.684 SL + 390.145, r2 = 0.111, n = 602, p < 0.0001) and Arctogadus glacialis (open symbols, SN×LN = 2.905 SL + 693.989, r2 = 0.040, n = 110, p = 0.0365)...... 98

Figure 4.4 Frequency distribution of lapillus nucleus area (SN×LN) for Boreogadus saida (upturned distribution and right axis) and Arctogadus glacialis, and the logistic regression giving the probability p (left axis) of being A. glacialis as a function of nucleus area (Ln [p/(1-p)] = 0.02687 SN×LN - 17.5466, r2 = 0.724, n = 523)...... 99

Figure 5.1 Bathymetric map of southeast Beaufort Sea with sampling locations in 2004 (closed symbols) and 2008 (open symbols). Stars indicate sampling conducted from sea ice. CB: Cape Bathurst, FB: Franklin Bay, CP: Cape Parry, DB: Darnley Bay. . 112

Figure 5.2 Linear regression of age against notochord length and the cubic root of the product of standard length by body depth at the anus (LH1/3) for Boreogadus saida and Arctogadus glacialis aged by otolith analysis and identified to species by genetic and/or otolith nucleus size in 2004 (red) and 2008 (blue) and pooled years (black line)...... 115

Figure 5.3 Monthly frequency distributions of notochord length and the cubic root of the product of notochord length by body depth at the anus (LH1/3) for Boreogadus saida (black) and Arctogadus glacialis (grey) identified to the species by genetic and/or otolith nucleus size in 2004 and 2008. The percentage of fish identified to the species is indicated on the left panel. n indicate the number of B. saida, A. glacialis. Mean values for B. saida (black) and A. glacialis (grey) are indicated on each panel...... 116

Figure 5.4 Vertical distribution of larval and juvenile Boreogadus saida and Arctogadus glacialis collected by day (open bars) and by night (filled bars) over 14 dates from early June to early August 2004 in southeast Beaufort Sea...... 128

Figure 5.5 Vertical distribution by date of the early life stages of Boreogadus saida and Arctogadus glacialis collected in the summer 2004 in southeast Beaufort Sea...... 129

Figure 5.6 Monthly abundance (no. 1000 m-3) of Boreogadus saida and Arctogadus glacialis larvae and juveniles collected by oblique DSN tows in summer 2004 (red) and summer 2008 (blue)...... 130

Figure 5.7 Hatch date frequency distributions of Boreogadus saida and Arctogadus glacialis uncorrected for mortality in 2004 and 2008 for fish aged by otolith analysis and identified to the species by genetic and/or otolith nucleus size (black) and for all fish after species and age redistribution (grey). Number above every second bar indicates mean age of the sub-cohort. n indicates the number of fish identified to species, the total number of fish assigned to the species...... 131

Figure 5.8 Catch at age curves for Boreogadus saida (4-d age bins) and Arctogadus glacialis (6-d age bins) in 2004 (red) and 2008 (blue). The slopes of the regressions were used to estimate the mortality rate: B. saida 2004: Ln number = 6.912 - 0.043 age, r2 = 0.808; B. saida 2008: Ln number = 6.115 - 0.034 age, r2 = 0.882; A. glacialis 2004: Ln number = 3.507 - 0.018 age, r2 = 0.386; A. glacialis 2008: Ln number = 3.657 - 0.023 age, r2 = 0.498...... 132

Figure 5.9 Hatch date frequency distributions corrected for mortality for all Boreogadus saida and Arctogadus glacialis collected in the Beaufort Sea in 2004 and 2008 after species redistribution. The black line indicates weekly ice concentration in the study area. The timing of ice algae and phytoplankton production in Franklin Bay (2004) and Amundsen Gulf (2008) are indicated above each panel...... 133 xvi

Figure 5.10 Percent diet composition by number (a-c) and carbon (d-f) for three length classes of Boreogadus saida and Arctogadus glacialis sampled in 2004. n indicates the number of B. saida, A. glacialis analyzed. SI is Schoener’s index of diet overlap between the two species. The category ''Other'' includes Acartia spp., Eurytemora spp., Harpaticoida, Limnocalanus spp., Metridia longa, Metridia nauplii, Microcalanus spp., Microcalanus nauplii, Oncaea parila, Paraeuchaeta glacialis, unidentified calanoid copepod, unidentified cyclopoid copepod, unidentified copepod nauplii, Amphipoda, Appendicularia, Chaetognatha, Cirripeda, Cnidaria, Gastropoda, Ostracoda, Polychaeta and digested material...... 134

xvii

À Louis-Philippe

xix

«Le génie est fait d’un pour cent d’inspiration et de quatre-vingt-dix-neuf pour cent de transpiration.»

Thomas Edison

xxi

Remerciements

J’ai été très bien entourée pendant les années de mon doctorat et je pourrais ici faire l’éloge de nombreuses personnes. Mais je tenterai de rester brève dans mes remerciements. D’abord, toute ma reconnaissance au directeur de mes travaux, Louis Fortier, un grand scientifique, vraiment.

Merci aux membres de mon comité d’évaluation pour avoir investi de leur précieux temps : Jean-Éric Tremblay, Louis Bernatchez, Nicolas Derome (membre remplaçant), Warwick Vincent (examen doctoral), Martin Castonguay (examinateur externe) et Frédéric Maps (président de la soutenance).

Mon doctorat a été ponctué d’incursions dans plusieurs laboratoires où j’ai été à chaque fois très bien accueillie. Merci à mes collègues du Norwegian College of Fishery Science, Svein-Erik Fevolden, Jørgen Schou Christansen et Matias Madsen. À l’Université du Québec à Chicoutimi, je tiens à remercier Pascal Sirois, Anne-Lise Fortin et Joëlle Guérin du laboratoire d’écologie aquatique, qui m’ont permis de faire mes premiers pas en matière de chimie des otolithes. Merci à Simon Thorrold pour son accueil à WHOI et une collaboration très instructive, et à Scot Birdwhistell pour son aide avec l’analyse chimique des otolithes. Finalement, je tiens à témoigner de ma grande appréciation pour mes collègues de UVic, John Nelson et Stephanie Puckett. Dans tous ces magnifiques endroits, merci à tous ceux et celles qui ont fait en sorte que chacun des mes séjours soit une expérience inoubliable.

J’aimerais aussi remercier des gens qui ne le sont pas assez souvent, mais qui font certainement une différence dans le cheminement des étudiants chercheurs. Il s’agit de membres du personnel de Québec-Océan, et d’ArcticNet, notamment Brigitte Robineau, Lynn Bélanger, Guylaine Potvin, Richard Marquis, Keith Lévesque, Josée Michaud, Luc Michaud, Sylvain Blondeau, Pascal Massot, Steeve Gagné, Pierre-Yves Simard. Ce qu’aurait été mon doctorat sans technicien en informatique pour déboguer tous mes problèmes!

xxiii

Je tiens aussi à souligner le travail des capitaines et membres d’équipage du NGCC Amundsen, du NGCC Pierre Radisson, du IB Kapitan Dranitsyn et du RV Viktor Buynitskiy, qui ont permis aux missions océanographiques tant canadiennes que sibériennes d’être menées à bien. Merci aussi à tous les participants de ces missions qui m’ont aidée dans mon travail, physiquement ou moralement. D’ailleurs, je tiens à remercier chaleureusement Catherine Lalande sans qui certaines traversées des mers sibériennes m’auraient paru encore plus longues!

Évidemment, un gros merci à tous les membres du labo Fortier, présents, passés ou d’adoption, pour leur appui scientifique ou métaphysique (au bar, disons): Louis Létourneau, Dominique Robert, Gérald Darnis, Anna Prokopowicz, Pascale Lafrance, Vincent Perron, Simon Lebel, Marc Ringuette, Alexandre Forest, Makoto Sampei, Hélen Cloutier, Delphine Benoît, Stéphane Thanassekos, Catherine Lalande, Myriam Paquet- Gauthier, Sebastian Taborszky, Maxime Geoffroy, Samuel Lauzon, Marianne Falardeau- Coté, Cyril Aubry, Shani Rousseau, Amélie Marceau, Keita Suzuki, Salomé Mollard et Marianne Caouette.

Finalement, je ne saurai jamais assez remercier mes parents Margot et Régis, ma soeur Stéphanie et mes nièces Léa et Lydia, ainsi qu’à tous les membres de ma famille élargie et de ma belle-famille, pour tout le plaisir que j’ai à me retrouver auprès d’eux, même si cela n’arrive pas aussi souvent que je l’aimerais. Et, bien sûr, tout mon amour pour Louis- Philippe!

xxiv

Avant-Propos

La présente thèse de doctorat en océanographie regroupe une introduction générale (chapitre 1), quatre articles scientifiques rédigés en anglais (chapitres 2 à 5) et une conclusion générale (chapitre 6). L’objectif de cet ouvrage est de mettre en commun les connaissances antérieures, les résultats de mes propres études et les perspectives de recherche futures concernant l’écologie des jeunes stades de Boreogadus saida et d’Arctogadus glacialis*. Le chapitre 2 compare les saisons d’éclosion de Boreogadus dans plusieurs régions en lien avec les apports fluviaux. Le chapitre 3 met à profit la chimie des otolithes et se penche sur les régions d’éclosion et les mouvements migratoires de Boreogadus. Le chapitre 4 présente une méthode basée sur le noyau de l’otolithe qui permet de différencier les jeunes stades de Boreogadus et d’Arctogadus. Le chapitre 5 compare l’écologie larvaire et juvénile de Boreogadus et d’Arctogadus en mer de Beaufort. Les articles scientifiques, dont voici les références, ont été publiés dans des revues scientifiques ou le seront prochainement.

Chapitre 2: Bouchard C, Fortier L (2011) Circum-arctic comparison of the hatching season of polar cod Boreogadus saida: A test of the freshwater winter refuge hypothesis. Progress in Oceanography 90:105-116 (article reproduit avec la permission de l’éditeur).

Chapitre 3: Bouchard C, Thorrold S, Fortier L. Spatial segregation, dispersion and migration in early stages of polar cod Boreogadus saida revealed by otolith chemistry.

Chapitre 4: Bouchard C, Robert D, Nelson RJ, Fortier L (2013) The nucleus of the lapillar otolith discriminates the early life stages of Boreogadus saida and Arctogadus glacialis. Polar Biology 36:1537-1542 (article reproduit avec la permission de l’éditeur).

Chapitre 5: Bouchard C, Mollard S, Robert D, Suzuki K, Fortier L. Compared early life history of sympatric polar cod Boreogadus saida and ice cod Arctogadus glacialis in southeastern Beaufort Sea.

xxv

Le chapitre 2 est résumé pour les décideurs dans: Science for Environmental Policy, April 2012, Polar cod survival may be enhanced by climate warming, European Commission DG Environment News Alert Service. Thematic Issue 31: Arctic Science, p. 10. http://ec.europa.eu/environment/integration/research/newsalert/pdf/31si.pdf

De plus, je suis co-auteure principale des publications suivantes : Nelson RJ, Bouchard C, Madsen M, Praebel K, Rondeau E, Schalburg K, Leong J, Jantzen S, Sandwith Z, Puckett S, Messmer A, Fevolden S-E, Koop B (2013) Microsatellite loci for genetic analysis of the arctic gadids Boreogadus saida and Arctogadus glacialis. Conservation Genetics Resources 5:445-448.

Nelson RJ, Bouchard C (2013) Arctic cod (Boreogadus saida) population structure and connectivity as examined with molecular genetics. North Pacific Research Board Project 1125 Final Report, 37 p.

J’ai aussi donné des séminaires présentant mes résultats lors des congrès scientifiques nationaux et internationaux suivants:

Bouchard C, Fortier L (2010) Polar cod hatching season around the Arctic: influence of freshwater. Conférence de l'Année polaire internationale, Oslo, Norvège.

Bouchard C, Fortier L (2012) Early survival in frozen seas: synthesis of the factors influencing survival of polar cod (Boreogadus saida) larvae and juveniles. Réunion scientifique annuelle ArcticNet, Vancouver, Canada.

Bouchard C, Fortier L (2013) Morue arctique et changements climatiques: ce que 20 ans d’écologie larvaire nous ont appris. 81e Congrès de l’Acfas, Québec, Canada.

Les échantillons utilisés dans cette thèse doctorale ont été récoltés par plusieurs personnes du laboratoire de Louis Fortier, dont moi-même, lors de missions océanographiques sous la xxvi

direction d’ArcticNet (un réseau de centres d’excellence du Canada) dans l’Arctique canadien et du programme Nansen and Amundsen Basins Observational System (NABOS) dans l’Arctique sibérien. J’ai effectué les analyses de laboratoire avec l’aide d’excellents collaborateurs et collaboratrices à l’Université Laval, au Norwegian College of Fishery Science (Tromsø), au Woods Hole Oceanographic Institution et à l’Université de Victoria. J’ai personnellement fait les analyses statistiques, l’interprétation des résultats de même que la rédaction du présent document, avec de judicieux conseils de mes collègues. Au cours de mon doctorat, j’ai été appuyée financièrement par plusieurs organismes dont voici une liste non exhaustive : Fonds québécois de la recherche sur la nature et les technologies (Bourse de doctorat en recherche et Programme de stages internationaux), Ministère des Affaires Indiennes et du Nord Canada (Programme de formation scientifique dans le Nord et Prix commémoratif Malcolm Ramsay), Pan Arctic cluster for Climate forcing of the Arctic Marine Ecosystem (Projet de l’Année polaire internationale, Bourse pour programme d’échange étudiant), Université Laval (Fonds de soutien au doctorat, Bourse d’admission au doctorat, Aide au congrès), Québec-Océan (Soutien au doctorat, Aide au congrès, Aide aux formations de courtes durées, Prime à la publication).

* Note sur la taxonomie : Les deux espèces étudiées dans cette thèse, Boreogadus saida et Arctogadus glacialis, se partagent non seulement l’Arctique mais aussi les noms communs. En anglais, Boreogadus saida est généralement nommé Arctic cod en Amérique et polar cod en Eurasie. Inversement, Arctogadus glacialis est plus souvent désigné par le terme polar cod sur le continent américain et par celui d’Arctic cod ailleurs. Selon le Système d’Information Taxonomique Intégré (ITIS), la référence principale en taxonomie des espèces, Boreogadus saida et Arctogadus glacialis ont chacun deux noms communs en anglais, les mêmes, soit Arctic cod et polar cod. L’American Fisheries Society (AFS), importante référence en taxonomie des poissons, désigne Boreogadus saida par Arctic cod et Arctogadus glacialis par polar cod. La liste d’espèces publié par l’Aquatic Sciences and Fisheries Information System (ASFIS), qui fait office de référence pour les noms communs au sein de L’Organisation des Nations Unies pour l’alimentation et l’agriculture (FAO), utilise plutôt polar cod pour Boreogadus saida et Arctic cod pour Arctogadus glacialis. En français, Boreogadus saida est désigné par saïda franc (ITIS et AFS) ou morue polaire

xxvii

(FAO) tandis qu’Arctogadus glacialis n’a aucun nom français pour ces trois organisations. Cette situation a déjà entrainé des erreurs, notamment dans les pages de la très respectée revue Nature (Schiermeier 2007). Pour éviter toute confusion, les sections françaises de la présente thèse désignent les espèces par leur noms scientifiques uniquement. Aussi, puisque Boreogadus et Arctogadus représente chacun un genre monospécifique (voir section 5.3), ceux-ci sont utilisés dans la thèse pour en faciliter la lecture. Dans les sections anglaises de la thèse, les noms communs utilisés sont polar cod pour Boreogadus saida et ice cod pour Arctogadus glacialis.

xxviii

xxix

Chapitre 1 – Introduction générale

1.1 Les écosystèmes marins arctiques en changement

L’océan Arctique couvre moins de quatre pour cent de l’océan mondial, mais revêt une grande importance écologique, géopolitique et économique. Il est entouré de terres appartenant à cinq pays (Canada, Danemark, États-Unis, Norvège, Russie), desquels il reçoit beaucoup d’eau douce, de sédiments et de nutriments via de nombreux fleuves. En tant que région polaire, l’Arctique est extrêmement vulnérable aux changements climatiques actuels et futurs, et les preuves des impacts de ceux-ci sur les espèces, les communautés et les écosystèmes s’accumulent (IPCC 2013). L’augmentation des concentrations de gaz à effet de serre dans l’atmosphère depuis l’ère industrielle a entraîné un réchauffement global de la planète provoquant des bouleversements à l’échelle planétaire tel que la hausse du niveau des mers et l’acidification des océans. Le réchauffement climatique est fortement accentué aux pôles et l’Arctique en est affecté de plusieurs manières. Parmi les conséquences physiques les plus importantes en Arctique, la diminution draconienne de l’étendue du couvert de glace de mer (Stroeve et al. 2007), de son épaisseur moyenne (Rothrock & Zhang 2005) et de son volume (Schweiger et al. 2011) sont certainement les plus évidentes, et la disparition complète de la glace de mer en été est prédite d’ici la fin du siècle par les modèles climatiques (Holland et al. 2006, Stroeve et al. 2008, Boe et al. 2009, Wang & Overland 2009, Rampal et al. 2011). D’autres phénomènes cruciaux ont récemment été observés, notamment une augmentation des températures de surface de 0,12 à 0,70°C dans les différentes mers arctiques entre 1982 et 2006 (Belkin 2009), une augmentation des précipitations et des débits des grands fleuves (Peterson et al. 2002, Peterson et al. 2006, Zhang et al. 2013), et des intrusions de masses d’eau aux températures anormalement élevées (Polyakov et al. 2005, Shimada et al. 2006, Dmitrenko et al. 2008).

Le réchauffement climatique et l’éventail des changements physiques qui y sont reliés ont des conséquences biologiques dans tous les écosystèmes marins majeurs (Brierley & Kingsford 2009) y compris l’océan Arctique (Wassmann et al. 2011). Les effets directs des changements climatiques sur les espèces marines arctiques sont certainement nombreux,

1 bien que souvent ils soient difficiles à documenter ou inclus au sein d’interactions complexes dans la communauté ou l’écosystème entier. Certains effets sont notés au niveau de la phénologie, c’est-à-dire le synchronisme annuel des événements d’histoire de vie dans une population. Par exemple, le moment de la reproduction a été devancé dans la population de guillemot de Brünnich (Uria lomvia) du Nord de la baie d’Hudson suite au devancement de la fonte de la banquise printanière qui libère plus tôt l’accès aux poissons (Gaston et al. 2005). Plusieurs espèces répondent au réchauffement de leur environnement par des changements de leur aire de répartition, ce qui se manifeste par de nouvelles distributions chez les espèces arctiques (e.g. Rand & Logerwell 2011), mais aussi par l’arrivée d’espèces des zones tempérées dont l’aire de répartition remonte vers le nord (e.g. Perry et al. 2005, Hegseth & Sundfjord 2008, Mueter et al. 2009, Lenoir et al. 2011). Aussi, la fonte de la glace de mer pose un problème majeur pour ce qu’elle représente en tant que perte d’habitat pour nombre d’espèces arctiques telles que l’ours polaire (Ursus maritimus), le morse (Odobenus rosmarus), le phoque barbu (Erignathus barbatus) et le phoque annelé (Pusa hispida) (Laidre et al. 2008, Moore & Huntington 2008).

Mais la plupart des effets du réchauffement global dans l’océan Arctique sont indirects, c’est-à-dire s’affichent au travers la complexité des communautés et des écosystèmes. En mer du Nord, un écosystème ressemblant à celui de l’océan Arctique de par les fluctuations saisonnières très marquées dans la disponibilité de la nourriture, le niveau de réponse des différentes espèces aux changements climatiques n’est pas le même, ce qui amène un décalage temporel entre certaines espèces de plancton et leur ressource alimentaire (Edwards & Richardson 2004). De telles altérations phénologiques peuvent provoquer une cascade d’effets affectant tous les niveaux trophiques et ultimement de profonds changements écosystémiques. Dans l’Arctique, la production primaire globale augmentera vraisemblablement dans les prochaines années sous l’effet de la diminution du couvert de glace (Arrigo et al. 2008). Mais un phénomène probablement plus important que cet accroissement de productivité concerne l’effet de la disparition hâtive de la glace au printemps sur le synchronisme du bloom phytoplanctonique. En mer de Béring par exemple, lorsque la glace reste tard au printemps, le bloom phytoplanctonique apparaît sous la glace en avril, consomme la majorité des nutriments de la couche de mélange (qui se

maintient elle-même grâce à la glace), et, vu la faible abondance de zooplancton herbivore dans le milieu, coule vers le fond, favorisant ainsi le benthos (Overland & Stabeno 2004). Par contre, lorsque la glace disparaît tôt, la couche de mélange ne peut se maintenir et le bloom de phytoplancton arrive seulement lorsque la stratification thermique s’est installée, en mai ou juin. À cette période, une biomasse maximale de zooplancton herbivore est présente pour consommer cette production tardive, et l’énergie demeure ainsi dans le système pélagique au détriment du benthos (Overland & Stabeno 2004). Ce phénomène expliquerait, avec l’effet cumulatif d’autres changements climatiques, la transformation de l’écosystème observée ces dernières années en mer de Béring, où une réduction de la biomasse benthique et une augmentation du nombre de poissons pélagiques ont été notées (Grebmeier et al. 2006). En bref, la forte saisonalité caractéristique des mers arctiques s’érode peu à peu avec les chanegements climatiques, et ce phénomène bouleverse la structure et le fonctionnement des écosystèmes marins arctiques de plusieurs manières (Wassmann & Reigstad 2011).

Au sein de ces écosystèmes marins arctiques se trouve Boreogadus saida. Présente dans toutes les mers arctiques, c’est une espèce que l’on pourrait qualifier de plaque tournante dans son environnement puisqu’elle canalise le flux d’énergie de la production secondaire (le zooplancton) vers les niveaux trophiques supérieurs que sont les oiseaux et mammifères marins (Bain & Sekerak 1978, Bradstreet & Cross 1982, Hobson & Welch 1992a, Welch et al. 1992). Le régime alimentaire de plusieurs espèces d’oiseaux et de mammifères marins est constitué en grande majorité de Boreogadus (e.g. Finley et al. 1990, Welch et al. 1992, Weslawski et al. 1994, Bluhm & Gradinger 2008). L’espèce est assurément très abondante en Arctique, mais quantifier à quel point reste encore un Graal. L’espèce forme en été, près des côtes, des bancs parfois immenses représentant une biomasse apparemment inépuisable (Bradstreet et al. 1986, Crawford & Jorgenson 1993, Welch et al. 1993, Crawford & Jorgenson 1996) qui est largement exploitée par des milliers de prédateurs marins et aviaires (e.g. Hobson & Welch 1992b). Mais selon une étude effectuée dans le détroit de Lancaster, l’énorme biomasse contenue dans ces bancs ne représente qu’une infime proportion de ce qui est réellement consommé par les prédateurs (Welch et al. 1992), constatation qui a donné naissance au mystère des Boreogadus manquants. On sait

3 maintenant que Boreogadus passe l’hiver densément agrégé en eaux profondes (Benoit et al. 2008, Geoffroy et al. 2011), ce qui lève un peu le voile sur le mystère bien que les estimés d’abondance systématiques demeurent rares pour cette espèce. La population la mieux documentée est celle de la mer de Barents dont le stock est estimé annuellement depuis 1986 (Gjøsæter 2009). La biomasse de Boreogadus dans cette région de 1.4 million de km2 où Boreogadus n’est pas dominant et côtoie de nombreuses autres espèces de poissons compétitrices et prédatrices était de 1.2 à 1.8 million de tonnes entre 2000 et 2005 (Gjøsæter 2009). Dans la partie étasunienne des mers de Chukchi et de Beaufort, une région de 105 083 km2 où 67% de la biomasse de poissons est constituée de Boreogadus, la biomasse estimée atteint seulement 42 000 tonnes (NPFMC 2009).

Boreogadus est une espèce cruciale dans un écosystème où les effets des changements climatiques sont parmi les plus intenses de la planète. Cela est en soit une excellente raison pour s’atteler à l’étude de ce petit poisson. Cependant, un autre bon leitmotiv existe pour pousser à fond notre connaissance de l’espèce, et ce, le plus vite possible. Étant difficilement accessibles, les mers arctiques ont été relativement peu affectées par les activités humaines et leurs ressources naturelles très peu exploitées jusqu’à présent. Mais cette situation est sur le point de changer. La diminution croissante du couvert de glace, associée au contexte socio-économique actuel, entraîne une multiplication des activités humaines dans l’Arctique qui peuvent avoir des conséquences écologiques plus ou moins prévisibles. Le risque de pollution par les hydrocarbures et de marées noires s’accentue avec l’augmentation de la circulation maritime et des opérations d’exploration et d’exploitation pétrolière. De nouvelles zones de pêche deviennent accessibles. Pour ce qui est de Boreogadus, il a été exploité intensivement (jusqu’à 350 000 tonnes par année) au début des années 1970 par l’ancienne URSS, la Norvège, le Danemark et la République Démocratique Allemande qui le pêchaient dans l’Atlantique Nord-Est, en mer de Barents et en mer Blanche, principalement pour la fabrication de farine et d’huile de poisson (FAO 2012). Depuis, l’espèce n’est exploitée qu’à de faibles niveaux par la Russie (généralement moins de 50 000 tonnes par année, FAO 2012). Vu son accessibilité très limitée, sa petite taille et la faible qualité de sa chair, Boreogadus n’a jamais été exploité commercialement du côté canadien. Pour leur part, les États-Unis ont approuvé en février 2009 l’Arctic

Fishery Management Plan qui impose un moratoire sur toute pêche commerciale (excepté lee espèces de saumon du Pacifique et le flétan du Pacifique) dans les eaux arctiques au nord du détroit de Béring jusqu’à ce que de l’information suffisante existe pour autoriser un programme de pêcheries durables (NPFMC 2009).

Donc, les études de cette thèse sont empreintes des effets des changements climatiques, mais peu ou pas des autres facteurs anthropiques. Il est donc pressant de collecter autant d’information que possible entourant la biologie de l’espèce avant qu’elle ne devienne exploitée (elle-même ou en tant que prise accessoire), ou que différents effets anthropiques ne viennent brouiller les cartes en s’ajoutant aux changements climatiques en tant que perturbateurs. Aussi faut-il tenir compte du fait qu’en tant qu’espèce spécialiste des eaux glaciales, Boreogadus tendra à se faire remplacer peu à peu par d’autres espèces venant des régions tempérées/froides comme le capelan (Mallotus villosus). Il faudra donc une compréhension très approfondie de tous les aspects concernant Boreogadus pouvant servir de base scientifique solide à tout programme de gestion des ressources (biologiques ou non) dans l’Arctique. Boreogadus, qui a historiquement surtout été un sujet de science fondamentale, devient donc progressivement une ressource halieutique, ce qui implique notamment de définir des règles entourant la pêche (quotas, zones de pêches), d’établir des aires marines protégées et de documenter l’impact des exploitations pétrolières et gazières.

1.2 Boreogadus saida

L’importante valeur écologique de Boreogadus saida lui vaut une attention scientifique croissante depuis une vingtaine d’années. Cependant, certains aspects de la vie de ce poisson demeurent à ce jour empreints de mystère ou du moins de grand questionnement. Notamment, les sites de reproduction et d’éclosion, la structure des populations et les migrations sont parmi les aspects les plus méconnus. Mais avant d’aller plus loin avec les mystères, exposons d’abord sommairement l’état des connaissances sur Boreogadus.

Boreogadus est une espèce pélagique qui a une répartition géographique circumpolaire : on le retrouve dans toutes les mers arctiques, l’archipel canadien, la baie d’Hudson, la baie de Baffin de même que dans l’Atlantique canadien le long des côtes du Labrador (Scott & Scott 1988). Il possède des protéines antigel (Osuga & Feeney 1978) qui lui permettent de détenir le record de hautes latitudes parmi les poissons (Jensen 1948, Andriyashev 1954). Sa distribution suit en fait de près celle de la glace de mer, un élément majeur, sinon vital pour Boreogadus puisque les jeunes stades (œufs, larves et juvéniles) et les jeunes adultes utilisent la banquise et ses anfractuosités comme protection contre les prédateurs aquatiques et aviaires (Lønne & Gulliksen 1989, Gradinger & Bluhm 2004). Le mode de distribution de l’espèce est encore mal connu, mais semble très variable, allant des poissons dispersés (Crawford & Jorgenson 1990), aux petites agrégations de quelques dizaines de poissons (Lønne & Gulliksen 1989), aux bancs massifs formés l’été près de la surface (Welch et al. 1993, Crawford & Jorgenson 1996) et aux denses agrégations hivernales (Benoit et al. 2008, Geoffroy et al. 2011, Melnikov & Chernova 2013).

Selon les connaissances actuelles, la reproduction de Boreogadus aurait lieu de novembre à mars sous la glace ou près de ses marges (Craig et al. 1982, Bradstreet et al. 1986). Des aires de fraie spécifiques ont été suggérées dans les zones côtières des mers de Beaufort (Craig et al. 1982) et de Barents (Baranenkova et al. 1966) mais la présence de larves nouvellement écloses dans plusieurs régions de l’océan Arctique et celle d’individus ayant récemment frayé loin des côtes suggèrent que la reproduction se fait en de nombreux endroits et/ou sur de grandes étendues. La maturité sexuelle est atteinte à 2 ou 3 ans (Bain & Sekerak 1978) et l’espèce est itéropare (Hop et al. 1995). L’âge des individus dépasse rarement 6 ans et leur taille une trentaine de centimètres. Les œufs, au nombre de 9 000 à 21 000 par femelle, sont relativement gros pour un gadidé (1.5 à 1.9 mm) et, ayant une flottabilité positive, se retrouvent généralement sous la glace (Rass 1968, Graham & Hop 1995), où ils éclosent après une incubation d’un à trois mois, selon la température de l’eau (Rass 1968, Graham & Hop 1995).

À l’éclosion, les larves mesurent de 3.5 mm à 6.5 mm (e.g. Michaud et al. 1996, Ponomarenko 2000) et possèdent un grand sac vitellin, ce qui leur confère de faibles

capacités natatoires, mais leur permet de survivre une vingtaine de jours sans nourriture avant le début de l’alimentation exogène (Graham & Hop 1995). Au début, les larves se nourrissent principalement d’œufs et de nauplii de copépodes. Au cours de leur croissance, elles intègrent à leur régime alimentaire des petits copépodes cyclopoïdes et des copépodites de plus en plus gros (Ponomarenko 1967, Drolet et al. 1991, Fortier et al. 1995, Michaud et al. 1996). La période larvaire dure de trois à quatre mois et la métamorphose en juvénile se produit à une taille de 27-35 mm, ce qui correspond à un âge de 3 à 4 mois. Les larves restent près de la surface jusqu’au début août, les juvéniles descendent plus profondément pendant les mois d’août et septembre et vers la fin septembre, les juvéniles migrent dans la couche d’eau près du fond et y demeurent jusqu’en mars-avril (Ponomarenko 2000). Pendant cette période hivernale, les jeunes Boreogadus participent probablement aux migrations verticales journalières typiques de l’espèce (Benoit et al. 2010, Geoffroy et al. 2011).

Outre ces migrations verticales de nature journalières et saisonnières, la question des mouvements migratoires, actifs ou passifs, n’a pas été beaucoup étudiée. Il a été suggéré que la migration verticale saisonnière de l’espèce lors de laquelle les poissons descendent près du fond pour y passer l’hiver, peut dans certaines conditions de courant, entraîner une migration horizontale passive sur une distance importante (Benoit et al. 2008). Il serait donc possible que d’autres cas de migrations passives entraînent les poissons avec les différents courants. Quant à savoir si Boreogadus est capable de migrations actives, les avis divergent. Certains ont suggéré des migrations extensives des parties centrales du bassin arctique, où elles se nourrissent, vers les aires de fraie à la limite du pack (Ponomarenko 1968, Welch et al. 1993) alors que d’autres ont conclu à une faible activité de l’espèce en milieu naturel puisque son plus important investissement métabolique serait dans la capture et la digestion des proies et non dans l’activité de nage (Hop & Graham 1995).

Génétiquement parlant, Boreogadus a été très peu étudié jusqu’à maintenant. Les séquences complètes de l’ADN mitochondrial de Boreogadus et d’Arctogadus, récemment determinées, ont permis de confirmer que les deux taxons représente des groupes frères, c’est-à-dire phylogénétiquement très proches l’un de l’autre (Breines et al. 2008). La

7 première étude portant sur la génétique de populations de Boreogadus, utilisant des marqueurs RAPD (random amplified polymorphic DNA) sur des individus de l’Atlantique Nord, n’a pas trouvé de différenciation génétique au niveau des populations (Fevolden et al. 1999). Depuis, une analyse de l’ADN mitochondrial de Boreogadus capturés autour du Groenland, a détecté deux lignées distinctes mais aucune différenciation importante au niveau populationnel (Pálsson et al. 2009). De nouveaux marqueurs microsatellites développés pour Boreogadus et Arctogadus permettent maintenant d’étudier plus finement la stucture de populations de ces espèces (Nelson et al. 2013). L’utilisation de ces marqueurs sur un grand nombre de Boreogadus provenant de plusieurs régions a permis d’examiner la génétique de populations de l’espèce à trois niveaux : circumpolaire, régional, et sympatrique (Nelson & Bouchard 2013). À l’échelle circumpolaire, trois grands groupes génétiquement distincts ont été identifiés et correspondent avec la géographie: le groupe “Est” contenant tous les échantillons canadiens à l’est de Resolute (e.g. baie d’Hudson, golfe du Saint-Laurent, mer du Labrador); le groupe “Ouest’’ rassemblant les échantillons de l’Est du Groenland et des mers de Laptev, Chukchi, Bering et Beaufort (coté américain); et le groupe “Central’’ incluant les individus collectés du coté canadien de la mer de Beaufort. Ces observations démontrent que Boreogadus n’est pas panmictique à travers son aire de répartition. Au niveau régional, il apparaît une différenciation génétique significative entre les échantillons des mers de Chukchi, Bering et Beaufort. Au niveau sympatrique, aucune évidence de différenciation génétique n’a été trouvée entre deux cohortes larvaires (éclosion printanière et estivale) de Boreogadus du golfe d’Amundsen (Nelson & Bouchard 2013).

1.3 Arctogadus glacialis

Étant commun dans les mers arctiques mais beaucoup moins abondant que Boreogadus, Arctogadus glacialis a été relativement peu étudié. Ainsi, l’information disponible sur la biologie de l’espèce est très limitée et son rôle dans l’écosystème pratiquement inconnu. Comme Boreogadus, Arctogadus est cryopélagique, c’est-à-dire essentiellement pélagique mais associé à la glace de mer pour au moins une partie de son cycle vital (Andriyashev

1970). Arctogadus a une aire de répartition circumpolaire qui semble restreinte aux plateaux continentaux situés à l’intérieur du cercle polaire (Aschan et al. 2009). Il est généralement retrouvé dans des eaux oscillant entre 0.6 et 1.5°C, occasionnellement à des températures plus élevées, jusqu’à 3.4°C, et aussi parfois en eaux saumâtres (Nielsen & Jensen 1967). Au stade adulte, Arctogadus se nourrit principalement de copépodes calanoïdes, de mysidacés, d’amphipodes, et de Boreogadus (Coad et al. 1995, Süfke et al. 1998, Christiansen et al. 2012). Parfois, Arctogadus peut constituer une forte proportion des contenus stomacaux de mammifères marins, comme l’ont observé Finley & Gibb (1982) pour des narvals (Monodon monoceros) en été dans la baie de Pond Inlet, et Holst et al. (2001) pour des phoques annelés (Pusa hispida) du fjord Grise.

Arctogadus atteint la maturité sexuelle à 3 ou 4 ans, vit jusqu’à 9 ans, et atteint une quarantaine de centimètres (Coad et al. 1995). La période de fraie d’Arctogadus est incertaine et pourrait avoir lieu en hiver (e.g. Süfke et al. 1998) ou en été (voir Aschan et al. 2009). Les oeufs seraient démersaux, et la taille des larves à l’éclosion est inconnue (Fahay 2007). Les larves et les juvéniles d’Arctogadus et de Boreogadus sont morphologiquement très similaires (Fahay 2007). La seule différence détectable est une pigmentation présentant des ramifications sur la surface ventrolatérale de l’estomac chez Arctogadus (Fig. 1.1). Ce critère est cependant très difficile à évaluer, surtout chez les plus grandes larves. Par conséquent, les collections de gadidés arctiques larvaires et juvéniles, sont considérées n’être constituées que de Boreogadus, alors qu’un mélange des deux espèces est probable (e.g. Sekerak 1982). Cette incapacité à différencier les jeunes stades des deux espèces entraîne des biais dans les estimations des taux vitaux de l’espèce dominante Boreogadus, et empêche l’étude de l’écologie des jeunes stades de l’espèce relativement rare Arctogadus.

1.4 La survie larvaire : importance de l’eau douce

La plupart des populations de poissons marins connaissent de fortes fluctuations interannuelles de leur abondance et celles-ci sont en grande partie déterminées par le nombre de jeunes stades (œufs, larves et juvéniles) qui ont survécu jusqu’au stade adulte une année donnée, ce que l’on nomme recrutement (Bailey & Houde 1989, Leggett & Deblois 1994). Comprendre les facteurs qui affectent le recrutement et tenter de le prédire est devenu un enjeu majeur pour les biologistes en général et encore plus pour les spécialistes de l’océanographie des pêches. Typiquement, on cherche à corréler une mesure de recrutement avec une ou plusieurs variables environnementales abiotiques tel que la température, la salinité ou le vent, facteurs qui affectent la survie des jeunes stades principalement via leurs effets sur le succès d’alimentation et la croissance (Leggett & Frank 2008). D’autres facteurs abiotiques moins couramment étudiés se sont aussi avérés influencer la survie larvaire, notamment la photopériode pour deux espèces de gadidés (Buckley et al. 2006) et la disponibilité de la lumière pour Boreogadus (Gilbert et al. 1992, Fortier et al. 1996). Parmi les facteurs biotiques, l’abondance absolue de proies au début de la période d’alimentation (l’hypothèse de la période critique de Hjort 1914) n’aurait pas d’effet majeur sur le recrutement (Leggett & Deblois 1994). Par contre, le synchronisme entre l’éclosion des larves et la production de leurs proies (l’hypothèse du « match/mismatch », Cushing 1990) apparaît comme un facteur affectant le recrutement mais moins fortement que ne le suggère de nombreuses études (Leggett & Deblois 1994). Évidemment, d’autres relations intra- ou interspécifiques incluant la prédation (e.g. Fortier & Quiñonez-Velazquez 1998), la compétition (e.g. Fortier & Harris 1989), le cannibalisme (e.g. Fogarty et al. 2001) et le parasitisme (e.g. Bourque et al. 2006) influencent aussi la survie des larves de poissons. Cependant, les variations dans le recrutement sont toujours le résultat de l’interaction de plusieurs facteurs abiotiques et biotiques et l’isolement d’un seul effet est très difficile (Rose 2000).

Une croissance rapide des jeunes stades apparaît comme un facteur favorisant leur survie et plusieurs hypothèses ont été élaborées pour expliquer le mécanisme sous-jacent (Leggett & Frank 2008). Chez les juvéniles, la mortalité est généralement inversement proportionnelle à la taille puisqu’une grande taille améliore l’évitement des prédateurs (meilleure capacité

11 natatoire), la résistance à la famine (plus grandes réserves énergétiques) et la tolérance physiologique (meilleure résistance aux extrêmes de température, de salinité et d’oxygène dissous), trois attributs qui sont particulièrement importants pour contrer la mortalité hivernale (Sogard 1997). La réponse adaptative de plusieurs espèces des mers tempérées et polaires à cette mortalité hivernale dépendante de la taille consiste à maximiser la taille pré- hivernale en optimisant les dates d’éclosion et la croissance (e.g. Conover 1992, Fortier et al. 2006, Bouchard & Fortier 2008). Ainsi, la saison d’éclosion commencera dès que le ou les facteurs environnementaux limitants (souvent la température) permettent la survie des larves nouvellement écloses, allongeant ainsi au maximum la période de croissance avant l’hiver. À cet égard, le cas de Boreogadus en mer de Laptev est assez éloquent. La mer de Laptev est caractérisée par un vaste réseau de polynies (étendues libres de glace) qui s’ouvrent de façon récurrente pendant l’hiver et le printemps aux mêmes endroits, mais avec de fortes variations interannuelles dans la durée et l’étendue (Zakharov 1997). La saison d’éclosion de Boreogadus en mer de Laptev s’étend sur une longue période et commence tôt en hiver. L’ouverture des polynies en hiver laisse entrer de la lumière dans la colonne d’eau, ce qui permet la survie des larves qui sont des prédateurs visuels et qui ne pourraient se nourrir sous une épaisse couche de glace. À la fin de l’été, les larves qui sont nées en janvier sont donc beaucoup plus grosses que leurs consœurs nées en juin, ce qui leur donne de meilleures chances de survivre à l’hiver. Les années de faibles ouvertures des polynies sont associées à une mortalité massive des larves écloses en hiver, mais l’éclosion hâtive se maintient tout de même au sein de la population puisque les juvéniles aux grandes tailles pré-hivernales qu’elle engendre ont une valeur sélective maximale (Bouchard & Fortier 2008).

Tous les poissons marins à l’exception des chondrichtyens et des cœlacanthes sont hypo- osmotiques par rapport à leur milieu et conséquemment, ont des dépenses métaboliques liées à l’osmorégulation. Les jeunes stades ont une capacité osmorégulatrice qui s’accroît généralement au cours du développement et qui implique différents mécanismes et structures physiologiques (Varsamos et al. 2005). L’osmorégulation est de plus en plus énergétique à mesure que la salinité de l’eau ambiante s’éloigne de la valeur isosmotique de 10 à 12 et les larves de plusieurs espèces de poissons marins ont des taux de croissance et

de survie accrues lorsqu’elles sont élevées à des salinités plus faibles que la salinité ambiante (Holliday 1969). En milieu naturel, de telles diminutions de salinités se retrouvent dans les estuaires et au large de ceux-ci dans les panaches des fleuves. Depuis longtemps déjà, les estuaires ont été identifiés comme des pouponnières où les jeunes stades de plusieurs espèces de poissons tirent avantage d’une osmorégulation moins énergivore (Potter et al. 1990), de températures plus favorables, d’un risque de prédation plus faible et, probablement d’une nourriture plus abondante (Miller et al. 1985). Le degré d’utilisation de l’habitat estuarien est très variable selon l’espèce (résidante, diadrome, opportuniste, etc) mais des différences ontogéniques, géographiques, annuelles et saisonnières existent aussi au sein d’une même espèce (Able 2005). Par exemple, les harengs atlantiques juvéniles (Clupea harengus) de la mer du Nord, migrent des régions de fraie, situées au large, jusqu’en estuaires à des périodes précises au printemps et en hiver, comportement qui résulte en une augmentation considérable de leurs chances de survie (Maes et al. 2005). Les avantages de l’eau douce peuvent aussi se faire sentir en dehors de l’estuaire même. Ainsi, les larves d’albacore (Thunnus albacares) qui se développent près du panache du fleuve Mississippi dans le golfe du Mexique ont de meilleurs taux de croissance et de survie que celles qui demeurent plus au large (Lang et al. 1994).

L’Arctique, le plus petit et le moins profond des océans, est couvert à 53% par des plateaux continentaux (Jakobsson et al. 2003) et est donc grandement influencé par l’afflux d’eau douce des grands fleuves qui s’y jettent (Fig. 1.2). L’eau douce apportée des fleuves arctiques pourrait être très importante pour Boreogadus en tant que facteur affectant le déterminisme de la saison d’éclosion, qui se veut un équilibre entre la survie des larves et celle des juvéniles. En effet, le facteur principal pour la survie des larves nouvellement écloses est la température puisqu’elles ne survivent pas aux froids extrêmes (Michaud et al. 1996, Fortier et al. 2006). Ainsi, les larves qui éclosent en hiver arrivent dans une eau très froide et survivent peu alors que celles qui arrivent en été dans une eau plus chaude survivent mieux. Au stade juvénile, le facteur dominant pour la survie devient la taille puisqu’une grande taille à l’automne augmente les chances de survivre à l’hiver. Donc, pour maximiser la taille des juvéniles avant l’hiver, l’éclosion des larves doit commencer le plus tôt possible, selon les conditions environnementales. Puisque l’eau douce a un point de

13 congélation plus élevé que l’eau de mer (0°C versus -1.8°C), l’effet de la salinité sur la température de congélation pourrait contrôler la saison d’éclosion de Boreogadus en offrant aux larves un refuge thermique pendant l’hiver (Bouchard & Fortier 2008). Selon cette hypothèse, l’éclosion hivernale se retrouverait uniquement près d’une source d’eau douce alors que dans les régions peu influencées par l’eau douce, l’éclosion serait retardée jusqu’au réchauffement des eaux de surface au printemps. Les données disponibles supportent cette hypothèse : de courtes saisons d’éclosion centrées sur la débâcle sont observées dans la polynie des eaux du Nord et celle des eaux du Nord-Est (deux régions aux apports riverains pratiquement nuls), tandis qu’en mers de Beaufort, de Laptev et de Kara, où de grands fleuves débouchent, l’éclosion prolongée commence en hiver (Bouchard & Fortier 2008). Le fait que les individus les plus petits/jeunes se retrouvent en eau moins salée que les plus vieux (Bouchard & Fortier 2008, Parker-Stetter et al. 2011, Wong et al. 2013) concorde aussi avec cette hypothèse.

Le réchauffement global entraîne une diminution de l’étendue et de l’épaisseur du couvert de glace (Rothrock & Zhang 2005, Stroeve et al. 2007), une augmentation des précipitations et du débit des fleuves arctiques (Peterson et al. 2002, 2006, Zhang et al. 2013), ainsi qu’un rétrécissement des glaciers arctiques (Dyurgerov & Carter 2004, Dyurgerov et al. 2010, Gardner et al. 2011) et de l’inlandsis du Groenland (Rignot et al. 2008, van den Broeke et al. 2009, Velicogna 2009), phénomènes qui ajoutent de l’eau douce à l’océan Arctique. La salinité de surface tend ainsi à diminuer bien que cette tendance soit inversée dans certaines régions, dont les bassins centraux (Polyakov et al. 2008) et les mers de Laptev et de Sibérie Orientale (Steele & Ermold 2004). L’eau douce représente donc, au même titre que la température ou l’intrusion d’espèces notamment, un facteur très important pour comprendre la réponse de Boreogadus aux changements climatiques.

1.5 Objectifs

Au départ, ma thèse devait se concentrer uniquement sur Boreogadus saida et les principaux objectifs étaient d’étudier la structure des populations à l’aide de la génétique et de tester l’hypothèse du refuge thermique hivernal, élaborée au cours de ma maîtrise, à l’aide de trois approches différentes. L’étude génétique des populations de Boreogadus, que j’ai initiée au début de mon doctorat, a évoluée en projet internationnal de grande ampleur, maintenant dirigé par John R. Nelson de l’Université de Victoria. C’est en étudiant la différenciation génétique de deux cohortes larvaires de Boreogadus que j’ai découvert que mes échantillons, présumés monospécifiques, contenaient un certain nombre d’Arctogadus glacialis. J’ai donc inclus Arctogadus dans mes travaux. J’ai d’abord considéré Arctogadus en tant qu’élément de contamination dans l’étude de l’écologie larvaire de Boreogadus (chapitre 2), puis j’ai développé une méthode pour l’identifier (chapitre 4), et enfin j’ai étudié son écologie larvaire en parallèle avec celle de Boreogadus (chapitre 5).

L’objectif général de ma thèse est donc de documenter l’écologie des jeunes stades de Boreogadus et d’Arctogadus. Le premier chapitre a révisé sommairement l’état des connaissances reliées au sujet qui me préoccupe. Le second chapitre vise à déterminer les liens entre l’apport d’eau douce par les fleuves et la saison d’éclosion de Boreogadus en comparant plusieurs régions de l’Arctique. Le troisième chapitre vérifie, à l’aide de la chimie des otolithes, l’hypothèse selon laquelle certaines larves de Boreogadus éclosent en hiver près des panaches des fleuves, et examine la ségrégation spatiale des jeunes stades. Le quatrième chapitre décrit une nouvelle méthode basée sur les otolithes pour discerner les jeunes stades de Boreogadus et d’Arctogadus. Le cinquième chapitre décrit les jeunes stades de Boreogadus et d’Arctogadus de la mer de Beaufort en terme de distribution spatiotemporelle, de saison d’éclosion, de taille à l’éclosion, de croissance et de mortalité. Finalement, en guise de conclusion générale, le sixième chapitre résume l’essentiel des avancées scientifiques présentées dans la thèse et expose quelques possibilités de recherches futures.

1.6 Approches utilisées

Le présent projet de doctorat met à profit trois approches très distinctes, mais complémentaires pour répondre à des questions clés sur la biologie de Boreogadus saida et d’Arctogadus glacialis. La microstructure des otolithes est le fil conducteur de la thèse : elle est utilisée pour tous les chapitres. Aux chapitres 2 et 5, l’analyse des anneaux journaliers permet de documenter la saison d’éclosion, la croissance et la survie larvaire des deux espèces. Les lectures d’âge faites sur les otolithes sont aussi utilisées au chapitre 3 pour tester le lien entre la date d’éclosion et la composition chimique des otolithes. Finalement, un aspect très précis de l’otolithe, la taille du noyau, est utilisé au chapitre 4 comme outil permettant de distinguer Boreogadus d’Arctogadus. La chimie des otolithes est utilisée au chapitre 3 pour tester l’hypothèse du refuge thermique hivernal et apporte aussi de l’information sur la ségrégation spatiale et les mouvements migratoires de Boreogadus. La génétique est un outil précieux utilisé tout au long de la thèse pour identifier Boreogadus et Arctogadus. Au chapitre 2, quelques poissons sont génétiquement identifiés comme Arctogadus à l’aide d’un marqueur microsatellite (selon une méthode développée par des collègues norvégiens), donnant un premier indice sur l’abondance relative des deux espèces. Au chapitre 4, des analyses génétiques plus poussées (génotypage de 19 microsatellites, séquençage de l’ADN mitochondrial) servent à distinguer Boreogadus d’Arctogadus dans le développement de la méthode d’identification basée sur la taille du noyau de l’otolithe. Au chapitre 5, les deux méthodes (génétique et otolithométrique) sont utilisées conjointement pour identifier à l’espèce le plus grand nombre possible de jeunes gadidés composant les échantillons. Les travaux en génétique de populations entrepris au cours du présent projet incluent le développement de marqueurs microsatellites pour l’étude génétique de Boreogadus et d’Arctogadus (Nelson et al. 2013) ainsi qu’une étude circumpolaire de la structure des populations de Boreogadus. Cette dernière étude sera soumise pour publication en 2014 mais les résultats préliminaires sont présentement disponibles (Nelson & Bouchard 2013). Les sections suivantes expliquent brièvement les approches utilisées aux niveaux théorique et méthodologique.

1.6.1 Microstructure des otolithes

Tous les poissons osseux possèdent trois paires d’otolithes (les sagittae, les lapilli et les asterici) qui se distinguent par leur emplacement dans l’oreille interne, leur grandeur et leur forme. Ces structures inorganiques de carbonate de calcium font partie du système vestibulaire lequel joue un rôle primordial dans l’équilibre et la détection du mouvement. Mais pour les ichtyologistes, les otolithes sont particulièrement intéressants vu leur utilité pour déterminer l’âge des poissons. En effet, la croissance des otolithes se fait un peu à l’image de celle des arbres, par l’addition successive de couches concentriques de matériau, ce qui forme des anneaux annuels qui peuvent être dénombrés. La première identification d’anneaux annuels dans les otolithes (attribuée à Johannes Reibisch) date de 1899 et demeure de nos jours une technique d’estimation d’âge largement utilisée pour les poissons (Jackson 2007). En comparaison, la découverte d’accroissements journaliers dans les otolithes est relativement récente (Pannella 1971). L’étude de la microstructure des otolithes (aussi appelée otolithométrie) a dès lors suscité un intérêt marqué notamment parmi les spécialistes d’écologie larvaire qui depuis ont mis cette méthode à profit dans plusieurs centaines d’études portant sur la structure d’âge, les dates d’éclosion, la survie, le recrutement, la croissance et les transitions dans l’histoire de vie des jeunes stades (Campana & Neilson 1985, Jones 1986, 1992).

La présence d’accroissements journaliers dans les otolithes apparaît comme une caractéristique universelle chez toutes les espèces de poissons et dans tous les milieux. Au microscope, les accroissements journaliers prennent la forme d’une succession de bandes sombres et claires, appelées respectivement zones discontinues (étroites et pauvres en carbonate de calcium) et zones d’accrétion (riches en CaCO3 et définissant la largeur de l’accroissement journalier). Ce motif généré lors de la croissance des otolithes reflète un rythme circadien endogène qui serait amorcé chez les larves par la photopériode et maintenu ensuite même en absence de celle-ci (Campana & Neilson 1985). D’autres facteurs environnementaux et physiologiques comme les variations de température, l’alimentation et l’activité, ont une influence sur la formation des accroissements, ce qui peut parfois entraîner l’apparition d’accroissements supplémentaires (dits subjournaliers), bien qu’en conditions naturelles, les cycles de ces différents facteurs viennent généralement

renforcer le cycle endogène journalier (Campana & Neilson 1985). Ainsi, pour chaque nouvelle espèce étudiée à l’aide de la microstructure des otolithes, il est souhaitable d’avoir au préalable confirmé que les accroissements sont bel et bien déposés de façon journalière. Pour ce faire, il existe plusieurs méthodes de validation du taux de déposition, notamment le marquage des otolithes à l’aide de composés chimiques fluorescents (Geffen 1992).

L’étude de la microstructure des otolithes comporte plusieurs étapes. D’abord, l’échantillonnage et la conservation des individus doivent être appropriés (Butler 1992). L’examen des otolithes étant très chronophage, il est généralement nécessaire de sélectionner un sous-échantillon représentatif pour les analyses de microstructure. Les otolithes utilisés pour l’étude (les lapilli par exemple), sont soigneusement extraits de la tête des individus, montés sur une lame de microscope à l’aide d’une colle adaptée puis finement polis jusqu’à ce que les accroissements journaliers deviennent facilement observables au microscope (Secor et al. 1992). L’étape suivante constitue l’interprétation de la microstructure, aussi appelée la lecture des otolithes (Campana & Jones 1992). À l’aide d’un microscope à fort grossissement relié à un ordinateur ayant un logiciel d’analyse d’images, un technicien identifie sur l’otolithe chaque accroissement et enregistre le nombre (estimation de l’âge) ainsi que l’espace entre chacun (estimation de la croissance journalière). La lecture des otolithes demande une bonne expérience et est toujours empreinte d’une certaine subjectivité. C’est pourquoi il est fortement recommandé d’effectuer des contrôles de qualité, par exemple en comparant les estimations d’âges effectuées par différents lecteurs ou par un même lecteur à plusieurs reprises (Campana et al. 1995). Les données otolithométriques permettent notamment la définition de modèles de croissance, le rétrocalcul de la taille des individus à un temps donné et l’analyse de la fréquence des dates d’éclosion (Campana & Jones 1992).

1.6.2 Chimie des otolithes

En croissant, la matrice de carbonate de calcium des otolithes incorpore aussi des éléments retrouvés en faible quantité dans l’environnement. Puisque l’otolithe est acellulaire et métaboliquement inerte, les éléments qui s’y incrustent y demeurent de manière

19 permanente. Une chronologie chimique s’enregistre donc dans les otolithes et celle-ci reflète certains aspects de l’environnement auxquels le poisson a été exposé tout au long de sa vie. L’analyse de la composition chimique des otolithes (aussi appelée, empreinte élémentaire, signature chimique ou géochimie des otolithes) constitue ainsi un outil puissant pour documenter des sujets tels les migrations ou autres mouvements d’individus, la philopatrie, les traits d’histoire de vie, la connectivité au sein des populations et l’identification des stocks. Depuis ses débuts, il y a une trentaine d’années, le domaine de la chimie des otolithes s’est accru de façon exponentielle pour devenir l’aspect le plus étudié de la science des otolithes (Campana 1999, Campana 2005, Miller et al. 2010). Plusieurs synthèses y ont été consacrées (Campana 1999, Thresher 1999, Elsdon & Gillanders 2003a, Elsdon et al. 2008), de même qu’une session à chacune des quatre éditions de l’International Otolith Symposium (Secor et al. 1995, Fossum et al. 2000, Begg et al. 2005, Miller et al. 2010).

Plus d’une trentaine d’éléments peuvent se retrouver dans les otolithes de poissons marins (Campana 1999). La détection de ces éléments dans les otolithes peut être faite à l’aide de divers appareils, les plus utilisés étant la spectrométrie de masse avec plasma inductif (ICPMS), l’émission X induite par proton (PIXE) et les microsondes électroniques (Campana et al. 1997). Peu importe l’instrument de mesure, deux types de méthodes peuvent être utilisées pour les analyses. Dans la première, chaque otolithe est entièrement dissous dans l’acide et analysé en tant que solution, ce qui permet d’avoir une signature chimique représentant la vie entière du poisson. La seconde méthode consiste à échantillonner une portion de l’otolithe seulement, ce qui est possible en utilisant l’ablation au laser notamment. Selon les objectifs de l’étude, l’otolithe peut être analysé en un seul endroit reflétant une période spécifique de la vie d’un poisson (le bord de l’otolithe pour documenter les derniers jours de vie, par exemple), ou en plusieurs endroits représentant différents stades de vie d’un même individu. Cette dernière technique vise généralement à relier les mouvements des poissons aux changements de composition chimique des otolithes.

Les éléments retrouvés dans l’eau de mer passent par le sang, puis par l’endolymphe, avant d’être incorporés aux otolithes. Entre chaque étape, un mécanisme physiologique (assimilation branchiale, transport cellulaire, puis cristallisation) intervient et concentre ou dilue les éléments (Campana 1999). Cette discrimination n’est pas la même pour tous les éléments et peut être quantifiée à l’aide du coefficient de discrimination (D), c’est-à-dire le rapport élément:calcium dans l’eau divisé par le rapport élément:calcium dans l’otolithe. Certains éléments assujettis à une forte régulation physiologique (par exemple Na, K, S, P, Cl) ont des valeurs de D très faibles (< 0,05) indiquant qu’ils ne sont pratiquement pas influencés par leur concentration relative dans l’environnement et donc d’une utilité limitée pour étudier la ségrégation spatiale ou les migrations. Cependant, le coefficient de discrimination de plusieurs éléments (par exemple Li, Mg, Mn, Fe, Sr, Ba, Pb) est assez élevé (0.14 à presque 1.0) pour affirmer que leur concentration dans les otolithes reflète celle dans l’environnement. L’eau environnante est considérée comme la source principale d’éléments incorporés dans les otolithes, bien qu’une petite proportion puisse provenir de l’alimentation (Walther & Thorrold 2006). Toutefois, l’incorporation des éléments dans les otolithes est influencée par des processus physiologiques (temps d’exposition, ontogénie, taux de croissance) et par des variables environnementales comme la salinité et la température (Elsdon & Gillanders 2003b). Les effets de ces variables ne sont pas les mêmes pour tous les éléments ni pour toutes les espèces, sont souvent non linéaires, et peuvent agir de façon interactive (Elsdon & Gillanders 2002, 2003b, 2004, Martin & Thorrold 2005). Les études d’identification des stocks se basent sur l’hypothèse selon laquelle les otolithes acquièrent une signature chimique unique à la masse d’eau particulière dans laquelle le poisson a évolué au cours de sa vie. Il n’est donc pas essentiel pour ce type d’étude de connaître l’effet des facteurs environnementaux qui affectent l’incorporation des éléments (Campana & Thorrold 2001). L’utilisation de la chimie des otolithes pour documenter les migrations ou pour reconstruire l’histoire environnementale des poissons implique cependant beaucoup plus de prémisses et demande une bonne compréhension des facteurs influençant la chimie des otolithes (Elsdon et al. 2008).

1.6.3 Génétique des populations

Toute espèce est soumise aux quatre forces évolutives soit la sélection naturelle, la dérive génétique, le flux génique (migration et dispersion) et les mutations. Ces processus interagissent ensemble et avec les processus historiques (phylogéniques) pour créer la diversité génétique et structurer les populations à différentes échelles spatiales et temporelles. L’étude de la structure génétique des populations permet de documenter plusieurs aspects d’une espèce notamment en ce qui concerne l’isolement reproductif, la connectivité, l’utilisation de l’habitat, les comportements reproductifs et les patrons de migrations ou d’autres types de mouvements. Plusieurs espèces de poissons marins ont des taux de différenciation génétique très faibles puisque la taille immense de leurs populations et leur fécondité élevée limitent la dérive génétique, alors que leur comportement migratoire, leurs jeunes stades pélagiques, leurs aires de reproduction mal délimitées ou chevauchantes ainsi que la rareté des barrières géographiques maximisent le flux génique (Ward et al. 1994, DeWoody & Avise 2000). Mais grâce à de puissants outils moléculaires comme les microsatellites, des structures de population ont été détectées chez plusieurs espèces de poissons marins, dont le hareng de l’Atlantique (Shaw et al. 1999), la morue franche (Knutsen et al. 2003) et le thon rouge de l’Atlantique (Carlsson et al. 2004). La structure génétique des populations de poissons permet d’améliorer notre compréhension fondamentale des espèces. Elle s’avère aussi très importante pour la gestion et la conservation des ressources halieutiques, et la génétique des pêches est aujourd’hui un domaine en pleine expansion (e.g. Ward 2000, Svedang et al. 2010).

L’intégration de données génétiques, écologiques et environnementales enrichit considérablement nos capacités à expliquer certains processus naturels. Ce type d’approche multidisciplinaire, appelée génétique du paysage, s’est développé rapidement au cours des dernières années et est de plus en plus appliqué au milieu marin (Manel et al. 2003, Storfer et al. 2007, Selkoe et al. 2008, Sork & Waits 2010, Storfer et al. 2010). En effet, plusieurs facteurs peuvent contribuer à structurer les populations marines et l’application de la génétique du paysage permet l’identification de ceux-ci. Dans les cas d’isolement par la distance, le taux de différenciation génétique augmente avec la séparation spatiale entre les paires de populations ou de sous-populations d’une espèce (Wright 1943). Ce phénomène a

été démontré chez certaines populations de poissons marins, et ce, même pour des espèces relativement mobiles tel que la morue franche (Pogson et al. 2001), le lieu de l’Alaska (O'Reilly et al. 2004) et la plie d’Europe (Watts et al. 2010). Parmi les paramètres environnementaux pouvant influencer la structure génétique des poissons marins et suggérés par la génétique du paysage, notons la température et la salinité (e.g. Cimmaruta et al. 2005, Jørgensen et al. 2005), les courants et la circulation océanique (e.g. Was et al. 2008) et les caractéristiques bathymétriques (e.g. Hemmer-Hansen et al. 2007). Plusieurs mécanismes écologiques peuvent aussi expliquer la différenciation génétique, notamment les comportements reproducteurs (e.g. Was et al. 2008), les migrations alimentaires (e.g. Gaggiotti et al. 2009), la philopatrie (e.g. Nesbø et al. 2000) et l’adaptation locale (e.g. Vilas et al. 2010).

Les microsatellites sont des marqueurs moléculaires très performants largement utilisés depuis une trentaine d’années en génétique des populations, du paysage et des pêches (Wright & Bentzen 1994, O'Connell & Wright 1997, Chistiakov et al. 2006). Ces séquences d’ADN sont constituées d’éléments d’une à six paires de bases répétées en tandem plusieurs fois (par exemple, GCGCGCGCGCGCGC). Les loci microsatellites sont très abondants dans le génome. Ils sont délimités par des régions flanquantes, spécifiques à chaque locus, conservées à l’intérieur de l’espèce et souvent aussi au sein des groupes phylogénétiques (Rico et al. 1996). Le polymorphisme des microsatellites provient de la variation du nombre d’éléments répétés (taille des allèles en paires de bases) causée par l’insertion ou la délétion d’éléments lors de la méiose dans un processus mutationnel pas-à- pas, c’est-à-dire un élément à la fois (Bell & Jurka 1997). Puisque le taux de mutation est très élevé (Ellegren 2000), les microsatellites sont hypervariables, une caractéristique particulièrement adaptée à l’étude de la structure des populations. Les microsatellites sont des marqueurs considérés comme neutres, c’est-à-dire qu’ils se situent en très grande majorité dans des régions non-codantes de l’ADN et donc que la diversité génétique de ceux-ci est façonnée principalement par les forces évolutives neutres (dérive, mutation, flux génique) et non par la sélection naturelle (e.g. Tóth et al. 2000, McCusker & Bentzen 2010). Techniquement, l’analyse d’échantillons avec des marqueurs microsatellites se fait en trois étapes, excluant l’étape préalable du développement d’amorces correspondantes

23 aux régions flanquantes des loci choisis. La première étape constitue l’amplification par PCR (réaction en chaîne de la polymérase) avec les amorces, la seconde, réalisée avec l’aide d’un séquenceur, est l’électrophorèse (séparation des allèles selon leur taille) alors que la dernière étape, celle du génotypage, consiste à visualisation les électrophorégrammes pour déterminer le ou les allèles de chaque échantillon.

Chapitre 2 – Circum-arctic comparison of the hatching season of polar cod Boreogadus saida: A test of the freshwater winter refuge hypothesis

2.1 Résumé

L’hypothèse selon laquelle les différences dans les températures de surface hivernales reliées à la salinité dictent les différences régionales dans la saison d’éclosion de Boreogadus saida est testée en comparant les distributions de fréquence de date d’éclosion de six régions océanographiques de l’Océan Arctique caractérisées par différents apports d’eau douce. Conformément à l’hypothèse, l’éclosion commence dès janvier et s’étend jusqu’en juillet dans les mers recevant de forts apports fluviaux (mers de Laptev et de Sibérie Orientale, baie d’Hudson et mer de Beaufort). En revanche, l’éclosion est restreinte à avril–juillet dans les régions où les apports d’eau douce sont limités (archipel arctique canadien, nord de la baie de Baffin et polynie des eaux du Nord-Est). La longueur (poids) à la fin de l’été (14 août) varie de <10 mm (<0.01 g) pour les larves nées en juillet à 50 mm (0.91 g) pour celles nées en janvier. Une débâcle hâtive de la glace de mer, des polynies hivernales plus fréquentes, une couche de surface plus chaude, et des débits fluviaux plus grands reliés aux changements climatiques pourraient favoriser la survie des Boreogadus juvéniles 0+ en permettant à une plus grande proportion de la cohorte annuelle d’éclore tôt et d’atteindre une grande taille avant la migration automnale vers les sites d’hivernage profonds. Un test supplémentaire de l’hypothèse nécessiterait la vérification que l’éclosion hivernale de Boreogadus se produit réellement dans le refuge thermique fournit par les panaches des fleuves sous la glace.

2.2 Abstract

The hypothesis that salt-related differences in winter sea surface temperature dictate regional differences in the hatching season of polar cod Boreogadus saida is tested by contrasting hatch-date frequency distributions among six oceanographic regions of the Arctic Ocean characterized by different freshwater input. Consistent with the hypothesis, hatching started as early as January and extended to July in seas receiving large river discharge (Laptev/East Siberian Seas, Hudson Bay, and Beaufort Sea). By contrast, hatching was restricted to April–July in regions with little freshwater input (Canadian Archipelago, North Baffin Bay, and Northeast Water). Length (weight) in late-summer (14 August) varied from <10 mm (<0.01 g) in July hatchers to 50 mm (0.91 g) in January hatchers. An earlier ice break-up, more frequent winter polynyas, a warmer surface layer, and increased river discharge linked to climate warming could enhance the survival of juvenile 0+ polar cod by enabling a larger fraction of the annual cohort to hatch earlier and reach a larger size before the fall migration to the deep overwintering grounds. A further test of the hypothesis would require the verification that the early winter hatching of polar cod actually occurs in the thermal refuge provided by under-ice river plumes.

2.3 Introduction

The polar cod Boreogadus saida plays a central role in the relatively simple pelagic food web of Arctic seas by channeling a major fraction of the energy flow between plankton and vertebrates (Bradstreet et al. 1986, Welch et al. 1992). Polar cod spawn in late fall and early winter under the ice cover of arctic shelves, and fertilized eggs rise to the ice–water interface (Rass 1968, Craig et al. 1982, Graham & Hop 1995, Ponomarenko 2000). Embryonic development may take as long as 60–90 d at the sub-zero temperatures prevailing under the ice (Altukhov 1979, Ponomarenko 2000). Length at hatch ranges from 4 to 8 mm (Rass 1968, Aronovich et al. 1975, Graham & Hop 1995, Michaud et al. 1996). Metamorphosis into pelagic juveniles occurs at 27–35 mm, and the migration from the surface layer to the deeper overwintering grounds begins at 30 to 35 mm (Baranenkova et al. 1966, Ponomarenko 2000).

Starting in late-summer, polar cod fry are preyed upon by seabirds in the surface layer (Bradstreet 1982, Karnovsky & Hunt 2002) and then by their adult congeners as they migrate at depth to their overwintering grounds (Baranenkova et al. 1966). A large size at the end of the short arctic summer should reduce the vulnerability of juveniles to avian predation, cannibalism, and winter starvation. Hence, selection pressures should push hatching to occur as early in winter or spring as environmental conditions will allow, so as to maximize the duration of the growth season and late-summer size (Fortier et al. 2006). We hypothesized that salinity-induced variations in sea surface temperatures dictate regional differences in the hatching season of polar cod (Bouchard & Fortier 2008). In coastal seas influenced by large rivers, brackish conditions in under-ice river plumes would provide the larvae with temperature only slightly below 0°C, accelerating embryonic development and allowing successful first-feeding and survival in winter. In regions with little freshwater input, the – 1.8°C temperature prevailing under the ice in winter would slow egg development and limit first-feeding and survival. In such regions, hatching would have to be delayed until the vernal warming of the surface layer for the larvae to survive.

In the present study, we test the prediction of the freshwater thermal refuge hypothesis that the hatching of polar cod starts in winter in regions of the Arctic Ocean influenced by freshwater, and is delayed until spring elsewhere. Based on new and published data, interannual variations and regional differences in the hatch-date frequency distribution (HFD) of polar cod are contrasted among six oceanographic regions of the Arctic Ocean ranging from inland and coastal seas heavily influenced by rivers to recurrent polynyas with little freshwater input.

2.4 Materials and methods

2.4.1 Study areas

Previously published hatch-date frequency distributions (HFDs) and new HFDs based on the otolith aging of pelagic juveniles sampled in late-summer and early fall, were used as estimates of the hatching season of polar cod in six regions of the Arctic Ocean that differ widely in their surface salinity (Fig. 2.1). Depending on the availability of data, the hatching season of polar cod was estimated for one to three different years in each region.

The Laptev Sea covers a wide shallow continental shelf and part of the deep Nansen Basin. The Lena and other rivers inject a total annual freshwater input of 738 km3 in the shallow sector (Gordeev 2006). The resulting dilution of the surface layer extends far offshore (Fig. 2.1a). The Laptev Sea is typically covered with ice from October to June, with some sectors in the North and West remaining ice-covered throughout the year. Polynyas are important features of the Laptev Sea that form under certain wind conditions as an enlargement of the circumpolar flaw lead that separates the fixed landfast ice from the mobile central ice pack. Our sampling in this region extended to the fringe of the adjacent East Siberian Sea.

Hudson Bay is a shallow (average depth of 150 m) estuarine sub-arctic sea (Fig. 2.1b). Several large rivers and James Bay input an average annual freshwater volume of 714 km3 in Hudson Bay (Déry et al. 2005). Together with Hudson Strait and Foxe Basin, it is often referred to as the Hudson Bay System. Seasonal ice is present in the bay from October to July with maximum thickness and extent in April. The general circulation is cyclonic and

slow, with cold and salty water from Foxe Basin entering from the northwest, and warmer, fresher water exiting along the eastern coast (Saucier et al. 2004 and references therein).

Southeastern Beaufort Sea extends over the shallow and wide Mackenzie Shelf and the Amundsen Gulf (Fig. 2.1c). The surface layer of the Beaufort Sea is strongly diluted by the plume of the Mackenzie River, the third largest river discharging into the Arctic Ocean (330 km3 year-1) (Macdonald et al. 1998). SE Beaufort Sea is typically covered with ice from October to June. In winter and spring, the circum-arctic flaw lead follows the 20-m isobath on the Shelf (Arrigo & van Dijken 2004, Lukovich & Barber 2005). Throughout winter, floe rafting at the edge of the landfast ice builds the stamukhi, a thick linear hummock that dams the Mackenzie River plume to form the seasonal brackish Lake Herlinveaux (70 km3) under the ice cover of the inshore shelf (Macdonald et al. 1995). With ice break-up in June or July, the flaw lead widens to form the Cape Bathurst polynya that extends over the Shelf and the Amundsen Gulf (Arrigo & van Dijken 2004). The break-up of the stamukhi in early summer releases the turbid and brackish waters of the seasonal lake in the top 5–10 m of the surface layer on the shelf. By August or September, the region is normally ice-free except for the permanent central ice pack over the northern sector.

The channels of the Northwest Passage in the Canadian Archipelago (Fig. 2.1d) are usually covered from October to July by a mixture of landfast ice and pack ice advected from the Canadian Basin of the Arctic Ocean (Melling 2002). There is little direct river inflow in the region, but the easternmost reaches of the Mackenzie plume clearly affect surface salinity in the south-western region of the Archipelago (Fig. 2.1d).

Located between the Canadian Archipelago and Greenland, Baffin Bay is a large semi- enclosed sea with little freshwater input (Fig. 2.1e). It receives Arctic Ocean surface water through Nares Strait and Lancaster Sound, and Atlantic Water with the West Greenland current. Typically, most of Baffin Bay is covered with ice from October to July. Our sampling of polar cod juveniles was conducted primarily in the northern part of Baffin Bay

31 which is dominated by the North Water, the largest recurrent polynya in the Arctic. The North Water starts to enlarge in spring, when the flux of arctic ice through Nares Strait is blocked by the formation of an ice bridge in Smith Sound, and northerly winds push the remaining ice south along the coast of Ellesmere Island. Open water expands in April and reaches over 70 000 km2 by the end of July when the entire bay becomes ice-free (Mundy & Barber 2001). Sometimes compared to an oasis, the North Water is a biological hotspot that supports large populations of seabirds and marine mammals (Deming et al. 2002).

The Northeast Water is a large recurrent polynya that extends over the East Greenland Shelf (Fig. 2.1f). It is bounded by the Arctic Ocean to the north and by the polar front of the Greenland Sea to the east and south (Smith et al. 1990). The Northeast Water usually opens in April or May, reaches its maximum extent of ca. 44 000 km2 in July–August, and closes around September (Wadhams 1981, Smith et al. 1990, Barber & Massom 2007). Bathymetry ranges between 100 and 500 m and features two shallow banks (Ob and Belgica). No large river drains the area and surface salinity is mainly dictated by ice formation and melting. With the recent collapse of the landfast ice shelves that helped form the polynya by deflecting the arctic flux of ice, the Northeast Water has formed irregularly in recent years (Barber & Massom 2007).

2.4.2 Sampling of fish larvae and juveniles

Fish larvae and juveniles were collected in the Canadian Arctic during the ArcticNet annual mission of the CCGS Amundsen from 14 August to 2 October 2005 and from 4 September to 29 October 2006 (Table 2.1). Young fish were sampled with four different samplers: (1) a Double Square Net (DSN) consisting of a rectangular frame carrying two 6-m long, 1-m2 mouth aperture, square-conical 500-μm mesh nets; (2) a 8-m2 effective aperture, 1.6-mm mesh Rectangular Midwater Trawl (RMT); (3) a large Pelagic Trawl (PT) with mesh size decreasing from the mouth (5 mm) to the end of the net (1.6 mm); and (4) a 1-m2 aperture multi-layer sampler equipped with nine 6-m long, 333-μm mesh nets (EZNet®). All four samplers were towed obliquely from the side of the ship at a speed of 1 m s-1 (two knots) to

References 2008 Fortier & Bouchard 2008 Fortier & Bouchard This study This study This study This study data unpubl. al, et Ringuette This study This study This study This study 2006 al. et Fortier

Number of cod 170 427 169 47 54 66 1087 292 342 30 21 823

Number stations of 3 24 36 9 6 8 44 10 11 2 3 52

17 October

2 October October 23 – September – –

19 – 20 July 2 August – – 21 September 21 September 30 August 30 23 September – – – 11 September 14 September –

– – 29 September 4 Sampling dates Sampling 3 13 18 26 September 2 30 April 14 August 27 25 September 23 May

Year 2003 2005 2007 2005 2005 2006 1998 2005 2006 2005 2006 1993

Amundsen Amundsen Amundsen Radisson Pierre Amundsen Amundsen Amundsen Amundsen Victor Buynitskiy Victor Polarstern Kapitan Dranitsyn Dranitsyn Kapitan Dranitsyn Kapitan

ship Research I/B I/B RV CCGS CCGS CCGS CCGS CCGS CCGS CCGS CCGS RV

Program NABOS NABOS NABOS ArcticNet ArcticNet ArcticNet NOW ArcticNet ArcticNet ArcticNet ArcticNet NEW

Region Sea Laptev Bay Hudson Sea Beaufort Bay Baffin Passage Northwest Water Northeast Table primary sources samplingof data. and cod year, juvenile and of 2. 1 Details by region

33 a maximum depth of 90 m (DSN, RMT and EZNet®) or 150–200 m (PT). The nets of all samplers were fitted with rigid cod-ends to minimize the deterioration of fish larvae and zooplankton. TSK flow meters were mounted on all samplers to record filtered volumes.

In the Siberian Arctic, fish larvae and juveniles were collected during the Nansen and Amundsen Basins Observational System (NABOS) annual missions of the icebreaker Kapitan Dranitsyn in 2003 and 2005 (Bouchard & Fortier 2008), and the research vessel Victor Buynitskiy in 2007 (Table 2.1). Young fish were sampled with a DSN as described above.

At sea, the indistinguishable larvae and juveniles of Boreogadus saida and Arctogadus glacialis were sorted from the zooplankton samples and measured fresh (standard length, SL) before preservation in 95% ethanol. Up to 25 cod per sample were randomly selected and measured fresh. In the laboratory, all preserved fish were measured again and the fresh standard length of fish not measured fresh at sea was estimated from the preserved-length on fresh-length regression for the region of origin. The number of young cod captured, measured fresh, and aged by otolith analysis for each region is given in Table 2.2.

2.4.3 Discriminating Boreogadus saida and Arctogadus glacialis larvae and juveniles

The larvae and juveniles of Boreogadus saida and Arctogadus glacialis are almost impossible to discriminate morphologically. The two species co-occur in arctic seas, and some limited (<1%) contamination of B. saida collections by A. glacialis has been reported (Sekerak 1982). Recently, a simple genotyping method based on the microsatellite marker Gmo8, has been developed to distinguish the two species with >95% certainty (Madsen et al. 2009). To assess the contribution of A. glacialis to the assemblage of young cod, sub- sets of larvae and juveniles from five of the six regions (Laptev Sea, Hudson Bay, Beaufort Sea, Baffin Bay, and Northwest Passage) were analyzed with this method. In addition, assuming a daily deposition of increments in the otolith of A. glacialis, the age and hatch-

2 r 0.855 0.871 0.803 0.967 0.865

Intercept 4.357 1.437 9.353 2.957 3.020 . 1 - age regressions

Length Slope 0.196 0.235 0.182 0.223 0.215 as parameters of the

266

284 278 239 212 –

– – – –

(days) Age rangeAge 59 104 81 22 20

67.3 61.8 60.0

47.4 61.5 – – –

– –

Standard length range (mm) 14.4 25.5 21.0 8.4 9.0

age regression is an estimate of growth in mm d in growth of estimate an is regression age

Number aged 535 45 102 43 322

Number measured fresh 766 47 120 51 506

Total number captured 766 47 120 51 634

2006.

Region Laptev Sea Hudson Bay Beaufort Sea Northwest Passage Baffin Bay* For* 2005

regression of length of slope of length - The the on age by region. regression Number of young polar cod captured, measured fresh, and young captured, of aged Table cod by otolithmeasured well 2.2 Number polar as analysis,

date of 55 fish positively identified as A. glacialis collected in the Beaufort Sea in 2004 were determined (see Section 2.4.5 for methodology).

2.4.4 Validation of the ageing of young polar cod

The daily nature of increment deposition in the otolith of polar cod was verified using a chemical marking technique (Geffen 1992). Live juveniles collected in the 0–30 m surface layer of the North Water from 10 September to 5 October 1999 were delicately transferred to individual jars containing seawater at 0°C (representative of in situ temperature) and kept in an incubator on board the ship. The fish were fed once daily with fresh zooplankton. The light regime in the incubator was set at 12h of darkness and 12 h of light to simulate the photoperiod prevailing at the time in the region. Light intensity in the jars (21.2 μmol photon m2 s-1) was representative of daytime irradiance at depths of 30–60 m. After ca. 24 h of acclimation, juveniles were transferred for 12 h in a 400 mg/L oxytetracycline hydrochloride (OTC) seawater solution with pH re-adjusted to original value of seawater with TRIS buffer. The surviving fish were transferred back to clean seawater and normal rearing conditions resumed. Eleven polar cod juveniles survived the marking procedure. Following death or sacrifice after variable periods of rearing, these fish were measured and preserved in 95% ethanol. The lapilli were dissected from each fish and embedded in thermoplastic glue (Crystal Bond®). After polishing, the left lapillus of each fish was examined under a microscope equipped with a fluorescent light source (380 nm) to locate the OTC mark which appeared as a clear green ring. Once the OTC mark was located, the number of increments between the mark and the edge of the otolith was counted under normal transmitted light.

In some species, the increments deposited in the first days of life are sometime too narrow to be resolved under the light microscope, resulting in the underestimation of age (e.g. Campana et al. 1987). In particular, narrow initial increments linked to slow growth may occur in low-temperature environments. To verify that the initial increments in the otolith of polar cod are resolved accurately by light microscopy, counts were compared for a sub-

set of lapilli analyzed both under light microscopy and scanning electron microscopy (e.g. Jones & Brothers 1987). The left lapilli of 13 polar cod of varying age collected in 2005 in Baffin Bay and in 2008 in the Beaufort Sea were analyzed under the light microscope in 2006 and 2009 respectively. The same otoliths were re-analyzed in scanning electron microscopy (SEM) independently by the same operator in 2010. The operator could not remember the light microscopy count when interpreting the SEM image of the otolith. In preparation for SEM, the polished lapilli were etched by immersion in a 5% EDTA solution (KOH-buffered to pH 7.5) for 30 s to 2 min, then rinsed in distilled water, air dried for 24 h, and coated with gold.

2.4.5 Regional Hatch-date frequency distribution (HFD) of polar cod

Based on the aging of larvae and juveniles collected in late-summer, new unpublished HFD of polar cod are presented for five regions: Beaufort Sea, Laptev Sea, Baffin Bay, Northwest Passage, and Hudson Bay. In each region, between 51% and 96% of the fish collected were aged by otolith analysis (Table 2.2). A sub-set of polar cod stratified by length and region was selected for otolith aging. The two lapilli of each fish selected were dissected and mounted separately on microscope slides in Crystal Bond® thermoplastic glue. Each otolith was ground on its medial side on a 3-μm aluminum grit paper. The increments of one of the two lapilli (preferentially the left) were enumerated and measured by a first reader under a light microscope (1000× magnification) coupled to a camera and image analyzer system (Image Pro Plus®). To estimate aging precision (Campana 2001), an independent reading by a second reader was made on 200 otoliths from the different regions, yielding a mean coefficient of variation of 2.8%. The counts of the first reader were retained in subsequent analyses. The number of increments was strongly and linearly correlated to the radius of the otolith over the range of increments (20–284) and radii (31– 305 μm) analyzed (number = 0.906 radius - 0.881, r2 = 0.938, n = 1047, p < 0.0001).

The age of the remaining fish was estimated from their standard length using a region- specific redistribution procedure in which fish in a given 1 mm length class are randomly assigned an age according to the known age probability function for that 1 mm length class

(Kimura 1977). The hatch-date of an individual fish was determined by subtracting its age (in days) from its date of capture. The hatch-date frequency distribution (HFD) was built by tallying the number of fish hatched in the same 7-d hatch-date bin.

Two additional HFDs based on the collections of polar cod larvae in spring and early summer in the Northeast Water Polynya in 1993 (Fortier et al. 2006) and in the North Water Polynya in 1998 (M. Ringuette, Université Laval, unpublished data) respectively, were included in the present analysis (Table 2.1). Polar cod larvae were sampled with a DSN and a RMT in the Northeast Water (Fortier et al. 2006) and with a DSN in the North Water (M. Ringuette et al., unpublished data). In both studies, small larvae collected in 4- m2 aperture zooplankton nets towed vertically were included in the reconstruction of the HFD. Fish not aged by otolith analyses were aged with the redistribution method described above and the HFDs were built by classifying fish in 7-d hatch-date bins.

When HFDs from adjacent regions did not differ statistically, the data were pooled and a new HFD was calculated. This was the case for the HFDs from the Laptev and East Siberian Seas (Kolmogorov–Smirnov, p = 0.087) and the HFDs from the Lancaster Sound, Baffin Bay and North Water regions (Kolmogorov–Smirnov, p = 0.312). Henceforth, these pooled regions are referred to as Laptev Sea and Baffin Bay. To make them comparable, all HFDs were expressed in percent frequencies. The number of fish hatched in a given 7-d hatch-date bin was divided by the total number of fish collected in the region in a given year and multiplied by 100. When data from several years were available for a given region, the average regional HFD was calculated by averaging the frequencies over years.

To illustrate the differences in the pre-winter size of juveniles resulting from regional differences in the hatching season, the length frequency distribution of polar cod on 14 August was contrasted among the five regions sampled in late-summer (Northwest Passage, Baffin Bay, Laptev Sea, Beaufort Sea, and Hudson Bay). Standard length was linearly correlated to otolith radius (length = 1.547 + 0.201 radius, r2 = 0.930, n = 1047, p < 0.0001), enabling us to back-calculate the length of individual fish from the radius of the otolith on 14 August (the earliest capture date among the five regions) using the biological

intercept method (Campana & Jones 1992) and a length at hatch of 5.5 mm. The estimated length of juvenile polar cod on 14 August (mm) was converted into weight (mg) using the relationship ln weight = 3.095 ln length - 5.3 (after Ponomarenko 2000).

2.4.6 Regional long-term average surface salinity and temperature

Our objective was to relate regional differences in the hatching season of polar cod (an evolutionary trait expected to be adapted to long-term average ocean climate) to differences in the long-term average surface salinity and temperature of different oceanic regions. The World Ocean Atlas of the National Oceanographic Data Center (http://www.nodc.noaa.gov/OC5/WOA05/pubwoa05.html) provides objectively analyzed climatological monthly averages of oceanographic variables based on all available measurements between 1800 and 2005 (Boyer et al. 2006). Monthly values of salinity and temperature within each 1° latitude × 1° longitude area of each region studied were extracted from the latest version of the World Ocean Atlas (Antonov et al. 2006, Locarnini et al. 2006). For each region, monthly surface salinities in each 1° × 1° area were averaged over the months corresponding to the hatching season of polar cod (based on the HFDs) and were mapped with the Ocean Data View software (Schlitzer 2009, http://odv.awi.de). Monthly surface temperatures and salinities in the 0–10 m layer were averaged over all 1° × 1° areas in each region. To take into account that hatching does not start and end precisely with the month, mean regional temperatures over the hatching season of polar cod were calculated by attributing its monthly value to each weekly bin and then averaging the weekly values.

2.5 Results

2.5.1 Boreogadus saida versus Arctogadus glacialis

The percentage of Arctogadus glacialis in the collections of larval and juvenile cod varied from 0% in 2003 to 11% in 2005 in the Laptev Sea. It ranged from 0% to 6% among different regions of the Canadian Arctic (Table 2.3). A. glacialis was not detected in the Northwest Passage and in sub-arctic Hudson Bay, and was rare in Baffin Bay. The hatching

39 date of 55 A. glacialis sampled in the Beaufort Sea in 2004 ranged from 21 March to 24 May.

2.5.2 Validation of the daily deposition of increments in the otoliths of polar cod

Juvenile polar cod died or were sacrificed from 1 to 41 d after the otolith marking procedure. Standard length at death ranged from 29 to 62 mm. There was no significant difference (paired t-test, t = 1.027, n =11, p = 0.329) between the number of increments counted from the mark to the edge of the otolith and the number of days elapsed between marking and death (Fig. 2.2).

Light microscopy (LM) and scanning electron microscopy (SEM) revealed the same patterns in the microstructure of the lapillus, including a clear hatch mark (typical diameter around 21–23 μm) and equally-spaced growth increments (typically 1 μm wide) from the hatch mark to the edge (Fig. 2.3a). We found no evidence of small concatenated increments in the nucleus region of polar cod lapilli under SEM (Fig. 2.3b). There was no significant difference (paired t-test, t = 0.448, n = 13, p = 0.662) between increment counts from LM and SEM over the 7–98 d range of estimated ages of the polar cod analyzed (Fig. 2.4).

Region Year Number Number Arctogadus glacialis analysed amplified Number Percentage Laptev Sea 2003 32 26 0 0 Laptev Sea 2005 48 45 5 11 Hudson Bay 2005 38 38 0 0 Beaufort Sea 2006 48 47 3 6

Baffin Bay 2006 144 143 1 <1 Northwest Passage 2006 14 14 0 0

Figure 2.4 Relationship between increment counts in the lapilli of polar cod analyzed under the light microscope (LM) in 2006 and 2009 and re-analyzed under the scanning 2 electron microscope (SEM) in 2010 by the same operator (LM = 1.014 SEM – 0.027, r = 0.986, n = 13, p < 0.0001). The dashed line is the 1:1 line.

2.5.3 Length-age relationships and growth rates

The length of polar cod larvae and juveniles sampled in late-summer and early fall was linearly correlated to age (Table 2.2). Growth rate, as estimated by the slope of the regression, ranged from 0.182 mm d-1 in the Beaufort Sea to 0.235 mm d-1 in Hudson Bay. Growth was positively correlated to the average surface temperature in the region during the hatching season (Fig. 2.5).

2.5.4 Interannual variability in regional hatch-date frequency distributions

Interannual comparisons of polar cod HFDs were possible for four of the six regions studied (Fig. 2.6). With a few exceptions, the HFD in a given region was relatively consistent among years. In the Laptev Sea, the duration and shape of the prolonged hatching season (December–July) were similar in 2005 and 2007 (Fig. 2.6a). By contrast, except for one fish hatched in January, most of the polar cod sampled in 2003 were hatched between late March and early July. In the Beaufort Sea, a few fish hatched in December and January, but the main hatching started in mid February and ended in late June in both 2005 and 2006 (Fig. 2.6b). Based on the relatively few fish sampled in the Northwest Passage, the hatching season was earlier in 2006 (February to June) than in 2005 (April to early August) (Fig. 2.6c). In Baffin Bay, the hatching season in 1998 started in mid-April and ended in mid-July, but the sampling of polar cod in early fall 2005 and 2006 indicated an earlier onset with some larvae hatching as early as late February and hatching well under way in early April (Fig. 2.6d). For comparison, only 14% of the larvae were hatched before 22 May in 1998, whereas 55% and 70% hatched before that date in 2005 and 2006 respectively.

Figure 2.5 Regression of young polar cod growth rate and long-term average surface temperature during the hatching season in a given region (growth = 0.032 temperature + 2 0.236, r = 0.720, p = 0.033, n = 6). Surface temperatures were extracted from the World Ocean Atlas for the months of polar cod hatching.

2.5.5 Average hatching season in relation to regional river input

When contrasting the six regions, the average regional HFDs became shorter in duration and shifted from winter towards summer with decreasing regional freshwater input (Fig. 2.7, Table 2.4). Hatching started as early as December and January in the Laptev Sea, Hudson Bay and the Beaufort Sea where freshwater inputs are high and winter surface salinities are low (Fig. 2.7a–c). At the other end of the series, hatching was delayed until April or May in Baffin Bay and the Northeast Water where freshwater inputs are negligible and surface salinities are relatively high (Fig. 2.7e and f). In the Northwest Passage, the occasional early- hatching in February was at odds with the negligible local freshwater input. By comparison to the start of the hatching season, the end of the hatching season varied relatively little, hatching persisting until end of June to early August in all six regions (Fig. 2.7, Table 2.4).

2.5.6 Pre-winter size

Hatch-date explained most of the variability (90%) in the length attained in late-summer by polar cod juveniles (Fig. 2.8). Estimated length on 14 August ranged from <10 mm (<0.01g) in larvae hatched in July to as much as 50 mm (corresponding to 0.91 g) in juveniles hatched in December and January. The vast majority (97%) of polar cod that reached a length >35 mm on 14 August were collected in regions characterized by important freshwater input (Laptev Sea, Hudson Bay and Beaufort Sea). Among fish from the freshwater-influenced regions, 20.3% were larger than 35 mm on 14 August compared to 1.4% in the other regions.

The back-calculated length frequency distribution of polar cod juveniles on the common date of 14 August differed significantly among the five regions sampled in late fall (Fig. 2.9). As expected, length achieved in late-summer decreased with increasing lateness of the hatching season. At the two extremes of the series, the early-hatching population of Hudson Bay (January to June) averaged 33.1 mm on 14 August (Fig. 2.9a), compared to 19.2 mm

Table 2.4 Statistics of the hatching season of polar cod by region and year. SD: standard deviation.

Hatching Region Year Hatch-date season Mean Min. Max. SD duration (days) Laptev Sea 2003 12 May 10 January 4 July 24.4 175 2005 3 April 2 January 20 July 40.2 199 2007 6 April 20 December 10 July 46.8 202 Hudson Bay 2005 05 April 9 January 20 June 45.2 162 Beaufort Sea 2005 20 April 26 December 13 July 35.2 169 2006 11 April 12 January 22 July 30.6 161 Baffin Bay 1998 5 June 14 April 11 July 12.6 88 2005 22 May 26 March 31 July 20.9 127 2006 11 May 21 February 27 July 21.7 156 Northwest Passage 2005 17 June 4 April 5 August 43.2 123 2006 13 May 31 January 26 June 39.3 146 Northeast Water 1993 30 May 12 May 21 July 17.4 70

Figure 2.8 Regression of standard length on 14 August (SL) on hatch-date (HD) for polar cod collected in late-summer and early fall in regions characterized by strong river discharge (full circles, Laptev Sea, Hudson Bay and Beaufort Sea) or weak river discharge 2 (open circles, Northwest Passage and Baffin Bay). SL = 49.87 - 0.203 HD, r = 0.906, n = 1075, p < 0.0001.

for the late-hatching (April to August) population in the Northwest Passage (Fig. 2.9e). Differences in late-summer length were particularly marked between freshwater-influenced regions (27.0–33.1 mm) and regions with weak freshwater input (19.2–23.1 mm). Size in late-summer appeared unrelated or even inversely related to growth rate: the highest regional growth rate in Hudson Bay corresponded to the largest late-summer size, but size in late-summer was inversely related to growth for the four remaining regions (Fig. 2.9).

2.6 Discussion

2.6.1 The true age of the true polar cod

The overall contamination of Boreogadus saida samples by Arctogadus glacialis amounted to 9 out of 313 or 2.9%. While absent or rare in most regions, A. glacialis represented a sizable fraction of the cods in some years in some regions (6% in the Beaufort Sea in 2006 and 11% in the Laptev Sea in 2005). However, the hatching dates of A. glacialis (21 March to 24 May) fell well inside the hatching season of polar cod (January to July). Hence, the few A. glacialis in our collections could affect the shape of the estimated HFDs of B. saida slightly by adding false counts in the early-spring bins of the distribution. But contamination by A. glacialis would have little impact on the estimated start and end of the hatching season of Boreogadus saida, which is the main focus of this study. Therefore, no correction was made to the HFDs to account for the occurrence of A. glacialis in our collections.

The growth of newly-hatched fish can be slowed by suboptimal feeding or temperatures in the low end of the temperature range of a species. Under such conditions, otolith increments may be deposited at intervals longer than daily, or may become too narrow to be resolved by light microscopy (e.g. Umezawa & Tsukamoto 1991, Folkvord et al. 2004). The marking experiments confirmed the daily deposition of increments in polar cod juveniles >29 mm reared at 0°C. Furthermore, the examination of the core region of the otolith in scanning electron microscopy showed no evidence of thin increments that would not be resolved in light microscopy. These results are consistent with the observed daily deposition of increments in the otolith of larvae of the Antarctic fish Nototheniops

53 nudifrons reared at sub-zero temperature from hatching to 38 d, and growing at 0.13 mm d-1 (Hourigan & Radtke 1989). By comparison, polar cod larvae 0–70 d old grew at 0.234 mm d-1 in the Northeast Water (Fortier et al. 2006) and, in the present study, the growth of juveniles varied from 0.182–0.235 mm d-1 among the different regions. The typical width (1 µm) of the initial increments associated with such growth is resolved easily by conventional light microscopy. We conclude that the newly-hatched larvae of polar cod, a hyper specialist adapted to life at sub-zero temperatures, achieve initial growth rates that are sufficient for the daily deposition of clear increments on the otolith.

2.6.2 The winter thermal refuge hypothesis

A preliminary review of polar cod hatching season (Bouchard & Fortier 2008) indicated two hatching patterns: a short spring hatching season (May–June) centered on the ice break-up and the onset of biological production in the Northern Baffin Bay and the Greenland Sea (Sekerak 1982, Fortier et al. 2006); and a protracted winter–spring–summer hatching season (January–June/July) in the Kara and Laptev Seas, where some larvae emerge under sea ice in winter well before the spring bloom (Baranenkova et al. 1966, Bouchard & Fortier 2008). Fish larvae are visual predators that dwell in the surface photic layer. In Arctic seas in winter and early-spring, temperature under the ice cover varies with salinity from 0°C (S = 0) to ca. –1.8°C (S = 33). The motility of Atlantic cod Gadus morhua larvae (Valerio et al. 1992) and the feeding success of recently hatched polar cod (Michaud et al. 1996) are drastically reduced at temperatures < –1 oC. Bouchard & Fortier (2008) proposed that salinity-induced differences in the sub-zero temperatures that prevail under the ice could explain the observed regional differences in the hatching season of polar cod. Brackish conditions in the under-ice plume of large rivers would provide relatively warm temperatures (0 to –1 oC), allowing more rapid egg development and the motility needed for successful first-feeding in winter. In regions lacking a river plume, the extreme sub-zero temperatures (ca. –1.8°C) prevailing in winter would slow egg development and prevent successful first-feeding, leading to poor survival. In such regions hatching would be delayed until the ice break-up and the vernal warming of the surface layer.

Consistent with the winter thermal refuge hypothesis, the hatching of polar cod started in winter in regions strongly influenced by river discharge, and was delayed until spring in regions with weak freshwater input (Fig. 2.7). The hatching season was earliest in the Laptev Sea, Hudson Bay and the Beaufort Sea where freshwater influxes are large (738, 714 and 330 km3 year-1, respectively). Although the Beaufort Sea receives less than half the freshwater volume of the Laptev Sea and Hudson Bay, hatching there also started in early winter and the mean hatching date (15 April) was comparable to that in the Laptev Sea (17 April) and in Hudson Bay (5 April). This somewhat unexpected earliness of hatching in the Beaufort Sea may perhaps be linked to the presence of the stamukhi which, by damming the freshwater plume of the Mackenzie River, creates the brackish Lake Herlinveaux (Macdonald et al. 1995), an environment that could offer an early and particularly suitable winter thermal refuge for early hatchers. By contrast, with the exception of a few early hatchers in the Northwest Passage, hatching did not start until spring (April–May) in regions with little freshwater input (Northwest Passage, Baffin Bay, and the Northeast Water).

Diadromy in Gadidae is uncommon and is seen as a derived evolutionary state to exploit freshwater for reproduction (Dodson 1997). In estuaries of the western North Atlantic, a fraction of the population of tomcod Microgadus tomcod, a facultative diadromous Gadidae, migrate from estuaries to rivers to spawn under the ice in winter (Scott & Scott 1988). The polar cod, a Gadidae of similar size, share with the tomcod the challenges associated with a short season of biological production, seasonally ice-covered waters, and sub-zero temperatures in the surface layer in winter. Our study suggests that the polar cod may have developed some facultative partial diadromy that does not take the genitors into the rivers but nevertheless exploits the coastal under-ice plumes of rivers to hasten and lengthen the period of initial larval growth.

2.6.3 Winter hatching and the pre-winter size of polar cod juveniles

The winter thermal refuge hypothesis assumes that the evolutionary force driving hatching under the ice in winter is the need to maximize the pre-winter size of juveniles, so as to

55 minimize mortality over the first fall and winter of juvenile life (Fortier et al. 2006, Bouchard & Fortier 2008). Winter hatching resulted in spectacularly larger pre-winter size: by mid-August, January hatchers reach 0.9 g in weight and were typically 100–150 times heavier than July hatchers (<0.01 g). Regionally, differences in length between freshwater- influenced regions (27.0–33.1 mm) and purely marine regions (19.2–23.1 mm) translated into fish that were on average 3 times heavier in regions influenced by freshwater (0.13– 0.25 g versus 0.05–0.08 g). In juvenile fish, mortality is generally inversely related to size (see Sogard 1997 for a review). A larger size provides a survival edge through enhanced predator avoidance, resistance to starvation, and physiological tolerance: three attributes that are particularly important in fending off winter mortality (Sogard 1997). As the ice cover reaches a minimum in late-summer, polar cod fry in the surface layer become vulnerable to seabirds. These include the surface-feeder black-legged kittiwake Rissa tridactyla and several diving birds that can reach down to depths of 35–50 m, such as the dovekie Alle alle, Brünnich’s guillemot Uria lomvia and the black guillemot Cepphus grylle (Bradstreet 1982, Piatt & Nettleship 1985, Barrett & Furness 1990, Falk et al. 2000, Karnovsky & Hunt 2002). Since the migration to the deep overwintering grounds starts at lengths of 30–35 mm (Ponomarenko 2000), it can be expected that the larger the juveniles are at the end of summer, the earlier they will leave the surface layer and escape avian predation during the fall. As well, adult polar cod are likely the main predator of juvenile polar cod on the overwintering grounds (Baranenkova et al. 1966, Frost & Lowry 1984), and a large size should provide the juveniles with enhanced capacity to avoid their cannibalistic congeners (e.g. Miller et al. 1988). A feasible but logistically demanding approach to verify the size-dependence of vulnerability to avian predation and cannibalism would be to estimate the size-distribution of juvenile polar cod prey from the otoliths recovered from the gizzard of seabirds and the digestive tract of adult polar cod, for comparison to the size-distribution of the population.

Maximizing pre-winter size can be achieved by maximizing growth, by maximizing the growth season, or by maximizing both. Hatch-date explained 90% of the variance in pre- winter size (Fig. 2.8), and polar cod obviously relied primarily on maximizing the duration of the first growth season to achieve a large pre-winter size. We found little evidence that

fast growth also contributed to large pre-winter size. Regionally, with the exception of Hudson Bay where both growth and pre-winter size were highest, a larger pre-winter size was associated, somewhat paradoxically, with slower growth among the other four regions (Fig. 2.9). Hence, faster growth did not explain the larger pre-winter size in regions influenced by freshwater. Growth was rather dictated by the general surface temperature conditions prevailing in the sampling region during the hatching season and the early life of polar cod in the plankton (Fig. 2.2). This suggests that, in cold regions, the long growth season made possible by winter hatching in the thermal refuge provided by a river plume can actually overcompensate for slow early larval growth in producing large pre-winter sizes. Hudson Bay, where winter hatching combines with relatively warm surface temperatures during plankton drift, produced the largest pre-winter sizes.

2.6.4 The food of polar cod larvae under the ice in winter

A long standing tenet of the ecology of temperate and boreal fish is that the hatching of fish larvae coincides with the vernal production of their planktonic food to maximize food intake and early growth (see Leggett & Frank 2008 for a recent review). Our circum-arctic census of the hatching season of polar cod confirms that, in all regions studied, a large fraction of the larval population emerge in spring and early summer (April/May to June/July) during the short season of production of calanoid copepod eggs and nauplii, its main prey. Obviously however, wherever winter conditions allow, a fraction of the population will maximize the duration of the growth season by hatching as early as possible and well before the massive reproduction of calanoid copepods in spring. Bouchard & Fortier (2008) speculated that polar cod hatching in winter feed on the nauplii of small omnivorous copepods that reproduce all year long and on the eggs of calanoid copepods that release their eggs in winter such as Calanus hyperboreus and Metridia longa. Another possibility is that winter-hatched polar cod larvae prey on organisms such as rotifers associated with the brackish waters of river plumes and/or the microbial food web that remains active in winter in temperate and ice-covered seas (Mousseau et al. 1998, Garneau et al. 2008).

2.6.5 Climate change and the hatching season of polar cod

Changes in the phenology of populations and in the structure of ecosystems have been linked to climate change (Walther et al. 2002, Parmesan & Yohe 2003, Parmesan 2006), in particular at sub-arctic and arctic latitudes where reproduction must be synchronized with the extreme seasonality in light, temperature, and food availability (e.g. Réale et al. 2003, Gaston et al. 2005, Perry et al. 2005, Grebmeier et al. 2006). To a large extent, the response of the entire arctic pelagic ecosystem to climate change could depend on perturbation in the phenology of a few key elements of the low-diversity trophic web such as the polar cod (e.g. Tynan & DeMaster 1997).

On-going trends in arctic sea-ice cover (e.g. Bareiss & Gorgen 2005, Stroeve et al. 2007), sea surface temperature (e.g. Belkin 2009) and freshwater discharge (e.g. Peterson et al. 2006) all have the potential to alter the timing and success of the reproduction of polar cod. An earlier ice break-up, more frequent winter polynyas, a warmer surface layer, and increased river discharge would be expected to favor the feeding, growth and survival of winter hatchers (Fortier et al. 2006, Bouchard & Fortier 2008, this study), therefore shifting the hatch-date frequency distribution (HFD) of survivors sampled in late-summer and fall to the left. The present review yielded limited but interesting information on the interannual and inter-decadal variability in polar cod HFD in response to climate variability and change. In the Laptev Sea, juveniles sampled in the fall were hatched from January to July in 2005 and 2007 (Fig. 2.3a), two years characterized by record spring/summer sea-ice regression over the Siberian Shelves. By contrast, juveniles surviving to the end of summer were hatched no earlier than March in 2003, a year of closer-to-normal ice conditions. In Baffin Bay, the HFDs in 2005 and 2006 were shifted earlier by at least a month relative to 1998 (Fig. 2.3d). This inter-decadal shift is consistent with an increase of sea surface temperature of 0.47°C from 1982 to 2006, one of the most intense regional warming observed in the Arctic (Belkin 2009). Of course, although these shifts in the HFD are in the expected direction, any conclusion about the impact of climate change on the hatching season of polar cod will remain speculative until longer time series of observations are obtained in different regions of the Arctic Ocean.

2.7 Conclusion

Our review of hatching seasons of polar cod in relation to freshwater input generally supports the hypothesis that under-ice river plumes provide a thermal refuge that enables some polar cod to hatch, initiate first-feeding and grow in winter. In southeastern Hudson Bay in spring (late April to early June), newly-hatched polar cod larvae occur in the coastal zone influenced by the turbid freshwater plume of the Great Whale River that extends over the 0–5 m depth layer immediately under the ice cover (Drolet et al. 1991, Gilbert et al. 1992, Fortier et al. 1996). The seemingly euryhaline first-feeding larvae congregate in daytime in the brackish halocline (S = 5–25, T = –0.3 to –1.0oC) between the plume and the underlying marine layer, where the product of prey density by light intensity is maximum (Ponton & Fortier 1992). These observations support our assumption that polar cod larvae associate with under-ice river plumes in winter as well. However, a definitive test of the winter thermal refuge hypothesis would entail the direct verification that polar cod hatch, feed successfully, and grow in the halocline of the under-ice plumes of rivers in Arctic coastal seas in winter. The Laptev Sea, the area of the Beaufort Sea inside the stamukhi, and Hudson Bay would be particularly well suited to deploy the overwintering expeditions needed to sample polar cod larvae under sea ice in the extreme winter conditions of the Arctic Ocean.

Chapitre 3 – Spatial segregation, dispersion and migration in early stages of polar cod Boreogadus saida revealed by otolith chemistry

3.1 Résumé

La chimie des otolithes de Boreogadus saida juvéniles provenant de six régions océanographiques de l’océan Arctique a été examinée. Cinq rapports élémentaires (Li/Ca, Mg/Ca, Mn/Ca, Sr/Ca et Ba/Ca) ont été analysés par ablation au laser et spectrométrie de masse avec plasma inductif (ICP-MS) dans trois zones de l’otolithe représentant les stades d’œuf, de larve et de juvénile. La concentration de chacun des cinq éléments à la marge de l’otolithe, correspondant à une incorporation peu de temps avant la capture, est significativement corrélée à la salinité de surface au site et à la date de capture. En concordance avec ces relations et avec la distribution de certains éléments dans l’océan Arctique, la composition chimique des otolithes diffère entre les juvéniles des régions influencées par l’eau douce (golfe d’Amundsen, baie d’Hudson, mer de Laptev) et ceux des régions essentiellement marines (détroit de Lancaster, baie de Baffin, baie de Frobisher). Des analyses discriminantes incluant les cinq éléments ont apporté de l’information précieuse sur la structure de population de l’espèce et la dispersion des jeunes stades. La correspondance entre les concentrations otolithiques de Mn/Ca, Ba/Ca et les profils verticaux de Mn et Ba dissous dans la colonne d’eau pourrait reflèter la migration verticale ontogénique des Boreogadus juvéniles à la fin de l’été et au début de l’automne.

3.2 Abstract

Otolith chemistry of juvenile polar cod Boreogadus saida from six oceanographic regions of the Arctic Ocean was examined. Five elemental ratios (Li/Ca, Mg/Ca, Mn/Ca, Sr/Ca and Ba/Ca) were analysed by laser ablation inductively coupled plasma mass spectrometry (ICP-MS) in three otolith zones representing the egg, larval and juvenile stages. The concentration of each of the five elements at the edge of the otolith, corresponding to an incorporation shortly after capture, was significantly correlated to surface salinity temperature at capture site and date. In agreement with these relationships and with the distribution of some elements in the Arctic Ocean, otolith chemical compositions differed between juveniles from freshwater-influenced regions (Amundsen Gulf, Hudson Bay, Laptev Sea) and those from purely marine regions (Lancaster Sound, Baffin Bay, Frobisher Bay). Discriminant function analyses including all five elements provided valuable information on the species population structure and dispersion of early stages. The correspondence between otolith Mn/Ca, Ba/Ca and vertical profiles of dissolved Mn and Ba in the water column may reflect the ontogenetic vertical migration of juvenile polar cod in late-summer and fall.

3.3 Introduction The polar cod Boreogadus saida is extremely important in Arctic food webs (Bradstreet et al. 1986) and its response to ongoing climatic changes holds the potential to profoundly affect the entire Arctic ecosystem (Tynan & DeMaster 1997). Despite the recognition of the species crucial role and increasing efforts to understand different aspects of its biology, many key elements concerning polar cod remain uncertain. Notably, information on population structure and spatial segregation is very scarce. While no genetic structuring was detected by studies conducted with allozymes, RAPD and scnDNA (Fevolden & Christiansen 1997, Fevolden et al. 1999), more sensitive DNA markers have yet to be used. Furthermore, important questions remain unanswered in terms of spawning behavior and habitat use. Spawning occurs from November to March (Baranenkova et al. 1966) and, at least in the Amundsen Gulf region, adults spend this period aggregated near the bottom at depth greater than 140 m (Benoit et al. 2008, Geoffroy et al. 2011). In addition to these deep spawning aggregations, a number of discrete nearshore, shallow spawning grounds have been suggested in the North American (Craig et al. 1982, Thanassekos & Fortier 2012) and Siberian Arctic (see figure 1 of Fevolden & Christiansen 1997).

Progress has been made in our understanding of the ecology of polar cod early life history stages with case studies undertaken in many regions of the Arctic Ocean, including Hudson Bay (Drolet et al. 1991, Gilbert et al. 1992, Fortier et al. 1996) and the Northeast Water Polynya (Michaud et al. 1996, Fortier et al. 2006). Recently, it has been hypothesized that the main process dictating the hatching season of polar cod is not the match between first- feeding larvae and maximum abundance of their prey, but rather the maximization of pre- winter size to reduce predation (Fortier et al. 2006, Bouchard & Fortier 2008). Further, a significant proportion of polar cod hatch near estuaries during the winter, taking advantage of the thermal refuge provided by freshwater (Bouchard & Fortier 2008, 2011). Movements during each life stage of polar cod also require more systematic investigations. Seasonal vertical migrations, during which individuals gradually swim toward the bottom prior to winter, are observed in juveniles and adults (Ponomarenko 2000, Benoit et al. 2008, Geoffroy et al. 2011). Diel vertical migrations, taking place under the ice during the polar night, have also been documented (Benoit et al. 2010, Geoffroy et al. 2011). The

63 combination of seasonal migration and currents can lead to passive movement of adults over significant distances (Benoit et al. 2008) but it is unclear if polar cod perform long distance active horizontal migration.

Otolith chemistry has become a powerful tool to study fish movements including migration, natal homing and connectivity among sub-populations (Elsdon et al. 2008). The calcium carbonate matrix of fish otoliths incorporates some elements in proportion to their concentrations in the ambient environment. When matched with specific growth increments in otoliths, these data provide information on specific environments that individual fish have inhabited throughout their lives. For instance, Sr/Ca and Ba/Ca ratios in otoliths have been used to trace movements of anadromous fishes from spawning locations in rivers to ocean environments (e.g. McCulloch et al. 2005). Multi-element analyses of otolith cores have also revealed spatial structure in marine fish populations at spatial scale ranging from a few kilometers (e.g. Standish et al. 2008, Clarke et al. 2009) to thousands of kilometers (e.g. Ashford et al. 2008).

In the present study, we compared otolith elemental signatures of juvenile polar cod from six different regions across the Arctic Ocean with the main objective of testing the freshwater winter refuge hypothesis. Relationships between polar cod otolith chemistry, spatial segregation and ontogenetic migration are also investigated.

3.4 Materials and methods

3.4.1 Study areas

We collected samples from six regions of the Arctic Ocean characterized by differences in their physical and biological dynamics. The Laptev Sea, Hudson Bay and Amundsen Gulf are strongly influenced by freshwater, with river discharge of 738, 714 and 330 km3 y-1 respectively. Alternatively, Lancaster Sound, Frobisher Bay and Baffin Bay are located far from any large river and can be considered weakly influenced by riverine freshwater. Other important oceanographic features of the study areas, including the Laptev Sea polynyas, the

Beaufort Sea stamukhi and the North Water Polynya, are detailed in Bouchard and Fortier (2008, 2011)

3.4.2 Sampling

Gadid juveniles were collected from 22 August to 2 October 2005 and 15 September to 29 October 2006 during expeditions of the icebreakers CCGS Amundsen in the Canadian Arctic and Kapitan Dranitsyn in the Laptev Sea. Presumably all individuals were polar cod (Boreogadus saida) although a very low number (<1%) may have been ice cod Arctogadus glacialis (see Table 2.3). Detailed sampling methodology is given in Bouchard and Fortier (2008, 2011). In summary, juveniles collected at 38 stations (Fig. 3.1, Table 3.1) with four different samplers at depths of 0-200 m were preserved individually in 95% ethanol after being sorted from the zooplankton samples and measured fresh (standard length, SL, ranging 18-58 mm) on board. A CTD-rosette system was used to obtain vertical profiles of salinity and temperature.

3.4.3 Otolith preparation

Lapillal otoliths were removed, cleaned of adhering tissue and mounted separately on microscope slides in thermoplastic glue. One lapilli from each fish (preferentially the left) was used to estimate age and hatch date (details in Bouchard & Fortier 2008, 2011). The matching otolith from each pair was ground on its medial side on a fine-grain lapping paper in preparation for laser ablation ICP-MS analysis. The remaining preparation stages were conducted in a class 100 (ISO class 5) clean room. Otoliths were triple rinsed with Milli-Q water, sonicated for 2 min., rinsed with Milli-Q again, and allowed to dry for 24 h under the laminar flow hood. Otoliths were then transferred on clean petrographic slides using double-sided tape, insert in plastic Petri dishes and clean re-sealable zipper storage bag, and transported to the Woods Hole Oceanographic Institution Plasma Mass Spectrometry Facility.

SL n n otoliths Region Year Code Capture dates Hatch dates Age (d) (mm) stn C M E Amundsen Gulf 2005 AG05 26 - 58 2 - 14 Sep 26 Dec - 26 May 99-262 6 20 6 20 2006 AG06 29 - 55 29 Sep - 17 Oct 12 Jan - 24 May 131-278 7 22 21 31 Lancaster Sound 2005 LC05 22 - 40 22 Aug - 19 Sep 9 Apr - 24 May 102-163 2 19 0 18 2006 LC06 18 - 39 20 Sep - 21 Sep 29 Mar - 10 Jul 73-176 2 22 11 21 Hudson Bay 2005 HB 26 - 50 26 Sep - 2 Oct 25 Feb - 5 Jun 113-217 3 14 4 14 Frobisher Bay 2006 FB 40 - 50 29 Oct 6 Mar - 26 May 156-237 1 14 13 14 Baffin Bay 2006 BB 29 - 47 15 - 18 Sep 27 Feb - 30 May 108-200 3 26 15 32 Laptev Sea 2005 LS 19 - 57 14 Sep - 21 Sep 4 Jan - 9 Jul 70-260 14 36 18 36

a) b)

Edge

Edge Middle

Core Core

35 µm 35 µm

3.4.4 Otolith analysis

We determined isotope abundances (7Li, 25Mg, 48Ca, 55Mn, 88Sr and 138Ba) in the core, middle and edge zones of otoliths using a 193 nm excimer laser ablation system (New Wave Research) coupled to a Thermo Finnigan Element 2 ICP-MS (Thermo Electron Corporation). The laser software delimitated ablation zones that included a 35 μm-diameter spot centered on the otolith core, a 150-200 μm long by 35 μm wide curved line following growth increments the closest to the core (middle zone) and the close to the otolith edge (Fig. 3.2). For all analyses, laser scan speed was 5 μm s–1, repetition rate was 5 Hz and dwell time was 40 sec (30 sec for middle zones). Ablated otolith material was carried to the

ICP-MS with a carrier gas (He) and 2% HNO3 nebulized wet aerosol. Instrument blanks of

2% HNO3 were run at the beginning of each analytical session and then periodically after every 10 otolith samples. Dissolved standard reference materials (FEBS 1, Sturgeon et al. 2005) were used to correct for instrument mass bias and to determine instrument precision. -1 Both standards, digested in 2% HNO3 and diluted to a Ca concentration of 40 ug.g , were analysed after every blank. A total of 448 otolith samples were analysed over 11 analytical sessions. Detection limits were calculated by dividing 3 times the standard deviation of the blanks that were run throughout the analyses (n = 63) by the mean value of each isotope in all otolith samples and were (in %): Li (8.6), Mg (28.5), Ca (0.12), Mn (20.7), Sr (0.07) and Ba (11.4). Estimates of precision were determined by averaging within-run (n = 11) relative standard deviation of the standard (Japanese Reference) measurements and were: Li/Ca: 4.1%, Mg/Ca: 1.9%, Mn/Ca: 10.3%, Sr/Ca: 0.82% and Ba/Ca: 2.8%. A total of 18 core and 9 edge data were eliminated from subsequent analyses due to faulty positioning of the laser during sampling, or because the ablation caused significant damage to the otolith. Finally, a test for outliers found 13 core data with elemental ratios showing either anomalously high or low values, and these samples were also subsequently removed from the data set.

3.4.5 Data analysis

Elemental compositions were expressed as ratios to Ca, to account for fluctuations in the amount of material ablated (Sinclair et al. 1998). Following the results of a Box-Cox test, all elemental ratios were log10-transformed to better meet statistical assumptions of the

69 tests used (e.g. normality of errors and homogeneity of variances between groups for ANOVA). Because some cases of non-normality of errors or heterogeneity of variances remained after transformation, parametric tests and equivalent non-parametric tests were performed. Statistical analyses were performed with the software SAS© 9.2 and significance level used was α < 0.05 for all tests except for the Shapiro-Wilk test where data were considered non normal when prob < W < 0.01.

Relationships between elemental ratios, salinity, temperature, hatch date and capture date were assessed by linear regressions. T-tests (exact Wilcoxon when normality of scaled residuals was not met) were used to detect differences in elemental ratios between regions highly influenced by freshwater and those with weak freshwater input and to assess interannual differences in elemental ratios of otolith core, middle and edge zones in Amundsen Gulf and Lancaster Sound.

Among-region differences in the multi-elemental ratios were compared with one-factor multivariate analysis of variance (MANOVA). Mardia’s skewness and kurtosis tests showed multidimensionnal non-normality hence exact permutation statistics (Legendre & Legendre 1998) were used with 10 000 permutations. Discriminant function analyses with leave-one-out cross validation was used to graphically visualize the ability of otolith elemental ratios to classify juveniles to their region of origin. Analyses were performed separately for both years. Because homogeneity of variances-covariances matrix was not met (verified with chisquare test), quadratic discriminant function analysis were used.

Differences in elemental ratios between core, middle and edge zones of otolith were tested by repeated-measures ANOVA (PROC MIXED). In the case of non-normality, a non parametric test for longitudinal data was performed (Brunner et al. 2002).

3.5 Results

3.5.1 Relationships between elemental concentrations, salinity, temperature, hatch date and capture date

Elements found in the otolith edges, incorporated into the calcium carbonate matrix shortly before capture, were regressed against sub-surface (10-30 m) salinity and temperature at the station (and date) where fish were captured. The concentration of all five elements showed a significant relationship with salinity (Fig 3.3). Li/Ca, Mg/Ca, Sr/Ca and Ba/Ca were positively correlated with salinity while Mn/Ca showed a negative correlation with this variable (Fig 3.3). Concentrations of two elements, Li/Ca and Mn/Ca, were negatively correlated with temperature; no significant relationship with temperature was found for the other elements (Fig. 3.3).

Elemental ratios in the core, incorporated into the otolith shortly before hatching, were plotted against hatch date in order to detect temporal changes over the season from December to July (Fig. 3.4). Similarly, elemental ratios in otolith edges were plotted against capture date to document potential changes in otolith chemistry composition from August to October (Fig. 3.4). Core Li/Ca was negatively correlated to hatch date while core Mg/Ca was positively correlated to hatch date (Fig 3.4). Edge elemental ratios and capture date showed significant correlations with Mg/Ca, Mn/Ca, Sr/Ca and Ba/Ca but these relationships disappear when FB (with a single, late capture date) was removed (Fig. 3.4). Removing FB however brought a significant positive regression between edge Li/Ca and capture date.

3.5.2 Differences in otolith chemistry between regions with high and low freshwater input

Differences in elemental ratio between regions highly influenced by freshwater from nearby rivers (Amundsen Gulf, Hudson Bay and Laptev Sea) and weakly freshwater- influenced regions (Lancaster Sound, Frobisher Bay, Baffin Bay) were tested for the three otolith zones (Fig. 3.5). Li/Ca and Mg/Ca were higher in regions with low freshwater input

71 for middle and edge zones and core, middle and edge zones, respectively. Mn/Ca in the edge was significantly higher in freshwater-influenced regions. Sr/Ca values were higher in freshwater-influenced regions for the otolith core but higher in purely marine regions for the edge zones (t-tests or exact Wilcoxon, p < 0.0334). Ba/Ca values were higher in freshwater-influenced regions for the otolith core but higher in purely marine regions for the middle zones (exact Wilcoxon, p < 0.0112).

3.5.3 Differences in otolith chemistry among regions

Interannual differences in otolith chemistry were found for Amundsen Gulf and Lancaster Sound (t-tests or exact Wilcoxon, 0.0001 < p < 0.0411), hence each year was treated separately. For both years, multi-elemental compositions differed significantly among regions at the otolith core, middle and edge zones (MANOVA, p < 0.0008; exact permutation statistics < 0.0001), except for the middle zone in 2005 (MANOVA, p = 0.2110; exact permutation statistics = 0.2116), which had a low sample size (Table 3.1). Quadratic discriminant function analyses were used to visualize multivariate differences in elemental ratios among regions in the otolith core, middle and edge zones (Fig. 3.6). In 2005, overall jackknife reclassification success from the quadratic discriminant model was 54%, 30% and 56% for core, middle and edge, respectively (Table 3.2). Except for the middle zone which had a low sample size, these numbers are significantly higher than the 25% expected by chance. In 2006, reclassification success was 56%, 65% and 71% for the core, middle and edge, respectively (Table 3.2). The discriminant power of individual elements differed greatly between years and otolith zones (Fig. 3.6).

3.5.4 Differences in otolith chemistry among otolith zones

Repeated-measures ANOVAs and the equivalent non-parametric test showed significant differences among otolith zones for all elements (p < 0.0001) and multiple comparisons showed that all pairs were significantly differents (Fig. 3.7). The trends observed in the grouped regions are generally consistent among regions, with a few exceptions (Fig. 3.8). For each region, Li/Ca tended to increase and Mn/Ca to decrease from core to edge (Fig. 3.8). Mg/Ca generally decreased from highest in the middle, intermediate in the core, and

lowest in the edge. For Sr/Ca, the general pattern was the highest ratio in the edge, followed by core and the lowest values in the middle zone. However, this was not the case for Frobisher Bay, Lancaster Sound in 2006 (no significant differences between zones) and Hudson Bay (higher ratio in the core than in the middle). Ba/Ca ratios were much higher in the middle zones for all regions and generally higher in the edge than in the core, except for Amundsen Gulf in 2006 and Frobisher Bay where the ratios were significantly higher in the core than in the edge (Fig. 3.8).

Figure 3.5 Mean ± SE elemental ratio (µmol/mol) in core, middle and edge zones of otoliths from regions with low and high freshwater input. Asterisk above significantly higher values.

Table 3.2 Percentage of juveniles classified to each region by quadratic discriminant function analyses based on multi-elemental composition (Li, Mg, Mn, Sr, Ba) of otolith core, middle and edge zones for collection years 2005 and 2006 using jacknife leave-one- out crossvalidation. Correct classification percentage are in bold. See Table 3.1 for sampling codes.

Actual Predicted region Year Position Region AG05 HB LC05 LS 2005 core AG05 35 5 35 25 HB 0 57 14 29 LC05 42 0 47 11 LS 17 5 0 78 middle AG05 0 0 nd 100 HB 0 0 nd 100 LS 11 0 nd 89 edge AG05 50 15 0 35 HB 7 64 22 7 LC05 11 33 50 6 LS 36 3 3 58 2006 core AG06 BB IB LC06 AG06 50 23 0 27 BB 27 42 0 31 IB 0 7 72 21 LC06 14 27 0 59 middle AG06 83 6 0 11 BB 13 67 0 20 IB 0 8 69 23 LC06 0 30 30 40 edge AG06 74 16 0 10 BB 19 53 6 22 IB 7 0 93 0 LC06 10 28 0 62

3.6 Discussion

3.6.1 Effects of salinity and temperature on element incorporation in polar cod otoliths

Concentration of Li, Mg and Sr are higher in saltwater than in freshwater environments (Campana 1999). It can hence be hypothesised that the concentration of these elements in otolith edges (where elements incorporation occurred shortly before capture), should be positively correlated with salinity at the time of capture. Significant positive regressions between Li/Ca, Mg/Ca and Sr/Ca in the otolith of polar cod juveniles and salinity supported this hypothesis (Fig 3.3). Conversely, environmental concentrations of Mn and Ba are higher in freshwater than in saltwater (Campana 1999) and negative relationships should be expected between Mn/Ca and Ba/Ca in otolith edge and salinity. Mn/Ca in polar cod otolith edges was indeed negatively correlated with salinity (Fig 3.3). However, contrarily to our expectations, Ba/Ca showed a significant positive relationship with salinity (Fig 3.3).

Temperature has been shown to influence elemental incorporation in otoliths of larval and juvenile marine fishes (Bath et al. 2000, Martin et al. 2004, Martin & Thorrold 2005). In the present study, significant regressions between Li/Ca and Mn/Ca in otolith edges and sub-surface temperature at capture site and date, were found. No temperature effect was detected for the other elements (Fig. 3.3).

3.6.2 Polar cod larval ecology: a further test of the freshwater winter refuge hypothesis

We tested the freshwater winter refuge hypothesis (Bouchard & Fortier 2008, Bouchard & Fortier 2011), suggesting that polar cod larvae start to hatch in winter in freshwater- influenced regions but only later in the season in purely marine regions, by comparing otolith chemistry of juvenile polar cod collected in six regions with contrasting freshwater inputs. Since Li, Mg and Sr concentrations generally increase with ambient salinity (Campana 1999), we formulated the hypothesis that otoliths from purely marine regions should have higher concentrations of these elements than otoliths from regions highly

83 influenced by freshwater. With the exception of Sr/Ca in otolith cores, all significant differences observed in Li/Ca, Mg/Ca and Sr/Ca between purely marine and freshwater- influenced regions were in accordance with the hypothesis (Fig 3.5). On the other hand, Mn and Ba are known to decrease with salinity (Campana 1999), hence higher values of Mn/Ca and Ba/Ca should be found in otoliths from regions highly influenced by freshwater. Significant differences between region types in Ba/Ca from otolith cores and in Mn/Ca from otolith edges support this hypothesis (Fig 3.5). However, contrary to the prediction, the middle zone of otoliths from purely marine regions had significantly higher Ba/Ca than those from freshwater-influenced regions (Fig. 3.5).

One particular pattern in polar cod otolith chemistry leaded to the hypothesis that egg survival may be favored by the lower salinities characterizing the regions highly influenced by freshwater. The positive regression found between otolith core Mg/Ca and hatch date in regions highly influenced by freshwater (Fig. 3.4) could be underlined by the trend of otolith Mg/Ca to increase with salinity (Fig. 3.3) if egg survival was differently affected by salinity over the hatching season. The juveniles used in our study were in the environment for several months prior to capture and can be considered as survivors of the egg and larval cohort. Elemental signature in the otolith cores hence represents that of eggs that survived until the juvenile stage. We suggest that egg survival during the winter is enhanced by the lower salinities and concomitant higher temperatures associated with river plumes, as shown for the larvae (Bouchard & Fortier 2008, Bouchard & Fortier 2011). Survival of the eggs arriving in the environment later in the season would be less sensitive to freshwater input and a larger range of salinities and Mg/Ca would hence be represented in otolith cores (Fig. 3.4). This hypothesis could be tested by analyzing the effect of salinity on polar cod egg survival in laboratory.

3.6.3 Spatial segregation, dispersion and migration in polar cod early stages

Significant regional variability was found in the elemental composition of juvenile polar cod otoliths (Fig. 3.6). Reclassification success from the discriminant model, although low in some cases, indicates that a certain level of spatial segregation exists in polar cod early

life stages and can be detected by otolith chemistry (Table 3.2). In the present study, some reference groups were likely missing and the inclusion of other hatching regions in our analyses may have further increased classification success.

Otolith chemistry provided indirect evidence of significant displacement of eggs and early stage larvae from spawning locations. In 2006, classification success increased with the use of more recent otolith zones (56%, 65% and 71% for the core, middle and edge zone, respectively), suggesting that a significant number of larvae or early juveniles had moved into a different water mass at some stage before capture. Elemental ratios at the otolith edge represent the time shortly prior to capture (when individuals were in the same habitat) and higher classification success is reached from the discriminant models. These movements could correspond to larval dispersion, active migration or more likely a combination of both.

3.6.4 Is ontogenetic vertical migration of polar cod juveniles reflected in otolith chemistry?

Some patterns found in polar cod otoliths may reflect the ontogenetic vertical migration of the juveniles which descend progressively to their deep overwintering grounds in late- summer and early-fall (e.g. Ponomarenko 2000).

We found a large ontogenetic effect on otolith Mn/Ca that was consistent across all locations, with core values 5-10 times higher than middle or edge zones (Fig 3.8). A number of studies have also found elevated Mn/Ca in otolith cores that was apparently unrelated to dissolved Mn concentrations in the environment (Brophy et al. 2004, Ruttenberg et al. 2005, DiMaria et al. 2010). The exact mechanism generating high Mn/Ca values in otolith cores remains obscure. However outside the core zone, otolith Mn/Ca ratios are generally reflecting ambient Mn concentrations (Campana 1999). Rivers are the main source of dissolved manganese for the oceans and Mn/Ca ratios are typically higher in freshwater than in seawater (Campana 1999). In the Arctic, vertical profile of dissolved manganese is characterized by high concentrations in surface water and decreasing

85 concentrations with depth (Middag et al. 2011). Lower Mn/Ca in polar cod otolith edge zones than in middle zones (Fig. 3.8) suggest that juveniles where in waters relatively depleted in Mn compared to the larvae. This pattern may resulted from vertical migration of the juveniles, which leave surface waters before winter to reached their overwintering grounds.

Barium concentrations are higher in freshwater than in seawater and otolith Ba/Ca tends to accurately reflect ambient Ba/Ca (Campana 1999, Martin & Thorrold 2005). As in other oceanic regions, Ba in the Canadian Arctic displays a nutrient-type behavior with the highest surface concentrations observed at river mouths and typical profiles showing maximum concentration at the base of the surface layer decreasing above and below (Guay & Kenison Falkner 1997, Thomas et al. 2011). The lowest otolith edge Ba/Ca ratios found among polar cod juveniles from Frobisher Bay (Fig. 3.8), may result from their ontogenetic vertical migration. Juveniles from this region were sampled late in fall (29 October) and were likely in deep water depleted in Ba (sampling depth 200 m). The negative correlation found between otolith edges Ba/Ca and capture date (Fig. 3.4) may also reflects juvenile polar cod migration toward deeper waters in fall.

Strontium is the most widely used element in otolith chemistry studies and otolith Sr/Ca is generally positively correlated with salinity (Secor & Rooker 2000). Lower otolith edges Sr/Ca in marine regions than in freshwater-influenced regions would be consistent with the ontogenetic migration of juveniles prior to winter considering different age at capture between groups. Juveniles from freshwater-influenced regions were in average 175 days old at capture (40.2 mm) compared to 150 days old (35.5 mm) for those from purely marine regions. Since vertical migration begins at 30-35 mm (Baranenkova et al. 1966, Ponomarenko 2000), the first group was more likely to be found in deeper, saltier and Sr- enriched waters at the time of capture in fall. It is however impossible to determinate the vertical position of the juveniles collected since sampling was not stratified but integrated over the water column.

3.7 Conclusion

With the recent reduction of its ice cover, the Arctic Ocean is becoming more accessible to humans and resources exploitation projects are multiplying. Fisheries are declining worldwide and commercial exploitation of polar cod in a foreseeable future is conceivable. Being a small pelagic fish highly connected in its food web and representing a high proportion of the ecosystem biomass (Welch et al. 1992), the polar cod will likely be very sensitive to harvesting (Cury et al. 2000, Shannon et al. 2000, Smith et al. 2011). This vulnerability would be superimposed to the direct effects of climate change on the species, which alone can strongly modify the ecosystem. A better understanding of polar cod population structure, reproductive strategy and migratory behavior would help in the prediction of the response of the Arctic marine ecosystems to climate change and in the management of natural resources of the Arctic Ocean. Our study of the otolith chemistry of juvenile polar cod from six regions around the Arctic is the first to clearly show spatial segregation in the species, a step toward a finer description of population structuring. Salinity had an effect on otolith incorporation of the five elements analysed, which allows to detect freshwater influence and the winter thermal refuge provided by river plumes. Considering vertical profiles of some elements in the Arctic Ocean, notably Mn and Ba, we found that otolith chemistry may reflect the ontogenic vertical migration of juvenile polar cod. We think other migration and dispersion patterns of the species could be studied with the otolith chemistry procedure used in the present study, especially if joined with water masses characterization for the analysed elements and ideally to experimental quantification of the effects of environmental and physiological parameters on otolith incorporation of each element.

Chapitre 4 – The nucleus of the lapillar otolith discriminates the early life stages of Boreogadus saida and Arctogadus glacialis

4.1 Résumé

Boreogadus saida et Arctogadus glacialis sont sympatriques sur les plateaux continentaux de l’océan Arctique. Les larves et les juvéniles des deux espèces sont similaires et la discrimination basée sur la morphologie et la pigmentation est incertaine. Nous présentons un critère de discrimination basé sur la différence dans la taille du noyau de l’otolithe lapillus entre des Boreogadus (n = 441, longueurs standards de 4.2 à 55.0 mm) et des Arctogadus (n = 82, longueurs standards de 8.6 à 45.5 mm) identifiés génétiquement. Le produit du plus court et du plus long diamètres du noyau (SN×LN) était 58% plus grand chez Arctogadus que chez Boreogadus. La régression logistique (Ln [p/(1-p)] = 0.02687 SN×LN - 17.5466), où p est la probabilité qu’un poisson soit Arctogadus, a correctement réassigné 501 des 523 poissons (96%) utilisés pour construire le modèle à leur espèce génétiquement identifié (99% des Boreogadus et 80% des Arctogadus). La même régression a correctement classifié 97% des 189 poissons échantillonnés en 2002 et 2003 et non utilisés pour construire le modèle (99% des Boreogadus et 89% des Arctogadus).

4.2 Abstract

Polar cod (Boreogadus saida) and ice cod (Arctogadus glacialis) are sympatric on continental shelves of the Arctic Ocean. The larvae and early juveniles of the two species are similar and discrimination based on morphology and pigmentation is uncertain. We present a discrimination criterion based on the difference in lapillus nucleus size between genetically identified B. saida (n = 441, 4.2 to 55.0 mm standard length) and A. glacialis (n = 82, 8.6 to 45.5 mm). The product of the shortest and longest diameters of the nucleus (SN×LN) was 58% larger in A. glacialis than in B. saida. The logistic regression (Ln [p/(1- p)] = 0.02687 SN×LN - 17.5466), where p is the probability that the fish is A. glacialis, correctly reassigned 501 of the 523 fishes (96%) used to build the model to their genetically determined species (99% of B. saida and 80% of A. glacialis). The same regression correctly classified 97% of 189 fish sampled in 2002 and 2003 and not used in building the model (99% of B. saida and 89% of A. glacialis).

4.3 Introduction

The polar cod (Boreogadus saida) is a pivotal species in the food web of Arctic seas (e.g. Welch et al. 1992). A number of studies have recently reported on the hatch date frequency distribution (Bouchard & Fortier 2008, Bouchard & Fortier 2011), trophodynamics (Michaud et al. 1996, Walkusz et al. 2011), growth (Thanassekos & Fortier 2012) and survival (Fortier et al. 2006, Thanassekos et al. 2012) of the larvae and juveniles of polar cod. Field studies generally assume that the vast majority of young gadids sampled in arctic waters are polar cod Boreogadus saida. However, the post-yolk-sac larvae and juveniles of B. saida collected in Arctic seas are impossible to distinguish from those of the sympatric ice cod Arctogadus glacialis based on morphological criteria such as shape, myomere counts, or fin ray counts (Madsen et al. 2009, Bouchard & Fortier 2011). From ca. 10 to 20 mm in length, pigments on the ventrolateral surface of the gut tend to be larger and more ramose in A. glacialis than in B. saida, which may provide some indication of the species, with the caveat that intermediary pigment patterns are found in both species (Suzuki 2013). Hence, the larvae and juveniles of the two species are almost indistinguishable except through molecular genetics. Genetic markers have been developed to discriminate the two species (Madsen et al. 2009, Nelson et al. 2013). Based on these, A. glacialis may represent from 0 to 11% of the larval and juvenile fish collected in the field and identified as B. saida depending on region and season (Bouchard & Fortier 2011). Hence, misidentification may represent a source of bias when assessing the hatch date frequency distribution, early growth and survival of the two species. Molecular genetics to discriminate the young stages of the two cod species are time-consuming and expensive. Here, we present a method to discriminate the early life stages of the two species based on interspecific differences in the size of the nucleus of the lapillar otolith.

4.4 Materials and Methods

4.4.1 Larval fish sampling

Fish larvae and juveniles were collected during the annual expedition of the research icebreaker Amundsen to southeastern Beaufort Sea (Arctic Ocean). The young gadids used

91 in the present study were sampled under the ice and in open water in 2002, 2003, 2004 and 2008, using different plankton nets. The standard length (SL) of fish larvae and juveniles was measured fresh on board before preservation in 95% ethanol. A stratified subsample of gadid larvae was assembled for species identification by randomly selecting individuals from predetermined length-classes in each of the 4 years (Table 4.1).

4.4.2 Species determination based on genetic analyses

All selected fish from all years (n = 712) were assigned to species by amplifying locus Gmo8 following Madsen et al. (2009). In their analysis, Gmo8 alleles ranged in size from 122 to 134 base pairs (bp) in Boreogadus saida and from 146 to 342 bp in Arctogadus glacialis to the exception of two specimens out of 136 that presented 130 bp alleles (Madsen et al. 2009). In the present study, Gmo8 alleles in the range 109 to 142 were assigned to B. saida and from 161 to 321 bp to A. glacialis.

As part of a separate study of genetic variability in arctic gadids, 295 of the 523 fish selected in 2004 and 2008 were genotyped at 19 microsatellite loci (including Gmo8), following Nelson et al. (2013). The microsatellite analyses provided an opportunity to compare identification based on the 18 microsatellites (excluding Gmo8) with that based on Gmo8 only. The GENETIX software was used to cluster the 295 fish based on the factorial correspondence analysis of their microsatellite data (Belkhir et al. 2004).

In cases where the two methods assigned a fish to different species, a segment of the cytochrome b gene (Cytb) was sequenced to elucidate the ambiguity. Cytb was amplified with the forward primer CytbF GGCTGATTCGGAATATGCAYGCNAAYGG and the reverse primer CytbR GGGAATGGATCGTAGAATTGCRTA NGC RAA under PCR conditions of 95°C for 3 min, followed by 30 cycles of 95°C for 20 s, 50°C for 20 s, 72°C for 45 s, followed by 5 min at 72°C.

Species/Year Number Capture dates Standard length (mm) SN×LN (µm2) mean ± standard mean ± standard deviation (range) deviation (range) Boreogadus saida 2002 131 22 Sep - 14 Oct 29.0 ± 7.4 (15.8 - 45.0) 454 ± 104 (237 - 723) 2003 30 30 Sep - 28 Dec 39.2 ± 8.2 (25.1 - 51.5) 494 ± 105 (281 - 712) 2004 213 29 Apr - 12 Sep 22.8 ± 13.9 (4.9 - 55.0) 470 ± 91 (197 - 711) 2008 228 10 May - 03 Aug 13.2 ± 6.3 (4.2 - 29.4) 416 ± 96 (213 - 685) Arctogadus glacialis 2002 14 23 Sep - 14 Oct 33.7 ± 4.2 (26.4 - 40.9) 740 ± 151 (396 - 943) 2003 14 30 Sep - 23 Dec 37.8 ± 3.9 (33.3 - 45.5) 805 ± 122 (593 - 1119) 2004 52 19 Apr - 12 Sep 25.4 ± 8.2 (8.6 - 38.5) 797 ± 125 (442 - 1097) 2008 30 11 May - 02 Aug 18.6 ± 5.9 (8.7 - 31.2) 720 ± 136 (406 - 1016)

4.4.3 Otolith Analysis

The two lapillar otoliths of each fish were dissected, mounted with their flat-concave side down on a microscope slide in Crystalbound® thermoplastic polymer, and polished in the sagittal plane with 0.5-µm metallurgic lapping film until the hatch mark delineating the nucleus appeared clearly. Measurements were taken under a light microscope (1000× magnification) coupled to a camera and image analyzing system (Image Pro Plus®). The shortest (SN) and the longest (LN) diameters of the nucleus were measured on the left lapillus of each fish (Fig. 4.1). The product SN×LN was used as an index of otolith nucleus area.

The SN×LN index for the 523 fish sampled in 2004 and 2008 and identified by genetics was used to build a logistic regression of the form ln [p/(1-p)] = a + b(SN×LN) that estimates the probability p of a fish belonging to A. glacialis (or the probability 1-p of belonging to B. saida) for a given value of SN×LN (e.g. Scherrer 2009). Such a model should be validated with data other than those on which it is based. Thus, the logistic regression was used to classify 189 fish sampled in 2002 and 2003 and identified by Gmo8, which were not used in building the logistic regression.

4.5 Results

4.5.1 Molecular identification of the two species

The factorial correspondence analysis of the 18 microsatellites (excluding Gmo8) grouped the 295 fish into a tight cluster of 250 Boreogadus saida and a looser cluster of 45 Arctogadus glacialis (Fig. 4.2). Seventeen fish identified as B. saida by Gmo8 associated with the A. glacialis cluster (Fig. 4.2). Cytb was successfully amplified and sequenced for 14 of these ambiguous fish, all of which corresponded to the reference sequence for A. glacialis found in GenBank. Given that 14 out of 14 amplified Cytb corresponded to A. glacialis, the non-amplified three were most likely A. glacialis as well, and the 17 ambiguous fish were reclassified as A. glacialis. Similarly, 5 fish assigned to A. glacialis by Gmo8 grouped with the B. saida cluster. Cytb was successfully amplified in 4 of these 5

95 ambiguous fish and corresponded to B. saida in GenBank. All 5 fish were classified as B. saida.

Pooling the 4 years, molecular analyses identified 602 (85%) of the 712 young gadids as Boreogadus saida and 110 (15%) as Arctogadus glacialis (Table 4.1). The length range of both species over the 4 years spanned from the newly hatched larval stage to the late juvenile stage.

4.5.2 Classification of genetically identified fish using the area of the lapillus nucleus

In each of the 4 years, the average index of lapillus nucleus area (SN×LN) was larger in Arctogadus glacialis than in Boreogadus saida (t-tests, p < 0.0001, Table 4.1). Pooling all 4 years, SN×LN was 58% larger on average in A. glacialis (771 ± 136 µm2) than in B. saida (447 ± 100 µm2). The statistical distribution of nucleus area overlapped slightly between the two species (Fig. 4.3).

Average nucleus area varied among years in both Boreogadus saida (ANOVA, p < 0.0001) and Arctogadus glacialis (p = 0.048), and tended to increase with the mean standard length of fish, which in turn increased as captures took place later in the year (Table 4.1). At the individual level, nucleus area increased slowly but significantly with fish length in both species (Fig. 4.3). The rate of increase with length did not differ significantly (ANCOVA, p = 0.868) between B. saida (2.7 µm2 mm-1) and A. glacialis (2.9 µm2 mm-1).

To provide a sufficient number of the relatively rare Arctogadus glacialis, the 523 fish sampled in 2004 and 2008 (441 Boreogadus saida and 82 A. glacialis according to genetics) were pooled to build the logistic regression relating otolith nucleus area to species (Fig. 4.4). The model and parameter estimates of the regression Ln [p/(1-p)] = 0.02687 SN×LN - 17.5466, where p is the probability that a specimen is A. glacialis, were highly significant (p < 0.0001). When the value SN×LN = 653 µm2 corresponding to p = 0.5 was used as a threshold to reassign specimens to species, otolith nucleus area correctly classified 501 (96%) of the 523 fish identified by prior genetic analysis for 2004 and 2008,

Method Year Genetic species Logistic regression (otolith nucleus area) B. saida A. glacialis Gmo8 2004+2008 B. saida 191/191 (100%) 0/191 (0%) A. glacialis 4/37 (11%) 33/37 (89%)

18 microsatellites 2004+2008 B. saida 244/250 (98%) 6/250 (2%) A. glacialis 12/45 (27%) 33/45 (73%)

Gmo8 2002 B. saida 130/131 (99%) 1/131 (1%) A. glacialis 2/14 (14%) 12/14 (86%)

Gmo8 2003 B. saida 29/30 (97%) 1/30 (3%) A. glacialis 1/14 (7%) 13/14 (93%)

Total All years B. saida 594/602 (99%) 8/602 (1%) A. glacialis 19/110 (17%) 91/110 (83%)

100

i.e. the years used to build the model (Table 4.2). Identification based on otolith area was more accurate for B. saida (99%) than for A. glacialis (80%).

Otolith nucleus area correctly assigned 97% (99% of Boreogadus saida and 89% of Arctogadus glacialis) of the 189 fish collected in 2002 and 2003 (pooled data), which were not used in building the logistic regression. Reclassification success and errors were similar between the two years (Table 4.2). Combining all years, otolith nucleus area correctly assigned 96% (99% of B. saida and 83% of A. glacialis) of the 712 fish.

4.6 Discussion

Using easily identified adults, Madsen et al. (2009) found that most Boreogadus saida (96%) are homozygote bearing alleles 130 bp at Gmo8, whereas most Arctogadus glacialis (90%) are heterozygote at that locus. In their study, Gmo8 misclassified none of the 97 B. saida as A. glacialis. But two of their 136 A. glacialis (1.5%) bore the 130 bp allele homozygote signature typical of B. saida and therefore could be misidentified as B. saida using Gmo8 as the sole identification criteria. In the present study, assuming that the 18- microsatellites analysis buttressed by Cytb determination provided the correct identification, Gmo8 erroneously assigned 5 of 250 true B. saida (2%) to A. glacialis, compared to 0% in Madsen et al. (2009). By contrast, Gmo8 misidentified as much as 38% (17 out of 45) of A. glacialis as B. saida, compared to 1.5% in Madsen et al. (2009). Given the relative scarcity of A. glacialis in our collections (15%), this error rate implies that, in samples from the Beaufort Sea region, about 6% of the fish (17/17+245) identified as B. saida based on Gmo8 could be A. glacialis.

A first assumption in our approach to discriminate Boreogadus saida and Arctogadus glacialis based on otolith nucleus area, is that the fish used in building the logistic regression were accurately identified to species by molecular analyses. The 18- microsatellites and Gmo8 analyses diverged on the identification of 22 fish out of 295. Cytb was successfully amplified in 17 of these 22 ambiguous fish, and in 100% of the cases confirmed the identification by the 18-microsatellites analysis. Hence, we are confident that

101 out of the 523 fish sampled in 2004 and 2008 and used in building the logistic regression, 295 were correctly identified to species by the correspondence analysis of the 18 microsatellites. Assuming that the proportions of B. saida (84.7%) and A. glacialis (15.3%) were the same in the remaining 228 fish determined by Gmo8 only, 4 (2%) of the 193 B. saida were likely misidentified as A. glacialis and 13 (38%) of the 35 A. glacialis as B. saida. Thus, we estimate at 3.3% (17/523) the overall error in the molecular identification of the fish used in building the logistic regression. Curiously, while molecular misidentification should be more frequent in the fish analysed by Gmo8 only, identification by the logistic regression agreed better with the Gmo8 identification (100% for B. saida and 89% for A. glacialis) than with the 18-microsatellites analysis (98% for B. saida and 73% for A. glacialis) (Table 4.2).

A second assumption of our approach is that nucleus area is fixed at hatch and does not vary afterward. In both species, a slow but statistically significant increase in nucleus area with increasing fish length seemed to invalidate this assumption. A first potential explanation for this increase is that the nucleus grows after hatch, an interpretation that runs contrary to the notion that the otolith grows by the successive addition of daily layers onto a fixed and static nucleus (e.g. Campana & Neilson 1985). A more plausible interpretation is that a large nucleus is associated with a large size at hatch, which is selected for during the early life as reported in several studies (Vigliola & Meekan 2002, Raventos & Macpherson 2005, D'Alessandro et al. 2013). We conclude that the apparent increase in lapillus nucleus with increasing length reflects the selective survival of the larger fish at hatch rather than the implausible growth of the nucleus.

A final assumption of our approach to discriminate Boreogadus saida and Arctogadus glacialis based on otolith nucleus area is that the same logistic regression can be used for other regions and years. In the Beaufort Sea, the logistic regression built with fish sampled in 2004 and 2008 reclassified the fish sampled in either 2002 or 2003 with the same or slightly higher success (Table 4.2). Factors such as temperature (e.g. Høie et al. 1999) and pCO2 concentration (e.g. Maneja et al. 2013) may affect otolith size at hatching. Hence, the precise parameters of the logistic regression may change for other regions of the Arctic

102

Ocean, in which case a re-estimation of the parameters of the logistic regression may be advisable.

Typically, collections of larval and juvenile Boreogadus saida from arctic continental shelves comprise from 0 to 11% of Arctogadus glacialis misidentified as B. saida (Sekerak 1982, Bouchard & Fortier 2011). In the present study, two simple conservative properties of the lapillus nucleus, the longest and the shortest diameters, allowed to correctly classify 99% of B. saida and 83% of A. glacialis ranging in development from hatching to the juvenile stage. Hence, the analysis of the otolith nucleus has the potential to reduce the error in the discrimination of the two species to negligible levels. For example, by correctly identifying 8 of the 10 A. glacialis in a sample of 100 arctic gadids assumed to be B. saida, our approach would bring the error from 10% to about 2%. In studies of the early growth and hatch date frequency distribution of B. saida, the otoliths of several hundred larvae and juveniles are typically analysed (Fortier et al. 2006, Bouchard & Fortier 2008, Bouchard & Fortier 2011), providing a large sample size to assess the frequency of occurrence of A. glacialis by adding two simple measurements of the nucleus. We conclude that our otolith- based method constitutes an inexpensive alternative to molecular analyses for discriminating B. saida and A. glacialis. Reliable identification should help elucidate the early life of A. glacialis and the interactions between the two species during planktonic life.

103

Chapitre 5 – Compared early life history of sympatric polar cod Boreogadus saida and ice cod Arctogadus glacialis in southeastern Beaufort Sea

5.1 Résumé

Les larves et les juvéniles de Boreogadus saida et d’Arctogadus glacialis cohabitent sur les plateaux continentaux arctiques et sont morphologiquement indistinguables. Les deux espèces ont été échantillonnées dans le sud-est de la mer de Beaufort d’avril à août 2004 et 2008. Un sous-échantillon stratifié de 980 des 10 587 gadidés collectés a été identifié à l’espèce par la génétique ou la taille du noyau de l’otolithe lapillus. Les Arctogadus identifiés étaient plus longs à l’éclosion (6.9 vs 5.5 mm en 2004; 6.2 vs 4.9 mm en 2008) et systématiquement plus longs que Boreogadus d’avril à juin. Selon la taille au mois de capture, 5.8% des gadidés ont été assignés à Arctogadus en 2004 et 5.1% en 2008. Les deux espèces partagent la même saison d’éclosion de mars à juillet, avec un pic en avril- mai pendant la production maximale d’algues de glace. Sous la glace en avril-mai, Arctogadus était associé avec l’interface glace-eau tandis que Boreogadus était distribué dans les quarante premiers mètres au moins. En eau libre et partiellement libre de glace de juin à août, les deux espèces étaient principalement distribuées de la surface à 40 m. Les distributions spatiotemporelles estivales de Boreogadus et d’Arctogadus dans la mer de Beaufort étaient similaires, avec les deux espèces présentes à 76% des 96 stations. À des longueurs <15 mm, Boreogadus s’alimentait principalement de nauplii de Pseudocalanus et Arctogadus de nauplii de cyclopoïdes. À des longueurs >25 mm, les deux espèces partagent le même large spectre de proies, les grands calanoïdes Calanus glacialis and C. hyperboreus fournissant >50% de l’apport en carbone. Les taux de croissance variaient entre 0.161 et 0.176 mm j-1 et n’étaient pas significativement différents entre les espèces. Les pentes des régressions capture-à-l’âge indiquaient un taux de mortalité plus élevé pour Boreogadus que pour Arctogadus. Une plus grande taille à l’âge de l’éclosion à la métamorphose pourrait procurer à Arctogadus un certain avantage de survie sur Boreogadus.

105

5.2 Abstract

The larvae and early juvenile stages of polar cod Boreogadus saida and ice cod Arctogadus glacialis co-occur on Arctic shelves and are morphologically undistinguishable. The two species were sampled in southeastern Beaufort Sea from April to August of 2004 and 2008. A stratified subset of 980 of the 10587 gadids collected were identified to species by genetics and/or the size of the nucleus of the lapillar otolith. Identified A. glacialis were longer at hatch (6.9 vs 5.5 mm in 2004; 6.2 vs 4.9 mm in 2008) and consistently larger than B. saida from April to June. Based on size in month of capture, 5.8% of the cods were assigned to A. glacialis in 2004 and 5.1% in 2008. The two species shared the same hatching season from March to July, peaking in April-May during maximum production of ice microalgae. Under the ice in April and May, A. glacialis was associated with the ice- water interface while B. saida distributed over at least the top 40 m. In partially ice covered and open waters from June to August, both species were distributed primarily from the surface to 40 m. The summer spatiotemporal distribution of B. saida and A. glacialis in the Beaufort Sea was similar with both species occurring at 76% of 96 stations. At lengths <15 mm, B. saida preyed primarily on Pseudocalanus nauplii and A. glacialis on cyclopoid nauplii. At lengths >25 mm, the two species shared the same wide spectrum of prey, the large calanoids Calanus glacialis and C. hyperboreus providing >50% of the carbon intake. Growth rates ranged from 0.161 to 0.176 mm d-1 and were not significantly different between species. Slopes of the catch-at-age regressions indicated a higher mortality rate in B. saida than in A. glacialis. A larger size at age from hatching to metamorphosis may provide A. glacialis with some survival advantage over B. saida.

106

5.3 Introduction

The abundant Boreogadus saida (polar or arctic cod) and the relatively uncommon Arctogadus glacialis (ice cod) co-occur in arctic seas. Known interactions between the two gadids include predation on B. saida by A. glacialis (Coad et al. 1995) and possible resource competition between adults in NE Greenland (Christiansen et al. 2012). Despite its low abundance relative to B. saida, A. glacialis can be disproportionately important in the feeding of marine mammals at specific times and location. In years of low B. saida abundance, A. glacialis can be a replacement prey in the diet of narwhals (Monodon monoceros) summering near Pond Inlet (Finley & Gibb 1982). In summer 1998, A. glacialis accounted for 82% of the biomass in the stomach contents of adult ringed seals (Pusa hispida) from Grise Fiord (Holst et al. 2001).

Given its importance in the arctic marine food web (Bradstreet et al. 1986), the early life history and reproductive strategy of Boreogadus saida are relatively well documented. Boreogadus saida spawn in late fall and early winter (e.g. Baranenkova et al. 1966, Craig et al. 1982), near shore (Craig et al. 1982, Fevolden & Christiansen 1997) as well as in deeper water (Benoit et al. 2008, Geoffroy et al. 2011). The buoyant eggs rise to the ice-water interface (Graham & Hop 1995) and hatch after an incubation time that depends on temperature but differs between studies: 26-35 days at 0°C (Aronovich et al. 1975) and 58 days at 1.5°C (Graham & Hop 1995). Rass (1968) suggested an incubation time between 45 and 90 days at temperature typical of Arctic waters. Documented length at hatch range from 3.5 mm to 6.5 mm (Baranenkova et al. 1966, Aronovich et al. 1975, Altukhov 1981, Michaud et al. 1996, Ponomarenko 2000). In the laboratory at 1.5°C, exogenous feeding began at age 12-14 d and the large yolk sac was completely absorbed 18-20 d post hatch (Aronovich et al. 1975). Average growth rates range from 0.18 to 0.24 mm d-1 in different Arctic seas (Bouchard & Fortier 2011), but an individual-based model of the early growth of larvae sampled in two polynyas around Greenland suggests rates varying from 0.12 to 0.36 mm d-1 (Thanassekos et al. 2012). Metamorphosis from planktonic larvae to pelagic juveniles with improved swimming ability occurs from 27 to 35 mm (Baranenkova et al. 1966). Depending on their hatch date and region of origin, larvae and juveniles range in

107 size from 7 to 57 mm in August (Bouchard & Fortier 2011) at the onset of the migration of juveniles from the surface layer to their deep overwintering habitat (Graham & Hop 1995).

Arctogadus glacialis is now considered to include the synonymised A. borisovi, making Arctogadus a monotypic genus (Møller et al. 2002, Jordan et al. 2003). A. glacialis is found mainly on Arctic continental shelves, often in coastal habitats, and sometimes in brackish waters (Nielsen & Jensen 1967, Aschan et al. 2009). The species is believed to spawn in winter (e.g. Süfke et al. 1998) but contradictory observations point to summer spawning (Jordan et al. 2003, Aschan et al. 2009). The larval and early juvenile stages of A. glacialis are morphologically similar to, and often confounded with those of B. saida (Madsen et al. 2009, Bouchard & Fortier 2011, Bouchard et al. 2013). As a result, the early life of A. glacialis remains poorly documented. The eggs and newly-hatched larvae remain undescribed, although the eggs have been reported to be demersal (Fahay 2007). Cannibalism can be intense with juveniles and young adults making over 90% of the gut content of adults in summer in Cambridge Bay (Boulva 1970).

In the present study, molecular genetics and a recently developed otolithometric method are used to discriminate Boreogadus saida and Arctogadus glacialis in a subset of the larvae and juveniles sampled in southeastern Beaufort Sea in 2004 and 2008. Based on differences in size at date, all remaining gadids are assigned to either species. The early life histories of the two species, including spatiotemporal and vertical distributions, hatching season, length at hatch, diet, growth, and mortality are compared.

5.4 Materials and Methods

5.4.1 Study area

The main physiographic features of southeastern Beaufort Sea are the deep Canada Basin in the northwest, the shallow Mackenzie Shelf influenced by the freshwater plume of the Mackenzie River in the south, the Amundsen Gulf that connects the Beaufort Sea to the Canadian Archipelago in the southeast, and the shelf of Banks Island in the northeast (Fig. 5.1). Typically, ice formation starts in October at the coastal margins of Amundsen Gulf

108

and ice consolidation over the entire region takes place in December with the connection of landfast ice to the perennial Beaufort Sea ice pack west of Banks Island (Galley et al. 2008). In winter, the circum-arctic flaw lead separates the landfast ice and the mobile central ice pack (Arrigo & van Dijken 2004, Lukovich & Barber 2005). On average since 1979, sea ice has started to retreat in early June as the flaw lead enlarges to form the Cape Bathurst Polynya complex (Barber & Hanesiak 2004). Generally, the region becomes ice- free by August or September (except for the northern sector) and landfast ice starts forming again in October in the shallow shelf, reaching its maximum thickness in March. The extent and persistence of open water varies considerably from year to year (Arrigo & van Dijken 2004).

Weekly percent ice cover in 2004 and 2008 was obtained from the Canadian Ice Service IceGraph 2.0.5 application for the Mackenzie Shelf, Banks Island and western Amundsen Gulf predetermined areas (Fig. 5.1). Percent ice cover for the overall 194 350 km2 sampling region was calculated as a weighted average taking into account the relative surface of the three areas (61%, 21% and 18% respectively).

5.4.2 Ichthyoplankton sampling

In 2004, zooplankton and fish larvae were sampled in southeastern Beaufort Sea as part of the overwintering expedition of the research icebreaker CCGS Amundsen from October 2003 to August 2004 (Fig. 5.1). From 4 February to 1 June, sampling was conducted under the ice at two stations near the icebreaker fixed position (70°N, 126°W) in the landfast ice of Franklin Bay. The first ice station (225 m depth) was located 450 m east of the icebreaker position and the second station (110 m depth) 18 km to the WSW of the ship, off the mouth of the Horton River (Fig. 5.1). The under-ice sampler consisted of a rectangular metal frame carrying side-by-side two 6-m long, 1-m2 mouth aperture, square-conical nets with 750-μm mesh. This Double Square Net (DSN) was towed at ca. 1 m s-1 between two holes in the ice separated by a distance of 300 m, using a Bombardier BR180® tractor (Drolet et al. 1991). Once a week during this period, a first horizontal 5-min tow was completed with two freewheeling spherical buoys mounted on the frame to maintain the

109 sampler at the ice-water interface so as to sample the 0.5 to 1.5 m depth interval. After removing the buoys and lowering the sampler to a depth of 40 m, a second 5-min oblique tow was completed from 40 m to the surface.

From 1 June to 10 August 2004, sampling in partially ice-covered to open waters was conducted from the Amundsen over a grid of stations covering the study area (Fig. 5.1). Three different samplers were used in ice concentration ranging from 0 to 80%. The DSN was equipped with one 750 μm-mesh net throughout the sampling period while the mesh of the other net increased from 200 μm in June to 500 μm in July and August to match the growth of the young fish. The sampler was deployed at least once at all the open-water stations in a single oblique tow down to a depth of 81 ± 31 m (mean ± standard deviation, excluding 7 deep casts from 230 to 400 m), and a cable angle of 60º on the horizon.

At selected stations, a 1-m2 aperture EZNet® multi-layer sampler equipped with nine 6-m long, 200-μm mesh nets (333-μm in August) was deployed in a single oblique tow down to depths ranging from 20 to 500 m depending on the station. The nets were opened and closed sequentially starting at maximum cast depth. The nine sampling layers were tailored according to maximum cast depth and varied in thickness from as little as 1 m for shallow casts and up to 60 m for deep casts. To reconstruct the vertical distribution, each fish was attributed a capture depth equal to the mean depth of the sampling layer in which it was captured.

Larger juveniles were captured with an 8-m2 effective aperture, 1600-μm mesh Rectangular Midwater Trawl (RMT) deployed obliquely in shallow double oblique tows (from 25 to 90 m depths) or single deep oblique tows (100 to 530 m depths).

From mid-October 2007 to August 2008, the Amundsen remained mobile in the pack ice and open waters of the Cape Bathurst polynya and Amundsen Gulf (Fig. 5.1). From 2 April to 3 August, sampling in partially ice-covered and open waters was conducted over a grid of stations using a 500-μm mesh DSN (82 ± 18 m, n = 43) and a 1600-μm mesh RMT (86 ± 15 m, n = 20) both deployed in single oblique tows.

110

Fish larvae collected by other zooplankton nets towed vertically (1 m2 square-conical nets and 0.5 m2 aperture Hydrobios® multinet sampler) were included in the present analysis. The number of individuals per unit volume of water was estimated for the DSN collections in June-August 2004 and May-August 2008. The volume of water filtered through the net was estimated from the product of the mouth area of the net opening area (2 m2), ship speed (0.5 m s-1), and tow duration.

Under the ice in 2004, the DSN horizontal tows at the ice-water interface sampled a few relatively large larval gadids, while the oblique tows captured large numbers of small larvae (Table 5.1). In both years, the oblique DSN deployed at all stations collected most gadid larvae and early juveniles. The EZNet multi-layer sampler and the RMT used from June to early August 2004 collected larvae of all sizes. Small gadids were also collected in zooplankton ringnets and 1 m2 square nets deployed vertically through the ship’s moon pool in the landfast ice of Franklin Bay, or from the side in ice-free or partially ice-covered waters (Table 5.1).

5.4.3 Morphometric measurements

At sea, all fish larvae and juveniles including the numerically dominant gadids were sorted from the zooplankton samples. For each sample, fresh standard length (SL) and body depth at the anus (H) of all or up to a maximum of 50 randomly selected gadids from different length classes spanning the entire length range, were measured fresh to the nearest 0.1 mm before individual preservation of the fish in 95% ethanol in a separate vial. Non-measured fish were preserved in 95% ethanol as well. Back in the laboratory, preserved standard length (SLP) and body depth at the anus (HP) were measured on the fish measured fresh at sea, as well as on all intact preserved fish > ca. 20 mm and on a maximum of 50 fish < ca. 20 mm per sample, for a total of 3697 gadids in 2004 and 1284 in 2008. Preserved standard length and body height were corrected for shrinkage using the relationships SL = 1.073 SLP 2 2 + 0.253, (r = 0.995, n = 663) and H = 1.048 HP + 0.085, (r = 0.966, n = 321).

111

112

Table 5.1 Number of gadid collected by different sampling methods in 2004 and 2008, with average standard length L (± 1 standard deviation, SD) and percentage of Arctogadus glacialis.

Maximum Sampling SL ± SD % A. Year Deployment Conditions sampled n cast n gadid gear (mm) glacialis depth (m) 2004 DSN Horizontal Under ice 1.5 8 48 8.8 ± 2.3 54.2 DSN Oblique Under ice 40 24 3538 6.6 ± 0.9 2.9

DSN Oblique Open water 25 - 400 59 2264 12.5 ± 4.1 8.9

RMT Oblique Open water 25 - 530 9 111 16.8 ± 3.7 14.4

EZ-Net Oblique Open water 20 - 500 25 457 13.7 ± 3.2 6.8

Other Vertical Open water 85 - 375 12 61 6.9 ± 2.3 1.6

2008 DSN Oblique Open water 25 - 163 37 3223 8.2 ± 4.2 4.4 RMT Oblique Open water 60 - 92 12 369 13.8 ± 3.7 16.8

Other Vertical Under ice or 10 - 1080 58 516 7.2 ± 1.8 1.0 open water

113

Table 5.2 Total number of gadids collected, number and percentage of gadids identified to species by genetics and/or otolithometry, by sampling month in 2004 and 2008

Year Month n total n identified % identified 2004 Apr 36 30 83 May 3600 119 3

June 835 83 10

July 1839 311 17

Aug 169 108 64

2008 Apr 5 2 40 May 2346 41 2

June 1017 91 9

July 708 165 23

Aug 32 30 94

114

115

116

Gut contents taxa Equation taxa Equation Reference

-8 2.906 Hansen et al. 1999, Acartia spp. Acartia longiremis C = 1.023 * 10 * (PL * 1000) Fig. 9 Calanus glacialis Calanus glacialis C = 4.742 * PL3.452 Forest et al. 2011, Fig. 4b (summer) Calanus hyperboreus Calanus hyperboreus C = 7.263 * PL3.106 Forest et al. 2011, Fig. 4a (summer)

Calanus spp. nauplii Calanus finmarchicus C = 4.29 * 10-6 * (L * 1000)2.05 Hygum et al. 2000, Fig. nauplii 4 (high food resources)

Cyclopoid nauplii Oithona similis nauplii C = 5.545 * 10-8 * (PL*1000)2.71 Sabatini & Kiørboe 1994, Fig. 1 Eurytemora spp. Eurytemora herdmani C = (10 2.96*log (PL*1000) - 7.604) * 0.447 Middlebrook & Roff 1986, Equation 8, (1) Harpaticoida Macrosetella gracilis C = e 1.03 * ln (L*1000) -7.07 Satapoomin 1999, and Microsetella spp. Table 2

Limnocalanus spp. Copepoda C = 10 3.07 * log (PL * 1000) - 8.37 Uye 1982, Table 1

Metridia longa Metridia longa C = 7.498 *PL3.225 Forest et al. 2011, Fig. 4c (summer) Metridia spp. nauplii Acartia tonsa nauplii C = 3.18 * 10-9 * (PL * 1000)3.31 Berggreen et al. 1988, Fig. 3

Microcalanus spp. Copepoda C = 10 3.07 * log (PL * 1000) - 8.37 Uye 1982, Table 1

Microcalanus spp. Pseudocalanus C = (10 2.515 * log L + 0.975) * 0.447 Lee et al. 2003, Fig. 3, nauplii newmani nauplii (1)

Oithona similis Oithona similis C = 9.4676 * 10-7 * (PL*1000)2.16 Sabatini & Kiørboe 1994, Fig. 1 Oncaea parila Oithona similis C = 9.4676 * 10-7 * (PL*1000)2.16 Sabatini & Kiørboe 1994, Fig. 1 Paraeuchaeta glacialis Paraeuchaeta spp. C = (0.0075 PL 3.274 + (0.07 * Mumm 1991, Appendix 0.0075 * PL 3.274)) * 0.447 * 1000 C (1)

Pseudocalanus spp. Pseudocalanus spp. C = (10 2.85 * log (PL*1000) - 7.62) * 0.447 Liu & Hopcroft 2008, Fig. 1, (1)

Pseudocalanus spp. Pseudocalanus C = (10 2.515 * log L + 0.975) * 0.447 Lee et al. 2003, Fig. 3, nauplii newmani nauplii (1)

Triconia borealis Oithona similis C = 9.4676 * 10-7 * (PL*1000)2.16 Sabatini & Kiørboe

117

1994, Fig. 1 Unidentified copepod Copepoda C = 10 3.07 * log (PL * 1000) - 8.37 Uye 1982, Table 1

Unidentified copepod Pseudocalanus C = (10 2.515 * log L + 0.975) * 0.447 Lee et al. 2003, Fig. 3, nauplii newmani nauplii (1)

Unidentified cyclopoid Oithona similis C = 9.4676 * 10-7 * (PL*1000)2.16 Sabatini & Kiørboe 1994, Fig. 1

Amphipoda Scina crassicornis, C = 10 ((log L - 0.063)/0.277) * 0.285 Gorsky et al. 1988, Phrosina semilunata, Tables 1 & 2 Phronima sedentaria

-8 2.70 Appendicularia Oikopleura rufescens C = 8.20 * 10 * (LTR*1000) Sato et al. 2003, (synonym: (where LTR = trunk length) Equation 7 Oikopleura vanhoeffeni)

Bivalvia Mytilus edulis C = 3.06 * 10-8 * (L * 1000) 2.88 Fotel et al. 1999, Section 3.4 Chaetognatha Sagitta crassa C = 10 3.16 * log L -1.29 Uye 1982, Table 1

Cirripeda Mytilus edulis C = 3.06 * 10-8 * (L * 1000) 2.88 Fotel et al. 1999, Section 3.4 Cnidaria C = 0

Eggs 7 copepod species C = 0.14 * 10-6 * (4/3π (L * Kiørboe 1985 (Calanus 1000/2)3) finmarchicus, C. hyperboreus, Temora longicornis, Acartia longiremis, A. tonsa, Centropages hamatus and Pseudocalanus spp.)

Gastropoda Spiratella retroversa C = 10 3.102 logD + 1.469 (where D = Conover & Lalli 1974, diameter) Fig. 1 Ostracoda Conchoecia C = 0.346 * e3.868*L * 0,4 Ikeda 1990, Table 2 (2) pseudodiscophora Polychaeta Neanthes succinea 1.42 * 10-4 * (L*1000)1.47 Hansen 1999, Table 1

Digested material Digested material C = (109.08 * (π * (W / 2)2 * Sirois 1999, (2) L)0.9591) * 0.4 (where W = width) (1) Copepoda C = 44.7% DW Mauchline 1998 (2) Zooplankton C = 40% DW Legendre & Michaud 1998

118

5.4.4 Species determination

The early stages of Boreogadus saida and Arctogadus glacialis can be distinguished by genotyping the Gmo8 microsatellite locus (Madsen et al. 2009) or by measuring the nucleus of the lapillar otolith (Bouchard et al. 2013). A stratified sub-set of 1009 gadids (668 in 2004 and 341 in 2008) from all months and 83 of the 95 sampling stations were selected for species identification by Gmo8 and/or lapillus analysis (see Bouchard et al. 2013 for details). In summary, 802 gadids (494 in 2004 and 308 in 2008) were identified by genotyping the microsatellite locus Gmo8 (Madsen et al. 2009). Bouchard et al. (2013) estimate at 3.4% the overall misclassification error of larval gadids from the Beaufort Sea based on Gmo8. In addition, the lapillar otolith of 660 gadids (372 in 2004 and 288 in 2008, including 453 of the fish genotyped for Gmo8) was measured, and individuals with a product of the shortest and longest diameters of the nucleus SN×LN ≥ 653 µm2 were assigned to A. glacialis (Bouchard et al 2013). The analysis of 453 gadids by both methods (198 in 2004 and 255 in 2008) yielded the same identification in 94% of cases. The 29 ambiguous fish for which identification differed between methods were assigned to species by redistribution (see redistribution section). While a relatively large fraction of the few fish sampled in April (83% in 2004, 40% in 2008) were identified to species by either method, only a small proportion (2 to 23%) of the large number of fish sampled in May, June and July were analysed (Table 5.2).

5.4.5 Age determination

All intact gadid > 24 mm in length were aged by otolith analysis. For larvae < 24 mm, a sub-set of larvae including individuals in each 1 mm interval and representing all stations and capture months were selected for otolith age determination. Lapillar otoliths of each selected fish were extracted, mounted on microscope slides with thermoplastic polymer, and polished from one or both sides on a 0.5-μm metallurgical lapping film. Increments of the left lapillus were enumerated and measured under a microscope coupled to a camera and image analyzer system (Image Pro Plus®). A first reading was made on 380 lapilli from 2004 and 327 lapilli from 2008 (both species included). Ageing precision (Campana 2001) was estimated by comparing the first reading with a second independent reading

119 made on 73 otoliths from 2004 and 48 otoliths from 2008, resulting in a mean coefficient of variation of 4.3% for 2004 and 4.0% for 2008. The first counts were retained in subsequent analyses.

5.4.6 Redistribution of length, species and age

Measuring, identifying and aging each of the several thousand gadid larvae collected was beyond our analytical capacities. The sub-sets of larvae selected for morphometric measurements, aging, and species identification were designed to span the range of length in a given sample. Hence, only a minute fraction of the small larvae present in large numbers in some samples collected from May to July were analyzed. To obtain unbiased size frequency distributions and estimates of species occurrence, length, age and species were redistributed to the entire gadid population sampled based on the following procedures.

For samples containing more than 50 larvae < 20 mm, each non-measured gadid was assigned randomly a fresh standard length and body height based on the frequency distribution of SL and H for the 50 fish < 20 mm actually measured in that sample. Overall, the length and body height of 2782 gadids < 20 mm in 2004 (42.9% of the total number of fish of all sizes) and 2824 in 2008 (68.7%) were extrapolated this way.

The age of the larvae actually measured but not aged by otolith analysis (n = 3317 in 2004 and 957 in 2008) was estimated from their size. For both species, the age of identified and measured fish was strongly correlated to standard length, but the residuals of the linear regression were unevenly distributed in the 5-15 mm length interval (Fig. 5.2). The residuals of the regression of age on the cubic root of the product of standard length by body height (LH1/3) were more stationary (Fig. 5.2). Therefore, age was estimated from LH1/3, with fish in a given 0.5 mm2 LH1/3 class being attributed an age according to the known age probability function for that LH1/3 class in each sampling year. The small (< 20 mm) larvae not measured (n = 2782 in 2004 and 2824 in 2008) were attributed an age randomly, pro rata to the frequency distribution of the ages of measured larvae in the

120

sample they came from. Note that the vast majority of the unmeasured larvae < ca. 20 mm were actually small newly hatched fish.

Among the 980 gadids identified to species by molecular genetics or otolith analyses, the Arctogadus glacialis sampled on a given date were generally larger than the Boreogadus saida sampled on the same date. The size divergence between the two species was clearer in small fish and more marked in LH1/3 than in SL (Fig. 5.3). For a given sampling month, unidentified gadids in a given 0.5 mm2 LH1/3 class were assigned to species based on the percent occurrence of A. glacialis in this LH1/3 class. The overlap in LH1/3 between the two species, and hence the risk of misidentification, was relatively small in spring, in particular in May and June when large numbers of small larvae were sampled (Fig. 5.3, Table 5.2). The size overlap and hence the ambiguity in species attribution was maximum for the large fish sampled in August, 69% of which however were identified by genetics or lapillar nucleus size (Fig. 5.3, Table 5.2).

5.4.7 Hatch date frequency distributions

The hatch date of an individual fish was determined by subtracting its estimated age (in days) from its date of capture. The uncorrected hatch date frequency distributions (HFDs) of the young fish captured in a given month was built by tallying the number of fish hatched in the same 3-day bin. In uncorrected HFDs, a hatch date bin comprising old larvae will be under-represented relative to a bin comprising young larvae that have experienced mortality and dispersion for a shorter period (Yoklavich & Bailey 1990, Campana & Jones 1992, Fortier & Quiñonez-Velazquez 1998). Mortality-dispersion rates were estimated from the slope of the descending limb of the catch-at-age curve. Based on these rates, the HFDs were corrected by calculating for each hatch date bin the initial number of fish needed to account for the observed number given losses by mortality-dispersion during the interval between hatching and the average age in that bin.

121

5.4.8 Gut content analysis

The gut content of length-stratified subsets of 209 Boreogadus saida and 68 Arctogadus glacialis was examined under a stereomicroscope at magnifications from 50 to 80×. Each prey was identified to the lowest taxonomic level possible and measured (prosome length for copepods and total length for other taxa). The carbon content of each prey was estimated with published length–weight and carbon–weight relationships (Table 5.3). Specific relationships were used when available. Otherwise carbon content was estimated with generic or more general equations for closely related taxon. The overlap in the diet of the two species was measured with Schoener’s index (Schoener 1970).

5.5 Results

5.5.1 Compared occurrence, length, and growth of Arctogadus glacialis and Boreogadus saida

The 6479 larvae and juvenile gadids sampled from April to August 2004 ranged from 3.3 to 30.8 mm in length and from 0 to 138 days of age. The 4108 young gadids collected from April to August 2008 ranged from 4.1 to 31.2 mm in length and from 2 to 151 days of age. Among the 980 gadids identified to species by genetics and/or otolith nucleus diameter, 165 (16.8%) belonged to Arctogadus glacialis. After redistribution, A. glacialis represented 5.8% of the overall gadids sampled in 2004 and 5.1% in 2008.

Mean length was significantly larger in A. glacialis than in B. saida for all months in 2004 and for May, June and July in 2008 (t-tests, p < 0.0001, Table 5.4). Grouping 2004 and 2008, monthly frequency distributions of standard length (SL) and the cubic root of the product of length by body depth at the anus (LH1/3) showed little overlap between the two species for April, May and June, intermediate overlap for July and maximum overlap for August (Fig. 5.3).

Somatic growth estimated from the slope of the length-age regression for individuals identified to species was higher in 2004 than in 2008 (Table 5.5), significantly so for B.

122

saida (ANCOVA, p = 0.010) but not for A. glacialis (p = 0.143). In a given year, somatic growth did not differ significantly between the two species (ANCOVA, p > 0.630). In both years, intercepts pointed to a significantly smaller length at hatch in B. saida than in A. glacialis (ANCOVA, p < 0.0001).

5.5.2 Vertical distribution underice and in ice-free waters

Of the 3538 gadid larvae collected by oblique tows (40-0 m) under the ice of Franklin Bay in April and May 2004, only 2.9% were Arctogadus glacialis (Table 5.1). By contrast, 26 of the 48 gadids (54%) captured in horizontal tows immediately under the ice (0.5-1.5 m) were A. glacialis ranging from 7.9 to 14.4 mm in length and from 10 to 23 days in age.

When averaged over 14 profiles obtained at different dates from early June to early August, the vertical distributions of the two species was similar with 59% of Boreogadus saida and 43% of Arctogadus glacialis occurring within 10 m from the surface (Fig. 5.4). The remainder of the population distributed primarily between 10 and 40 m, except for a few B. saida sampled between 40 and 160 m. In both species, the fraction of the population close to the surface exhibited the same limited circadian migration, concentrating in the 5-10 m layer in daytime (06h00 to 18h00 local time) and moving upward in the 0-5 m layer at night (Fig. 5.4). Except for a few B. saida sampled between 40 and 160 m, the young gadids were distributed above 40 m and this pattern varied little over the summer (Fig. 5.5). Collection depth averaged 13 m over the season and did not differ significantly between the two species (t-test, p = 0.232, n = 438).

5.5.3 Monthly distribution of Arctogadus glacialis and Boreogadus saida in partially ice-covered southeast Beaufort Sea

After species redistribution, the monthly spatial distribution of the two species differed little within the partially ice-covered or ice-free areas sampled from June to August 2004 and May to August 2008 (Fig. 5.6). Among the 96 stations sampled with DSN oblique tows, only one (1%) yielded no young gadid while both species were collected at 73 stations (76%). In a given month, B. saida was usually one order of magnitude more

123 abundant than A. glacialis. In May 2008, both species were recorded at stations sampled in the flaw lead west and north of Banks Island. In that month, B. saida was relatively abundant in the Amundsen Gulf with a large number of small larvae collected near the fast ice edge off Cape Parry (Fig. 5.6). In June 2004 and 2008, high densities of B. saida (up to 434 ind. 1000 m-3) and A. glacialis (up to 42 ind. 1000 m-3) were found in Franklin Bay, Darnley Bay and the Amundsen Gulf (Fig. 5.6). B. saida tended to be less abundant at the few stations sampled on and at the edge of the Mackenzie Shelf than in Amundsen Gulf. High densities were also recorded in July 2004 and 2008, with maximum values of 1042 and 58 ind. m-3 for B. saida and A. glacialis, respectively. Relative to June, the highest densities of both species had shifted from Amundsen Gulf to the western Mackenzie Shelf and slope in July (Fig. 5.6). Sampling was limited to the Amundsen Gulf in August, where relatively low densities of juveniles (98 and 6 ind. 1000 m-3 for B. saida and A. glacialis respectively) were recorded.

5.5.4 Hatchdate frequency distributions of Boreogadus saida and Arctogadus glacialis

For both species, the uncorrected hatchdate frequency distribution (HFD) of the young fish identified to species and aged, spanned from March to early July (Fig. 5.7). In both years, the HFD of Boreogadus saida was shifted to the left when recalculated for all fish after the redistribution of ages. The centre of mass of the HFD of Arctogadus glacialis was little affected by the redistribution process, but its seasonal extent increased as fish hatched in March 2004 and February 2008 and June-July of both years were included (Fig. 5.7).

The average age at capture varied considerably among the hatchdate bins, with early hatchers being older on average at capture (and therefore having been decimated by mortality for longer) than late hatchers. This resulted in the under-representation of early hatchers and the over-representation of late hatchers in the uncorrected HFDs (Fig. 5.7). To correct for this bias, the rate of mortality of each species was assessed from catch-at-age curves (Fig. 5.8). The estimated mortality rate differed significantly between 2004 (4.3% d-1) and 2008 (3.4% d-1) for Boreogadus saida (ANCOVA, p < 0.049), but not for Arctogadus glacialis (1.8 and 2.3 % d-1). For each year, these rates were used to retro-

124

calculate the initial number of hatchlings needed to account for the observed numbers in each hatchdate bin once losses by mortality are offset.

Relative to the uncorrected HFDs, the resulting corrected HFDs were shifted to the left as increased numerical weight was given to older larvae that had suffered mortality for longer than young larvae (Fig. 5.9). In 2004, the hatching of Boreogadus saida started in early March, was most intense from mid-April to late May, and was nearly completed by the end of June with little hatching in July. In 2008, hatching started in the second half of February, increased progressively until a peak in late April and then declined until the end of June. In 2004, the protracted hatching of Arctogadus glacialis started in early March, peaked from mid-April to early May and then declined progressively until mid July (Fig. 5.9). In 2008, sporadic hatching was observed as early as mid February, peaked in mid April and declined until the end of June. In a given year, the hatching season of the two species started and ended simultaneously, but was hastened by ca. 13 d for B. saida and 20 d for A. glacialis in 2008 relative to 2004 (Fig. 5.9). The ice breakup was about 20 d earlier and sea-ice declined more rapidly in 2008 than in 2004. Average ice concentration in the study area reached <50% in mid-June 2008 and at the end of June 2004. By the end of July, only 3% of the area was covered by ice in 2008, compared to 31% in 2004 (Fig. 5.9).

5.5.5 Prey and carbon intake

A majority of the B. saida larvae collected under the ice in April and May presented empty guts (18 out of 21 examined). The three feeding larvae contained Pseudocalanus copepodites, Pseudocalanus nauplii, and invertebrate eggs. By contrast, only two of the 16 A. glacialis larvae collected under the ice and selected for gut content analysis contained no prey. Under the ice, A. glacialis larvae preyed primarily on Pseudocalanus nauplii (38% of carbon intake), cyclopoid nauplii (28%), and Calanus nauplii (20%).

Pooling data for all sampling months, the diet of the two species by length classes differed initially and then converged as the larval stages metamorphosed into juveniles. Schoener’s index of diet overlap was relatively low for the <15 mm length class (SI = 0.38) and

125

Table 5.4 Number and standard length SL (mean ± 1 standard deviation, SD) of Boreogadus saida and Arctogadus glacialis collected by sampling months in 2004 and 2008.

Year Month Boreogadus saida Arctogadus glacialis n SL ± SD (mm) n SL ± SD (mm) 2004 Apr 10 6.3 ± 1.0 26 10.5 ± 1.6 May 3496 6.5 ± 0.8 104 8.1 ± 1.9 Jun 745 8.1 ± 1.1 90 11.7 ± 2.5 Jul 1704 14.0 ± 2.6 135 17.0 ± 3.7 Aug 145 19.2 ± 3.4 24 23.3 ± 3.9

2008 Apr 2 7.2 ± 3.8 3 9.8 ± 0.5 May 2278 5.8 ± 0.9 68 11.0 ± 1.4 Jun 942 8.9 ± 1.5 75 12.9± 2.0 Jul 646 15.3 ± 3.7 62 18.0 ± 4.2 Aug 30 23.6 ± 2.5 2 24.5 ± 3.6

126

Number of Size at Growth rate Species Year otoliths -1 hatch (mm) (mm j ) analysed Boreogadus saida 2004 271 5.53 0.173 2008 267 4.90 0.161 Arctogadus glacialis 2004 78 6.86 0.176 2008 30 6.25 0.163

127

128

Figure 5.5 Vertical distribution by date of the early life stages of Boreogadus saida and Arctogadus glacialis collected in the summer 2004 in southeast Beaufort Sea.

129

Figure 5.6 Monthly abundance (no. 1000 m-3) of Boreogadus saida and Arctogadus glacialis larvae and juveniles collected by oblique DSN tows in summer 2004 (red) and summer 2008 (blue).

130

131

132

133

134

increased to 0.70 in fish >15 mm (Fig. 5.10). In larvae <15 mm, the most frequent prey of Boreogadus saida were Pseudocalanus spp. nauplii (42%) and Calanus spp. nauplii (16%), while Arctogadus glacialis preyed primarily (65%) on cyclopoid nauplii (Fig. 5.10a). From 15 to 25 mm, the diet of the two species became more similar, with continued preference of B. saida for Pseudocalanus spp. nauplii and a shift of A. glacialis from nauplii to copepodites of the cyclopoids Oithona similis and Triconia borealis (Fig. 5.10b). Beyond 25 mm in length, the diet of both species shifted from nauplii to larger copepodite prey, including Calanus glacialis, Pseudocalanus spp., O. similis, and to a lesser extent T. borealis (Fig. 5.10c).

In larvae <15 mm, Calanus nauplii (40%) and invertebrate eggs (22%) contributed the most to the carbon intake of B. saida (Fig. 5.10d). Cyclopoid nauplii (43%) and Calanus nauplii (20%) were the main sources of carbon for A. glacialis. In the 15-25 mm length class, Calanus nauplii represented 29% and 25% of the energy intake of B. saida and A. glacialis, respectively. The infrequent but large Calanus glacialis copepodite prey accounted for 24% of the carbon intake of B. saida 15-25 mm (Fig. 5.10e). Oithona similis (16%) and gastropod larvae (17%) were important carbon sources for A. glacialis 15-25 mm. At lengths >25 mm, the large copepodites C. glacialis and C. hyperboreus contributed 90% of the carbon intake of B. saida, while C. glacialis accounted for 63% of the carbon intake of A. glacialis (Fig. 5.10f).

5.6 Discussion

5.6.1 Discriminating young Boreogadus saida and Arctogadus glacialis in large plankton collections

The morphological similitude of Boreogadus saida and Arctogadus glacialis during early life in the plankton (Fahay 2007) has hindered the reliable discrimination of the two species in field collections. Recently, molecular (Madsen et al. 2009, Nelson et al. 2013) and otolithometric (Bouchard et al. 2013) identification methods have been developed. These methods are labour intensive and the analysis of large numbers of fish often remains prohibitive (Bouchard et al. 2013). In the present study, 9.3% of the young gadids collected

135 were measured, aged, and identified to species by genetics and/or otolithometry. For these confirmed fish, the larger size at hatch of A. glacialis, its slightly earlier hatching in a given year, and the equal growth rates of the two species resulted in A. glacialis being consistently larger at date than B. saida, at least from April to June (Fig. 5.3). This size difference allowed the attribution of the abundant newly hatched and small larvae collected in these months to either species with some confidence. By July, the size divergence between the two species had attenuated and, while our probabilistic extrapolation of the overall number of A. glacialis sampled in that month is probably correct, confidence in the identification of a given gadid of a given size from a given station is low. This uncertainty in the statistical redistribution of species based on size prevails for August as well when the overlap in the size distribution of the two species was greatest. However, a larger fraction (69%) of the relatively few fish sampled in that month were identified by genetics or their otolith nucleus. The a posteriori analysis of the otolith of a subset of the fish identified and aged on the basis of their size would provide an evaluation of the actual magnitude of the error in the species and age redistribution procedures.

5.6.2 The vertical distribution and feeding of Boreogadus saida and Arctogadus glacialis under the ice and in ice free waters

Fish larvae are visual predators that feed primarily on copepod eggs and nauplii. Under the landfast ice of coastal Hudson Bay in spring, the vertical distribution of first-feeding Boreogadus saida is dictated by the thickness of river plumes, the larvae aggregating just below the strong halocline separating fresh and salt water where the product of light by prey availability is maximum (Ponton & Fortier 1992, Fortier et al. 1996). Under the landfast ice of Franklin Bay in April and May, thousands of newly hatched B. saida larvae were captured in oblique tows from 40 m to the surface, but only 22 in horizontal tows immediately under the ice-water interface (0.5-1.5 m). Hence B. saida likely distributed over the shallow euphotic layer but avoided the ice-water interface. By contrast, post first- feeding Arctogadus glacialis (7.9 to 14.4 mm and 10 to 23 d old) clearly occupied the 0.5- 1.5 m layer near the ice-water interface. Although the larvae analyzed for gut content were of similar size (5.3 - 12.3 mm versus 6.6 - 14.4 mm, respectively), feeding incidence in B.

136

saida sampled under the ice was low (14%) relative to A. glacialis (88%). This suggests that the capacity to move close to the ice-water interface where light is more intense provided A. glacialis with some foraging advantage.

In spring and early summer, several copepod species dominated by Calanus glacialis, Pseudocalanus spp., Oithona similis and C. hyperboreus associate with a 0.3 to 0.5-m thick meltwater layer under the landfast ice of the Beaufort Sea (Hop et al. 2011). In C. glacialis, Pseudocalanus and likely O. similis, the early production of eggs and nauplii is fuelled by the production of ice algae before the phytoplankton bloom (e.g. Runge & Ingram 1991, Tourangeau & Runge 1991, Darnis 2013). The naupliar stages of these copepods were the primary carbon sources of Boreogadus saida and Arctogadus glacialis larvae under the ice. Thus, at least in April and May, the first-feeding success and early growth of both species could depend on the early reproduction of copepods grazing ice algae. The food of larvae hatched under the ice in February and March before the onset of microalgal production (Fig. 5.9) remains unknown, but the naupliar stages of small omnivorous copepods that reproduce year long and the floating eggs of Calanus hyperboreus laid in winter (Darnis 2013) are potential candidates.

In partially or fully ice-free waters in summer, the late larval and early juvenile stages of the two species shared essentially the same vertical distribution and circadian migration pattern (Fig. 5.4). Despite the midnight sun prevailing from June to early August, both species exhibited a small-amplitude diel vertical migration (DVM) with part of the population invading the near surface (0-5 m) layer during night hours and retreating to the 5-10 m layer in daytime. Such a DVM pattern is typical of situations in which greater food availability near the surface is offset by higher vulnerability to visual predators in daytime (Lampert 1989, Fortier et al. 2001). One interpretation of the observed vertical distributions is that the non-feeding fraction of the population remained below 10 m whereas the actively foraging fraction occupied the 0-10 m layer (Fig. 5.4). Assuming increased availability of prey near the surface, foraging larvae and juveniles minimized predation by remaining in the 5-10 m layer in daytime and invading the 0-5 m layer at night when the sun was low

137 enough on the horizon to hamper seabirds and other visual predators, while still providing enough light for the young fish to capture their prey.

The diet of the two species differed primarily during the early larval stage (<15 mm) when Arctogadus glacialis specialized on cyclopoid copepod nauplii and Boreogadus saida on calanoid copepod nauplii. This divergence in diet occurred under the spring ice cover. We suspect that it reflected the capacity unique to A. glacialis to occupy the near-surface diluted layer immediately under the ice-water interface where light would be sufficient to detect the small cyclopoid nauplii. As the larvae of the two species developed into pre- metamorphosis larvae (15-25 mm) in ice-free surface waters in summer, both their vertical distribution and their diet converged, to the exception of a tendency for A. glacialis to prey on cyclopoid copepodites. By metamorphosis, B. saida obtained most of its energy from Calanus hyperboreus and C. glacialis, and A. glacialis from C. glacialis. At that time, the two very similar species occupied essentially the same niche in the plankton of the Beaufort Sea.

5.6.3 Spatio-temporal sympatry of the planktonic stages of Boreogadus saida and Arctogadus glacialis in southeast Beaufort Sea

Spatially, the early stages of the two species shared the same geographical distribution, occupying most of the region sampled in a given month. This is consistent with the ubiquity of juvenile gadids in the surface layer (0-100 m) of southeastern Beaufort Sea as detected by acoustic surveys (Geoffroy and Fortier, in preparation). Adult A. glacialis are distributed in shallow coastal and estuarine habitats and on continental shelves (Nielsen & Jensen 1967, Aschan et al. 2009) while B. saida, at least in the Beaufort sea, often congregate in deep embayments and on the slope of the continental shelf (Benoit et al. 2008, Geoffroy et al. 2011). We found no evidence for such a general inshore-offshore segregation of the two species during the larval and juvenile stages.

The planktonic stages of Boreogadus saida and Arctogadus glacialis not only co-occurred spatially but also coincided in time, the two species presenting a similar protracted hatching

138

season from February to July. In subarctic seas, hatching is sometime synchronized with the vernal onset of primary production and the maximum seasonal production of the zooplankton prey of the larvae (e.g. the match/mismatch hypothesis, Cushing 1990). In Arctic seas influenced by large rivers, the protracted hatching season of Boreogadus saida starts in winter when the abundance of suitable zooplankton prey is at a minimum and ends in early summer when prey availability peaks (Bouchard & Fortier 2011). Obviously, some constraint other than matching the emergence of the larvae with peak availability of their prey dictates the hatching season of B. saida. A large size at the end of the short arctic summer should reduce the vulnerability of juvenile B. saida to avian predation, cannibalism, and winter starvation (Fortier et al. 2006). Hence, the paradox of winter- spring hatching when food is relatively scarce could be explained by the need to maximize the duration of the first growing season, so as to achieve a minimum threshold size in August (Bouchard & Fortier 2008, Bouchard & Fortier 2011). In the Beaufort Sea, the hatching season of Arctogadus glacialis was similar to that of B. saida, also extending from February to early July (Fig. 5.9). This suggests that the two species are equally constrained by this imperative of hatching early to achieve a threshold pre-winter size. However, arctic continental shelves are varying and unpredictable environments sometimes causing early hatched Boreogadus larval cohorts to vanish entirely, leaving the larvae hatched later in the season as the only survivors of the year for a given region (Fortier et al. 2001, Bouchard and Fortier 2008). In evolutionary terms, late hatching may has been maintained as a way to avoid a complete recruitment failure in years when environmental conditions preclude any survival of early hatchers. In other worlds, the long hatching season of Boreogadus and Arctogadus may have evolved, in response to the highly unpredictable arctic environment, as a diversified bet-hedging strategy (don’t put all eggs in one basket) in order to minimize the variation in fitness between years (Olofsson et al. 2009).

Interestingly, the hatching season of the two species varied in parallel interannually, starting and ending ca. 2 weeks earlier in 2008 than in 2004. In the Laptev Sea (Siberian Arctic Ocean), interannual variations in the hatching season and survival of B. saida have been linked to the frequency and extent of leads and polynyas during winter (Bouchard & Fortier 2008). The winter ice cover was particularly low in southeast Beaufort Sea in 2008

139 with numerous and extensive leads, and an early opening of the Cape Bathurst polynya (e.g. Barber et al. 2010), all of which reflected in an early ice breakup relative to 2004, a year of normal ice regime (Fig. 5.9). Hence, improved feeding conditions in winter resulting from more light reaching the surface layer through more frequent leads could explain the earlier hatching dates of the survivors of both species in 2008 compared to 2004 (Bouchard & Fortier 2008).

5.6.4 Ecological divergences and niche separation during life in the plankton

While adult Boreogadus saida and Arctogadus glacialis are morphologically and ecologically quite different, many aspects of their planktonic life proved remarkably similar. In addition to being almost indistinguishable morphologically, the young stages of the two species overlapped geographically, presented the same hatching season that varied in parallel between sampling years, grew at the same average rate, and, as pre-juveniles and juveniles were distributed in the same depth interval and shared essentially the same diet in open waters in summer.

Divergences in some early life history traits were also documented that may reduce competition between the two species and help explain the sympatric existence of the nearly identical larvae and juveniles in the plankton. First, Arctogadus glacialis was larger at hatch (6.2 to 6.9 mm) than Boreogadus saida (4.9 to 5.5 mm, see also Michaud et al. 1996, Ponomarenko 2000). This difference in initial size combined to a slightly earlier hatching season resulted in the modal length of A. glacialis being on average ca. 5 mm longer at a given date than that of B. saida from April to July (Fig. 5.3). Second, the estimated mortality rate of A. glacialis over the larval and juvenile stages was about half that of B. saida (Fig. 5.8). A lower mortality rate likely contributed to accrue the size difference at date between the two species, a smaller fraction of the population of B. saida surviving to a larger size. Third, under the ice in spring the first feeding larvae of the two species were partially segregated over depth, A. glacialis occupying the thin layer immediately under the ice-water interface while B. saida avoided it. This depth segregation coincided with a remarkably higher feeding occurrence (88%) in A. glacialis than in B. saida (14%) of

140

similar size ranges. It could also explain in part the initial specialization of A. glacialis on small cyclopoid nauplii. Finally, the scarcity of adult A. glacialis relative to B. saida was reflected in the relative abundance of their respective early life stages, with B. saida being approximately 20 times more frequent in our collections than A. glacialis.

Several hypotheses on the determinism of larval fish survival emphasize the advantage provided by fast growth and a large size (see Robert et al. in press for a recent review). Most other traits being equal in the morphologically similar early stages of the two species, a larger size at age and at date likely provided Arctogadus glacialis with some survival advantage over Boreogadus saida from hatching to metamorphosis. The low mortality rate of A. glacialis compared to B. saida is consistent with this interpretation. A more stable and predictable food source in the form of small cyclopoid nauplii and copepodites near the ice- water interface in spring could be the key ingredient allowing A. glacialis to persist as a species by producing fewer and larger larvae than B. saida. The validation of this hypothesis awaits the high-resolution sampling of the vertical co-distribution of larval gadids and their microzooplankton prey under the ice in winter-spring (using for example the methodology of Ponton and Fortier 1992), and the analysis of their isotopic composition to assess the relative importance of ice algae and phytoplankton in the feeding of their main copepod prey (e.g. Hobson et al. 2002, Pineault et al. 2013).

5.6.5 Conclusion

The presence of Arctogadus glacialis larvae in the thin layer immediately under the ice- water interface confirms its status of arctic specialist endemic to the High Arctic. As the ice regime of the Arctic Ocean and its ancillary seas shifts in response to global warming over the present century, the ice breakup is expected to occur progressively earlier in spring, potentially disrupting the timing and intensity of ice algae and phytoplankton production (Tremblay et al. 2011, Ji et al. 2013). In an extreme scenario, the sea-ice substrate could melt before the vernal increase in solar irradiance in April, stifling ice algal production and potentially hindering the reproduction of copepods that depend on this production. Our results suggest that, relative to Boreogadus saida, A. glacialis would be particularly

141 vulnerable to such a decline in ice algal production. Hence the importance of confirming the proposed dependence of first-feeding A. glacialis larvae on the zooplankton assemblage associated with the meltwater layer under the arctic ice cover in spring.

142

Chapitre 6 – Conclusion générale

6.1 Hypothèse du refuge thermique hivernal

De mes précédents travaux de maîtrise (Bouchard 2007) et de la présente thèse doctorale a émergé l’hypothèse du refuge thermique hivernal associé à l’eau douce. Selon cette hypothèse, les températures relativement élevées retrouvées près de l’embouchure des grands fleuves permettent à une certaine proportion des larves de Boreogadus saida d’éclore en hiver. Sous le couvert de glace hivernale, la température de l’eau atteint son point de congélation, donc environ –1.8°C en pleine mer où la salinité est de 33, une température à laquelle la survie des larves est compromise par la réduction de leur motilité et de leur succès d’alimentation (e.g. Michaud et al. 1996). Par contre, la salinité réduite du panache d’un fleuve en augmente le point de congélation à des températures favorables à la croissance et à la survie des larves en hiver (au-dessus de -1°C pour S < 20). En accord avec cette hypothèse, la saison d’éclosion dans les régions à proximité d’un apport fluvial important commence dès janvier alors que dans les régions éloignées des sources d’eau douce, l’éclosion débute au printemps seulement (chapitre 2). Les différences de rapports Li/Ca, Mg/Ca, Mn/Ca, Sr/Ca et Ba/Ca dans les otolithes entre les régions influencées par l’eau douce et celles qui ne le sont pas soutiennent aussi l’hypothèse du refuge thermique hivernal (chapitre 3). Aussi, la saison d’éclosion observée chez Arctogadus glacialis suggère qu'une éclosion hâtive soit une stratégie partagée par au moins deux espèces de poissons arctiques (chapitre 5). Dans l’ensemble, mes résultats témoignent de l’importance de l’eau douce pour la survie et la croissance larvaire de plusieurs espèces de poissons marins qui profiteraient des nombreux avantages associés aux estuaires et aux panaches des fleuves, notamment une osmorégulation moins énergivore, une température plus favorable, un risque de prédation plus faible et une nourriture plus abondante. Ils appuient aussi l’hypothèse selon laquelle la mortalité hivernale étant inversement proportionnelle à la taille (Sogard 1997), plusieurs espèces des mers tempérées et polaires tendent à maximiser la taille pré-hivernale des juvéniles en optimisant les dates d’éclosion et la croissance (e.g. Conover 1992, Fortier et al. 2006, Bouchard & Fortier 2008).

143

6.2 Le vrai âge du vrai Boreogadus saida

L’étude des jeunes stades de Boreogadus saida aura gagné en précision au cours de mon doctorat, d’abord avec la validation de la déposition journalière d’anneaux de croissance sur les otolithes (chapitre 2) et ensuite avec le développement d’une méthode permettant de différencier, à l’aide de caractéristiques simples du noyau de l’otolithe, Boreogadus saida d’Arctogadus glacialis (chapitre 4). L’estimation de l’âge des larves de poissons par le décompte des anneaux journaliers des otolithes nécessite pour chaque espèce une validation de la nature journalière de la déposition des anneaux de croissance (Geffen 1992). Cette validation, qui peut se faire de différentes manières, permet de détecter d’éventuels anneaux subjournaliers et de vérifier à quel âge se forme le premier anneau. Pour Boreogadus, la validation a été effectuée d’abord par le marquage d’otolithes à l’aide d’un composé chimique fluorescent (oxytétracycline), puis par la comparaison des anneaux de croissance observés au microscope optique et au microscope électronique (chapitre 2).

Les stades larvaires et juvéniles de Boreogadus et d’Arctogadus se retrouvent souvent ensemble sur les plateaux continentaux des mers arctiques et sont morphologiquement très similaires. Les collections présumées monospécifiques de Boreogadus contiennent donc régulièrement une certaine proportion d’Arctogadus (chapitre 2). L’utilisation de plusieurs marqueurs moléculaires pour identifier les deux espèces a permis de développer une méthode utilisant la taille du noyau de l’otolithe pour différencier les deux espèces (chapitre 4), de quantifier l’abondance d’Arctogadus dans nos échantillons de la mer de Beaufort, et d’étudier pour la première fois l’écologie larvaire et juvénile d’Arctogadus (chapitre 5). Le taux de succès de la méthode d’identification basée sur les otolithes (96%), combiné à sa rapidité, son faible coût et sa simplicité, en font un outil de choix qui permettra à l’avenir une détermination plus exacte des taux de croissance et de survie des jeunes stades des deux gadidés arctiques, de même qu’une amélioration de la robustesse et de la performance des modèles bio-énergétiques (e.g. Thanassekos & Fortier 2012, Thanassekos et al. 2012).

144

6.3 Quel futur pour les gadidés arctiques?

L’Arctique fait face à une pléthore de changements lesquels, amorcés ou anticipés, biologiques ou physiques, à l’échelle cellulaire ou globale, se combinent en un maelstrom d’effets potentiels pour son écosystème: réduction de la glace de mer, augmentation des précipitations et des apports fluviaux, introduction d’espèces boréales, accroissement de l’exploitation des ressources, acidification océanique, modifications des courants marins, etc. Dans ce contexte, champagne pour celui qui prédira avec exactitude le sort de Boreogadus saida et d’Arctogadus glacialis! Même si un certain consensus prévoit à moyen terme le remplacement de Boreogadus par des espèces compétitrices comme le capelan (Mallotus villosus) (e.g. Provencher et al. 2012), et prédatrices comme la morue franche (Gadus morhua) et l’aiglefin (Melanogrammus aeglefinus) (e.g. Renaud et al. 2012), de nombreux scénarios demeurent envisageables entretemps.

La réduction du couvert de glace peut être considérée sous plusieurs angles lorsqu’il s’agit de prédire ses effets, positifs ou négatifs, sur Boreogadus et Arctogadus. D’un côté, une débâcle plus hâtive et des polynies hivernales plus fréquentes pourraient favoriser la survie des larves écloses en hiver (chapitres 2 et 5). Mais d’un autre côté, un couvert de glace réduit est potentiellement synonyme de prédation accrue pour Boreogadus, autant pour les oeufs, qui flottent normalement sous la glace, que pour certains jeunes adultes habitant temporairement les interstices de la glace (Bluhm & Gradinger 2008). La relation entre la glace et la prédation sur les individus adultes est encore plus ambiguë puisque le couvert de glace hivernale sert à la fois de plateforme de chasse aux phoques annelés (Pusa hispida) se nourrissant dans les agrégations profondes de Boreogadus, et de barrière à une telle prédation de par la diminution de la quantité de lumière disponible aux phoques pour chasser en profondeur (Benoit et al. 2010, Geoffroy et al. 2011).

La survie des jeunes stades de Boreogadus et d’Arctogadus pourrait également être modulée par d’éventuelles modifications dans la communauté zooplanctonique. Par exemple, l’augmentation des remontées d’eaux profondes causée par la réduction du couvert de glace enrichie les eaux de surface en nutriments, ce qui favorise la croissance des algues de glace (Tremblay et al. 2011). Cette accroissement de biomasse algale sous la

145 glace à l’hiver et au printemps pourrait entrainer une augmentation de la quantité d’œufs et de nauplii de copépodes, principales proies des larves de Boreogadus et d’Arctogadus. Dans cette situation, un meilleur taux de survie des larves écloses en hiver et au printemps est envisagé. Par contre, dans un cas où la glace fond très hâtivement au printemps, la production d’algues de glace et des copépodes qui en dépendent pourrait devenir très faible et la survie des larves, particulièrement celles d’Arctogadus, fortement réduite.

Selon un scénario dans lequel la concentration atmosphérique de CO2 atteint le double de sa valeur préindustrielle (vraisemblablement d’ici 2050), l’écoulement fluvial annuel dans l’océan Arctique pourrait augmenter de 10 à 20%, et l’écoulement fluvial hivernal de 150 à 200% (Shiklomanov & Shiklomanov 2003). L’impact de tels changements sur les jeunes stades de Boreogadus pourrait aussi s’avérer dichotomique. D’une part, des débits fluviaux hivernaux accrus provoqueront un agrandissement de l’aire des refuges thermiques potentiellement disponibles pour les larves, ce qui pourrait favoriser leur survie. D’autre part, la faible intensité lumineuse d’un panache fluvial causée par sa forte turbidité peut entraîner une diminution du succès d’alimentation et de la survie des Boreogadus larvaires (Gilbert et al. 1992, Fortier et al. 1996).

L’accessibilité accrue de l’Arctique induite par la réduction du couvert de glace a entraîné l’accroissement du trafic maritime et du nombre de projets d’exploration et d’exploitation des ressources minières et pétrolières (Lasserre 2010, Lasserre & Pelletier 2011). Les divers polluants libérés dans l’environnement par ces activités pourraient causer des effets négatifs sur plusieurs constituants de l’écosystème marin arctique. Notamment, les hydrocarbures aromatiques polycycliques (HAP) sont connus pour entraver, entre autres, la croissance et la survie des poissons adultes (e.g. Sol et al. 2000), et des jeunes stades (e.g. Ingvarsdóttir et al. 2012). Finalement, Boreogadus n’est pas à l’abri de l’exploitation. Un brin de futurologie nous laissera entrevoir aisément le jour où l’industrie des pêches se tournera vers cette ressource naturelle encore très abondante. Mais l’espèce, un petit gadidé pélagique fortement interrelié à l’intérieur de son réseau trophique (Welch et al. 1992), se montrera probablement très sensible à la récolte (Smith et al. 2011).

146

6.4 Perspectives de recherche

En terminant, j’exposerai quelques avenues de recherche, lesquelles complémentent mes travaux de doctorat ou s’attaquent à des sujets qui me paraissent particulièrement importants dans le contexte actuel.

L’écologie larvaire de Boreogadus saida et d’Arctogadus glacialis fait intervenir plusieurs paramètres physiques interreliés. Ainsi, la température, la salinité, la glace, la turbidité, la quantité de lumière et les courants interagissent, entre eux et avec des facteurs biologiques, dans le déterminisme de la survie larvaire et juvénile. Pour mieux comprendre et quantifier l’effet de chacune de ces variables, une approche privilégiée inclurait des campagnes d’échantillonnage plus ciblées s’attardant aux régions près des embouchures des grands fleuves, idéalement associées à des travaux expérimentaux en microcosmes. Une paramétrisation précise de ces facteurs permettrait d’accroitre la fiabilité des modèles individus-centrés ou écosystémiques. Les études futures consacrées aux jeunes stades de Boreogadus devraient inclure l’identification à l’espèce d’une partie des spécimens, génétiquement ou à l’aide des otolithes. Continuer d’étudier l’écologie des jeunes stades d’Arctogadus et les interactions possibles avec Boreogadus m’apparaît aussi essentiel à la compréhension de l’écosystème marin arctique dans son ensemble.

En ce qui concerne les adultes Boreogadus et Arctogadus, leur monitorage à long terme et à grande échelle devrait constituer une priorité. Pour ce faire, l’estimation des biomasses par acoustique serait probablement la méthode la plus appropriée compte tenu des conditions difficiles d’échantillonnage en Arctique. Des études plus détaillées de la structure des populations utilisant la génétique et/ou la chimie des otolithes seraient aussi très importantes. Une approche multidisciplinaire d’étude des stocks intégrant plusieurs types de données (e.g. Higgins et al. 2010, Smith & Campana 2010) pourrait aussi aider à mieux comprendre la stratégie de reproduction, la ségrégation spatiale et les comportements migratoires des deux espèces. Les gadidés arctiques vivent dans un environnement changeant à grande vitesse et l’atteinte d’un niveau de connaissance suffisant pour permettre une gestion éclairée des ressources naturelles de l’océan Arctique constitue donc un objectif prioritaire à court terme.

147

Vu l’accélération du réchauffement climatique et la disparition imminente du couvert de glace estivale, la prochaine décennie pourrait marquer un tournant pour les écosystèmes marins arctiques. Et à ces bouleversements se surimposeront tôt ou tard les effets de l’augmentation des activités économiques dans l’Arctique. Le rôle clé de Boreogadus au sein de ces écosystèmes implique que tout changement dans cette espèce, aussi subtil soit- il, pourrait avoir des répercussions majeures sur l’ensemble du système. Quant à Arctogadus, hyperspécialiste de l’Arctique, son importance sur la structure et le fonctionnement de l’écosystème demeure pratiquement inconnue mais pourrait apparaître disproportionnée relativement à sa faible biomasse. Dans ce contexte, prédire l’avenir des gadidés arctiques représente un objectif capital mais empreint de complexité et qui nécessitera davantage d’efforts concertés misant sur le caractère pluridisciplinaire et international de la recherche.

148

Bibliographie

CAFF 2001. Arctic Flora and Fauna: Status and Conservation. Conservation of Arctic Flora and Fauna. Helsinki, Edita. http://www.grida.no/graphicslib/detail/major-river- systems-in-the-arctic_2c11

Able KW (2005) A re-examination of fish estuarine dependence: evidence for connectivity between estuarine and ocean habitats. Estuarine, Coastal and Shelf Science 64:5-17

Altukhov KA (1979) The reproduction and development of the Arctic cod, Boreogadus saida, in the White Sea. Journal of Ichthyology 19:93-101

Altukhov KA (1981) The reproduction and development of the arctic cod, Boreogadus saida, in the White Sea. Journal of Ichthyology 19:93 -101

Andriyashev A (1970) Cryopelagic fishes of the Arctic and Antarctic and their significance in polar ecosystems. In: Holdgate M (ed) Antarctic ecology, Book 1. Academic Press, New York, p. 297-304

Andriyashev AP (1954) Fishes of the northern seas of the U.S.S.R. Israel Program for Scientific Translations, Jerusalem, 617 p.

Antonov JI, Locarnini RA, Boyer TP, Mishonov AV, Garcia HE (2006) World Ocean Atlas 2005, Volume 2: Salinity. In: Levitus S. (ed.) NOAA Atlas NESDIS 62, US Government Printing Office, Washington, DC, 182 p.

Aronovich T, Doroshev S, Spectorova L, Makhotin V (1975) Egg incubation and larval rearing of Navaga (Eleginus Navaga Pall.), Polar cod (Boreogadus Saida Lepechin) and Arctic flounder (Liopsetta Glacialis Pall.) in the laboratory. Aquaculture 6:233- 242

Arrigo KR, van Dijken G, Pabi S (2008) Impact of a shrinking Arctic ice cover on marine primary production. Geophysical Research Letters 35:L19603

Arrigo KR, van Dijken GL (2004) Annual cycles of sea ice and phytoplankton in Cape Bathurst polynya, southeastern Beaufort Sea, Canadian Arctic. Geophysical Research Letters 31:L08304

Aschan M, Karamushko O, Byrkjedal I, Wienerroither R, Borkin I, Christiansen J (2009) Records of the gadoid fish Arctogadus glacialis (Peters, 1874) in the European Arctic. Polar Biology 32:963-970

Ashford JR, Jones CM, Hofmann EE, Everson I, Moreno CA, Duhamel G, Williams R (2008) Otolith chemistry indicates population structuring by the Antarctic

149

Circumpolar Current. Canadian Journal of Fisheries and Aquatic Sciences 65:135- 146

Bailey K, Houde E (1989) Predation on eggs and larvae of marine fishes and the recruitment problem. Advances in Marine Biology 25:1-67

Bain H, Sekerak A (1978) Aspects of biology of Arctic cod, Boreogadus saida, in the central Canadian Arctic. LGL ltd, Toronto, 104 p.

Baranenkova AS, Ponomarenko VP, Khokhlina NS (1966) The distribution, size and growth of the larvae and fry of Boreogadus saida (Lep.) in the Barents Sea. Voprosy Ikhtiologii 6:498-518. Fisheries and Marine Service Translation Series 4025, 37 p.

Barber DG, Asplin MG, Gratton Y, Lukovich JV, Galley RJ, Raddatz RL, Leitch D (2010) The International Polar Year (IPY) Circumpolar Flaw Lead (CFL) System Study: Overview and the Physical System. Atmosphere-ocean 48:225-243

Barber DG, Hanesiak JM (2004) Meteorological forcing of sea ice concentrations in the southern Beaufort Sea over the period 1979 to 2000. Journal of Geophysical Research-Oceans 109:C06014

Barber DG, Massom RA (2007) The role of sea ice in Arctic and Antarctic polynyas. In: Smith WO, Barber DG (eds) Polynyas: windows to the World. Elsevier Oceanography Series 74, Amsterdam, p. 1-54

Bareiss J, Gorgen K (2005) Spatial and temporal variability of sea ice in the Laptev Sea: analyses and review of satellite passive-microwave data and model results, 1979 to 2002. Global and Planetary Change 48:28-54

Barrett RT, Furness RW (1990) The prey and diving depths of seabirds on Hornøy, north Norway after a decrease in the Barents Sea capelin stocks. Ornis Scandinavica 21:179-186

Bath GE, Thorrold SR, Jones CM, Campana SE, McLaren JW, Lam JWH (2000) Strontium and barium uptake in aragonitic otoliths of marine fish. Geochimica et Cosmochimica Acta 64:1705-1714

Begg GA, Campana SE, Fowler AJ, Suthers IM (2005) Otolith research and application: current directions in innovation and implementation. Marine and Freshwater Research 56:477-483

Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F (2004) GENETIX 4.05, logiciel sous Windows TM pour la génétique des populations. Laboratoire Génome, Populations, Interactions, CNRS UMR 5171, Université de Montpellier II, Montpellier (France).

150

Belkin IM (2009) Rapid warming of large marine ecosystems. Progress in Oceanography 81:207-213

Bell GI, Jurka J (1997) The length distribution of perfect dimer repetitive DNA is consistent with its evolution by an unbiased single-step mutation process. Journal of Molecular Evolution 44:414-421

Benoit D, Simard Y, Fortier L (2008) Hydro-acoustic detection of large winter aggregations of Arctic cod (Boreogadus saida) at depth in ice-covered Franklin Bay (Beaufort Sea). Journal of Geophysical Research-Oceans 113:C06S90

Benoit D, Simard Y, Gagné J, Geoffroy M, Fortier L (2010) From polar night to midnight sun: photoperiod, seal predation, and the diel vertical migrations of polar cod (Boreogadus saida) under landfast ice in the Arctic Ocean. Polar Biology 33:1505- 1520

Bluhm BA, Gradinger R (2008) Regional variability in food availability for Arctic marine mammals. Ecological Applications 18:S77-S96

Boe J, Hall A, Qu X (2009) September sea-ice cover in the Arctic Ocean projected to vanish by 2100. Nature Geoscience 2:341-343

Bouchard C (2007) Saison d'éclosion, croissance initiale et survie de la morue arctique (Boreogadus saida) en mer de Laptev : impact des polynies. Mémoire de maîtrise, Université Laval, Québec, 55 p.

Bouchard C, Fortier L (2008) Effects of polynyas on the hatching season, early growth and survival of polar cod Boreogadus saida in the Laptev Sea. Marine Ecology Progress Series 355:247-256

Bouchard C, Fortier L (2011) Circum-arctic comparison of the hatching season of polar cod Boreogadus saida: A test of the freshwater winter refuge hypothesis. Progress in Oceanography 90:105-116

Bouchard C, Robert D, Nelson RJ, Fortier L (2013) The nucleus of the lapillar otolith discriminates the early life stages of Boreogadus saida and Arctogadus glacialis. Polar Biology

Boulva J (1970) The systematics of sympatric populations of the Arctic codfishes, Arctogadus borisovi Drjagin and A. glacialis (Peters), from Cambridge Bay, N.W.T. M. Sc. Thesis, Dalhousie University, Halifax, Canada:122 p.

Bourque J, Dodson J, Ryan D, Marcogliese D (2006) Cestode parasitism as a regulator of early life-history survival in an estuarine population of rainbow smelt Osmerus mordax. Marine Ecology Progrss Series 314:295-307

151

Boyer TP, Antonov JI, Garcia HE, Johnson DR, Locarnini RA, Mishonov AV, Pitcher MT, Baranova OK, Smolyar IV (2006) World Ocean Database 2005. In: Levitus S. (ed.), NOAA Atlas NESDIS 60, US Government Printing Office, Washington, DC, 190 p.

Bradstreet MSW (1982) Occurrence, habitat use, and behavior of seabirds, marine mammals and Arctic cod at the Pond Inlet ice edge. Arctic 35:28-40

Bradstreet MSW, Cross WE (1982) Trophic relationships at high Arctic ice edges. Arctic 35:1-12

Bradstreet MSW, Finley KJ, Sekerak AD, Griffiths WD, Evans CR, Fabijan MF, Stallard HE (1986) Aspects of the biology of Arctic cod Boreogadus saida and its importance in Arctic marine food chains. Canadian Technical Report of Fisheries and Aquatic Sciences 1491, 193 p.

Breines R, Ursvik A, Nymark M, Johansen S, Coucheron D (2008) Complete mitochondrial genome sequences of the Arctic Ocean codfishes Arctogadus glacialis and Boreogadus saida reveal oriL and tRNA gene duplications. Polar Biology 31:1245-1252

Brierley AS, Kingsford MJ (2009) Impacts of climate change on marine organisms and ecosystems. Current Biology 19:R602-R614

Brophy D, Jeffries TE, Danilowicz BS (2004) Elevated manganese concentrations at the cores of clupeid otoliths: possible environmental, physiological, or structural origins. Marine Biology 144:779-786

Brunner E, Domhof S, Langer F (2002) Nonparametric analysis of longitudinal data in factorial experiments. John Wiley & Sons, New York, 288 p.

Buckley LJ, Caldarone EM, Lough RG, St Onge-Burns JM (2006) Ontogenetic and seasonal trends in recent growth rates of Atlantic cod and haddock larvae on Georges Bank: effects of photoperiod and temperature. Marine Ecology Progress Series 325:205-226

Butler JL (1992) Collection and preservation of material for otolith analysis. In: Stevenson DK, Campana SE (eds) Otolith microstructure examination and analysis. Canadian Special Publication of Fisheries and Aquatic Sciences 117, p. 13-17

Campana S, Gagné J, Munro J (1987) Otolith microstructure of larval herring (Clupea harengus): image or reality? Canadian Journal of Fisheries and Aquatic Sciences 44:1922-1929

Campana SE (1999) Chemistry and composition of fish otoliths: pathways, mechanisms and applications. Marine Ecology Progress Series 188:263-297

152

Campana SE (2001) Accuracy, precision, and quality control in age determination, including a review of the use and abuse of age validation methods. Journal of Fish Biology 59:197-242

Campana SE (2005) Otolith science entering the 21st century. Marine and Freshwater Research 56:485-495

Campana SE, Annand MC, McMillan JI (1995) Graphical and statistical methods for determining the consistency of age determinations. Transactions of the American Fisheries Society 124:131-138

Campana SE, Jones CM (1992) Analysis of otolith microstructure data. In: Stevenson DK, Campana SE (eds) Otolith microstructure examination and analysis. Canadian Special Publication of Fisheries and Aquatic Sciences 117, p. 73-100

Campana SE, Neilson JD (1985) Microstructure of fish otoliths. Canadian Journal of Fisheries and Aquatic Sciences 42:1014-1032

Campana SE, Thorrold SR (2001) Otoliths, increments, and elements: keys to a comprehensive understanding of fish populations? Canadian Journal of Fisheries and Aquatic Sciences 58:30-38

Campana SE, Thorrold SR, Jones CM, Gunther D, Tubrett M, Longerich H, Jackson S, Halden NM, Kalish JM, Piccoli P, dePontual H, Troadec H, Panfili J, Secor DH, Severin KP, Sie SH, Thresher R, Teesdale WJ, Campbell JL (1997) Comparison of accuracy, precision, and sensitivity in elemental assays of fish otoliths using the electron microprobe, proton-induced X-ray emission, and laser ablation inductively coupled plasma mass spectrometry. Canadian Journal of Fisheries and Aquatic Sciences 54:2068-2079

Carlsson J, McDowell JR, Diaz-Jaimes P, Carlsson JEL, Boles SB, Gold JR, Graves JE (2004) Microsatellite and mitochondrial DNA analyses of Atlantic bluefin tuna (Thunnus thynnus thynnus) population structure in the Mediterranean Sea. Molecular Ecology 13:3345-3356

Chistiakov DA, Hellemans B, Volckaert FAM (2006) Microsatellites and their genomic distribution, evolution, function and applications: a review with special reference to fish genetics. Aquaculture 255:1-29

Christiansen J, Hop H, Nilssen E, Joensen J (2012) Trophic ecology of sympatric Arctic gadoids, Arctogadus glacialis (Peters, 1872) and Boreogadus saida (Lepechin, 1774), in NE Greenland. Polar Biology 35:1247-1257

Cimmaruta R, Bondanelli P, Nascetti G (2005) Genetic structure and environmental heterogeneity in the European hake (Merluccius merluccius). Molecular Ecology 14:2577-2591

153

Clarke LM, Walther BD, Munch SB, Thorrold SR, Conover DO (2009) Chemical signatures in the otoliths of a coastal marine fish, Menidia menidia, from the northeastern United States: spatial and temporal differences. Marine Ecology Progress Series 384:261-271

Coad B, Waszczuk H, Labignan I (1995) Encyclopedia of Canadian Fishes. Canadian Museum of Nature & Canadian Sportfishing Productions Inc., Ottawa, 928 p.

Conover DO (1992) Seasonality and the scheduling of life history at different latitudes. Journal of Fish Biology 41:161-178

Craig PC, Griffiths WB, Haldorson L, McElderry H (1982) Ecological studies of Arctic cod (Boreogadus saida) in Beaufort Sea coastal water. Canadian Journal of Fisheries and Aquatic Sciences 39:395-406

Crawford R, Jorgenson J (1990) Density distribution of fish in the presence of whales at the Admiralty Inlet landfast ice edge. Arctic 43:215-222

Crawford RE, Jorgenson JK (1993) Schooling behaviour of Arctic cod, Boreogadus saida, in relation to drifting pack ice. Environmental Biology of Fishes 36:345-357

Crawford RE, Jorgenson JK (1996) Quantitative studies of Arctic cod (Boreogadus saida) schools: important energy stores in the Arctic food web. Arctic 49:181-193

Cury P, Bakun A, Crawford RJM, Jarre A, Quiñones RA, Shannon LJ, Verheye HM (2000) Small pelagics in upwelling systems: patterns of interaction and structural changes in “wasp-waist” ecosystems. ICES Journal of Marine Science 57:603-618

Cushing D (1990) Plankton production and year-class strength in fish populations: an update of the match/mismatch hypothesis. Advances in Marine Biology 26:249-294

D'Alessandro E, Sponaugle S, Cowen R (2013) Selective mortality during the larval and juvenile stages of snappers (Lutjanidae) and great barracuda Sphyraena barracuda Marine Ecology Progress Series 474:227-242

Darnis G (2013) Migration verticale du zooplancton et flux respiratoire de carbone en mer de Beaufort (Arctique canadien). Thèse de doctorat, Université Laval, Québec, 159 p.

Deming J, Fortier L, Fukuchi M (2002) The International North Water Polynya Study (NOW): a brief overview. Deep-Sea Research Part II -Topical Studies in Oceanography 49:4887-4892

Déry SJ, Stieglitz M, McKenna EC, Wood EF (2005) Characteristics and trends of river discharge into Hudson, James, and Ungava Bays, 1964-2000. Journal of Climate 18:2540-2557

154

DeWoody JA, Avise JC (2000) Microsatellite variation in marine, freshwater and anadromous fishes compared with other animals. Journal of Fish Biology 56:461- 473

DiMaria R, Miller J, Hurst T (2010) Temperature and growth effects on otolith elemental chemistry of larval Pacific cod, Gadus macrocephalus. Environmental Biology of Fishes 89:453-462

Dmitrenko IA, Polyakov IV, Kirillov SA, Timokhov LA, Frolov IE, Sokolov VT, Simmons HL, Ivanov VV, Walsh D (2008) Toward a warmer Arctic Ocean: spreading of the early 21st century Atlantic Water warm anomaly along the Eurasian Basin margins. Journal of Geophysical Research-Oceans 113:C05023

Dodson JJ (1997) Migration in fish: an evolutionary perspective. In: Godin J-GJ (ed) Behavioural ecology of teleost fishes. Oxford University Press, Oxford, p. 10-36

Drolet R, Fortier L, Ponton D, Gilbert M (1991) Production of fish larvae and their prey in subarctic southeastern Hudson Bay. Marine Ecology Progress Series 77:105-118

Dyurgerov M, Bring A, Destouni G (2010) Integrated assessment of changes in freshwater inflow to the Arctic Ocean. Journal of Geophysical Research-Atmospheres 115:D12116

Dyurgerov MB, Carter CL (2004) Observational evidence of increases in freshwater inflow to the Arctic Ocean. Arctic Antarctic and Alpine Research 36:117-122

Edwards M, Richardson AJ (2004) Impact of climate change on marine pelagic phenology and trophic mismatch. Nature 430:881-884

Ellegren H (2000) Microsatellite mutations in the germline: implications for evolutionary inference. Trends in Genetics 16:551-558

Elsdon TS, Gillanders BM (2002) Interactive effects of temperature and salinity on otolith chemistry: challenges for determining environmental histories of fish. Canadian Journal of Fisheries and Aquatic Sciences 59:1796

Elsdon TS, Gillanders BM (2003a) Reconstructing migratory patterns of fish based on environmental influences on otolith chemistry. Reviews in Fish Biology and Fisheries 13:217-235

Elsdon TS, Gillanders BM (2003b) Relationship between water and otolith elemental concentrations in juvenile black bream Acanthopagrus butcheri. Marine Ecology Progress Series 260:263-272

Elsdon TS, Gillanders BM (2004) Fish otolith chemistry influenced by exposure to multiple environmental variables. Journal of Experimental Marine Biology and Ecology 313:269-284

155

Elsdon TS, Wells BK, Campana SE, Gillanders BM, Jones CM, Limburg KE, Secor DH, Thorrold SR, Walther BD (2008) Otolith chemistry to describe movements and life- history parameters of fishes: Hypotheses, assumptions, limitations, and inferences. Oceanography and Marine Biology: An Annual Review 46:297-330

Fahay MP (2007) Early Stages of Fishes in the Western North Atlantic Ocean (Davis Strait, Southern Greenland and Flemish Cap to Cape Hatteras), 1696 p.

Falk K, Pedersen CE, Kampp K (2000) Measurements of diving depth in Dovekies (Alle alle). Auk 117:522-525

FAO (2012) Fisheries and Aquaculture Department species fact sheet: Boreogadus saida (Lepechin 1774). Site de la Food and Agriculture Organization of the United Nations, (En ligne), Page consultée le 4 avril, 2012. http://www.fao.org/fishery/species/2233/en

Fevolden SE, Christiansen JS (1997) Allozymic and scnDNA homogeneity in Polar cod (Boreogadus saida)(Gadiformes: Gadidae). Cybium 21:411-414

Fevolden SE, Martinez I, Christiansen JS (1999) RAPD and scnDNA analyses of polar cod, Boreogadus saida (Pisces, Gadidae), in the North Atlantic. Sarsia 84:99-103

Finley KJ, Bradstreet MSW, Miller GW (1990) Summer feeding ecology of harp seals (Phoca groenlandica) in relation to Arctic cod (Boreogadus saida) in the Canadian high Arctic. Polar Biology 10:609-618

Finley KJ, Gibb EJ (1982) Summer diet of the narwhal (Monodon monoceros) in Pond Inlet, northern Baffin Island. Canadian Journal of Zoology 60:3353-3363

Fogarty MJ, Myers RA, Bowen KG (2001) Recruitment of cod and haddock in the North Atlantic: a comparative analysis. ICES Journal of Marine Science 58:952-961

Folkvord A, Johannessen A, Moksness E (2004) Temperature-dependent otolith growth in Norwegian spring-spawning herring (Clupea harengus L.) larvae. Sarsia 89:297- 310

Fortier L, Gilbert M, Ponton D, Ingram RG, Robineau B, Legendre L (1996) Impact of freshwater on a subarctic coastal ecosystem under seasonal sea ice (southeastern Hudson Bay, Canada). III. Feeding success of marine fish larvae. Journal of Marine Systems 7:251-265

Fortier L, Harris RP (1989) Optimal foraging and density-dependent competition in marine fish larvae. Marine Ecology Progress Series 51:19-33

Fortier L, Ponton D, Gilbert M (1995) The match/mismatch hypothesis and the feeding success of fish larvae in ice-covered southeastern Hudson Bay. Marine Ecology Progress Series 120:11-27

156

Fortier L, Quiñonez-Velazquez C (1998) Dependence of survival on growth in larval pollock Pollachius virens and haddock Melanogrammus aeglefinus: a field study based on individual hatchdates. Marine Ecology Progress Series 174:1-12

Fortier L, Sirois P, Michaud J, Barber D (2006) Survival of Arctic cod larvae (Boreogadus saida) in relation to sea ice and temperature in the Northeast Water Polynya (Greenland Sea). Canadian Journal of Fisheries and Aquatic Sciences 63:1608-1616

Fortier M, Fortier L, Hattori H, Saito H, Legendre L (2001) Visual predators and the diel vertical migration of copepods under Arctic sea ice during the midnight sun. Journal of Plankton Research 23:1263-1278

Fossum P, Kalish J, Moksness E (2000) Special issue: 2nd International Symposium on Fish Otolith Research and Application, Bergen, Norway, 20-25 June 1998 - Foreword. Fisheries Research 46:1-2

Frost KJ, Lowry LF (1984) Trophic relationships of vertebrate consumers in the Alaskan Beaufort Sea. In: Barnes PW, Schell DM, Reimnitz E (eds) The Alaskan Beaufort Sea ecosystems and environments. Academic Press, London, UK, p. 383-402

Gaggiotti OE, Bekkevold D, Jørgensen HBH, Foll M, Carvalho GR, Andre C, Ruzzante DE (2009) Disentangling the effects of evolutionary, demographic, and environmental factors influencing genetic structure of natural populations: Atlantic herring as a case study Evolution 63:2939-2951

Galley RJ, Key E, Barber DG, Hwang BJ, Ehn JK (2008) Spatial and temporal variability of sea ice in the southern Beaufort Sea and Amundsen Gulf: 1980-2004. Journal of Geophysical Research-Oceans 113:C05S95

Gardner AS, Moholdt G, Wouters B, Wolken GJ, Burgess DO, Sharp MJ, Cogley JG, Braun C, Labine C (2011) Sharply increased mass loss from glaciers and ice caps in the Canadian Arctic Archipelago. Nature advance online publication

Garneau M-È, Roy S, Lovejoy C, Gratton Y, Vincent WF (2008) Seasonal dynamics of bacterial biomass and production in a coastal Arctic ecosystem: Franklin Bay, western Canadian Arctic. Journal of Geophysical Research-Oceans 113:C07S91

Gaston AJ, Gilchrist HG, Hipfner JM (2005) Climate change, ice conditions and reproduction in an Arctic nesting marine bird: Brunnich's guillemot (Uria lomvia L.). Journal of Animal Ecology 74:832-841

Geffen AJ (1992) Validation of otolith increment deposition rate. In: Stevenson DK, Campana SE (eds) Otolith microstructure examination and analysis. Canadian Special Publication of Fisheries and Aquatic Sciences 117, p. 101-113

157

Geoffroy M, Robert D, Darnis G, Fortier L (2011) The aggregation of polar cod (Boreogadus saida) in the deep Atlantic layer of ice-covered Amundsen Gulf (Beaufort Sea) in winter. Polar Biology 34:1959-1971

Gilbert M, Fortier L, Ponton D, Drolet R (1992) Feeding ecology of marine fish larvae across the Great Whale River plume in seasonally ice-covered southeastern Hudson Bay. Marine Ecology Progress Series 84:19-30

Gjøsæter H (2009) Commercial fisheries (fish, seafood and marine mammals). In: Sakshaug E, Johnsen G, Kovacs K (eds) Ecosystem Barents Sea. Tapir Academic Press, Trondheim, Norway, p. 373-414

Gordeev VV (2006) Fluvial sediment flux to the Arctic Ocean. Geomorphology 80:94-104

Gradinger RR, Bluhm BA (2004) In-situ observations on the distribution and behavior of amphipods and Arctic cod (Boreogadus saida) under the sea ice of the high Arctic Canada Basin. Polar Biology 27:595-603

Graham M, Hop H (1995) Aspects of reproduction and larval biology of Arctic cod (Boreogadus saida). Arctic 48:130-135

Grebmeier JM, Overland JE, Moore SE, Farley EV, Carmack EC, Cooper LW, Frey KE, Helle JH, McLaughlin FA, McNutt SL (2006) A major ecosystem shift in the northern Bering Sea. Science 311:1461-1464

Guay CK, Kenison Falkner K (1997) Barium as a tracer of Arctic halocline and river waters. Deep-Sea Research Part II -Topical Studies in Oceanography 44:1543-1569

Hegseth EN, Sundfjord A (2008) Intrusion and blooming of Atlantic phytoplankton species in the high Arctic. Journal of Marine Systems 74:108-119

Hemmer-Hansen J, Nielsen EE, Grønkjær P, Loeschcke V (2007) Evolutionary mechanisms shaping the genetic population structure of marine fishes: lessons from the European flounder (Platichthys flesus L.). Molecular Ecology 16:3104-3118

Higgins RM, Danilowicz BS, Balbuena JA, Danielsdottir AK, Geffen AJ, Meijer WG, Modin J, Montero FE, Pampoulie C, Perdiguero-Alonso D, Schreiber A, Stefansson MO, Wilson B (2010) Multi-disciplinary fingerprints reveal the harvest location of cod Gadus morhua in the northeast Atlantic. Marine Ecology Progress Series 404:197-206

Hjort J (1914) Fluctuations in the great fisheries of northern Europe. Rapports et Procès- Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 20:1- 228

158

Hobson K, Welch H (1992a) Determination of trophic relationships within a high Arctic marine food web using delta-13 C and delta-15 N analysis. Marine Ecology Progress Series 84:9-18

Hobson K, Welch H (1992b) Observations of foraging northern fulmars (Fulmarus glacialis) in the Canadian high Arctic. 45:150-153

Hobson KA, Fisk A, Karnovsky N, Holst M, Gagnon J-M, Fortier M (2002) A stable isotope (δ13C, δ15N) model for the North Water food web: implications for evaluating trophodynamics and the flow of energy and contaminants. Deep Sea Research Part II: Topical Studies in Oceanography 49:5131-5150

Høie H, Folkvord A, Johannessen A (1999) Maternal, paternal and temperature effects on otolith size of young herring (Clupea harengus L.) larvae. Journal of Experimental Marine Biology and Ecology 234:167-184

Holland MM, Bitz CM, Tremblay B (2006) Future abrupt reductions in the summer Arctic sea ice. Geophysical Research Letters 33:L23503

Holliday FGT (1969) The effects of salinity on the eggs and larvae of teleosts. In: Hoar WS, Randall DJ (eds) Fish Physiology, Vol. 1, . Academic Press, London, p. 293- 311

Holst M, Stirling I, Hobson KA (2001) Diet of ringed seals (Phoca hispida) on the east and west sides of the North Water Polynya, northern Baffin Bay. Marine Mammal Science 17:888-908 Hop H, Graham M (1995) Respiration of juvenile Arctic cod (Boreogadus saida): effects of acclimation, temperature, and food intake. Polar Biology 15:359-367

Hop H, Graham M, Trudeau VL (1995) Spawning energetics of Arctic cod (Boreogadus saida) in relation to seasonal development of the ovary and plasma-sex steroid levels. Canadian Journal of Fisheries and Aquatic Sciences 52:541-550

Hop H, Mundy CJ, Gosselin M, Rossnagel AL, Barber DG (2011) Zooplankton boom and ice amphipod bust below melting sea ice in the Amundsen Gulf, Arctic Canada. Polar Biology 34:1947-1958

Hourigan TF, Radtke RL (1989) Reproduction of the Antarctic fish Nototheniops nudifrons. Marine Biology 100:277-283

Ingvarsdóttir A, Bjørkblom C, Ravagnan E, Godal BF, Arnberg M, Joachim DL, Sanni S (2012) Effects of different concentrations of crude oil on first feeding larvae of Atlantic herring (Clupea harengus). Journal of Marine Systems 93:69-76

IPCC (2013) Climate Change 2013: The Physical Science Basis. Working Group I Contribution to the IPCC 5th Assessment Report, Intergovernmental Panel on Climate Change, Stockholm, Sweden, 2216 p.

159

Jackson JR (2007) Earliest references to age determination of fishes and their early application to the study of fisheries. Fisheries 32:321-328

Jakobsson M, Grantz A, Kristoffersen Y, Macnab R (2003) Bathymetry and physiography of the Arctic Ocean and its constituent Seas. In: Stein R, Macdonald RW (eds) The organic carbon cycle in the Arctic Ocean. Springer, Berlin Heidelberg New York, p. 1-6

Jensen AS (1948) Contribution to the ichtyofauna of Greenland, 8-24. Spolia Zoologica Musei Hauniensis 9:1-182

Ji R, Jin M, Varpe Ø (2013) Sea ice phenology and timing of primary production pulses in the Arctic Ocean. Global Change Biology 19:734-741

Jones C (1986) Determining age of larval fish with the otolith increment technique. Fishery Bulletin 84:91-104

Jones C (1992) Development and application of the otolith increment technique. In: Stevenson DK, Campana SE (eds) Otolith microstructure examination and analysis. Canadian Special Publication of Fisheries and Aquatic Sciences 117, p. 1-11

Jones C, Brothers E (1987) Validation of the otolith increment aging technique for striped bass, Morone saxatilis, larvae reared under suboptimal feeding conditions. Fishery Bulletin 85:171-178

Jordan AD, Møller PR, Nielsen JG (2003) Revision of the Arctic cod genus Arctogadus. Journal of Fish Biology 62:1339-1352

Jørgensen HBH, Hansen MM, Bekkevold D, Ruzzante DE, Loeschcke V (2005) Marine landscapes and population genetic structure of herring (Clupea harengus L.) in the Baltic Sea. Molecular Ecology 14:3219-3234

Karnovsky NJ, Hunt GL (2002) Estimation of carbon flux to dovekies (Alle alle) in the North Water. Deep-Sea Research Part II -Topical Studies in Oceanography 49:5117-5130

Kimura DK (1977) Statistical assessment of the age-length key. Journal of the Fisheries Research Board of Canada 34:317-324

Knutsen H, Jorde PE, Andre C, Stenseth NC (2003) Fine-scaled geographical population structuring in a highly mobile marine species: the Atlantic cod. Molecular Ecology 12:385-394

Laidre KL, Stirling I, Lowry LF, Wiig Ø, Heide-Jørgensen MP, Ferguson SH (2008) Quantifying the sensitivity of Arctic marine mammals to climate-induced habitat change. Ecological Applications 18:S97-S125

160

Lampert W (1989) The adaptive significance of diel vertical migration of zooplankton. Functional Ecology 3:21-27

Lang KL, Grimes CB, Shaw RF (1994) Variations in the age and growth of yellowfin tuna larvae, Thunnus-Albacares, collected about the Mississippi River plume. Environmental Biology of Fishes 39:259-270

Lasserre F (2010) Mines et pétrole. Vers une rapide expansion de l’exploitation des ressources naturelles du sous-sol dans l’Arctique? In: Lasserre F (ed) Passages et mers arctiques. Géopolitique d’une région en mutation. Presses de l’Université du Québec, Quebec City, p. 373-410

Lasserre F, Pelletier S (2011) Polar super seaways? Maritime transport in the Arctic: an analysis of shipowners’ intentions. Journal of Transport Geography 19:1465-1473

Legendre P, Legendre L (1998) Numerical ecology, 2nd english edn. Elsevier Science BV, Amsterdam, 870 p.

Leggett WC, Deblois E (1994) Recruitment in marine fishes: is it regulated by starvation and predation in the egg and larval stages? Netherlands Journal of Sea Research 32:119-134

Leggett WC, Frank KT (2008) Paradigms in fisheries oceanography. Oceanography and Marine Biology: An Annual Review 46:331-363

Lenoir S, Beaugrand G, Lecuyer É (2011) Modelled spatial distribution of marine fish and projected modifications in the North Atlantic Ocean. Global Change Biology 17:115-129

Locarnini RA, Mishonov AV, Antonov JI, Boyer TP, Garcia HE (2006) World Ocean Atlas 2005, Volume 1: Temperature. Levitus S. (ed) NOAA Atlas NESDIS 61, US Government Printing Office, Washington, DC, 182 p.

Lønne OJ, Gulliksen B (1989) Size, age and diet of polar cod, Boreogadus saida (Lepechin 1773) in ice covered waters. Polar Biology 9:189-191

Lukovich JV, Barber DG (2005) On sea ice concentration anomaly coherence in the southern Beaufort Sea. Geophysical Research Letters 32:L10705

Macdonald RW, Paton DW, Carmack EC, Omstedt A (1995) The fresh-water budget and under-Ice spreading of Mackenzie River water in the Canadian Beaufort Sea based on salinity and 18O/16O measurements in water and ice. Journal of Geophysical Research-Oceans 100:895-919

Macdonald RW, Solomon SM, Cranston RE, Welch HE, Yunker MB, Gobeil C (1998) A sediment and organic carbon budget for the Canadian Beaufort Shelf. Marine Geology 144:255-273

161

Madsen M, Fevolden S-E, Christiansen J (2009) A simple molecular approach to distinguish between two Arctic gadoid fishes Arctogadus glacialis (Peters, 1874) and Boreogadus saida (Lepechin, 1774). Polar Biology 32:937-939

Maes J, Limburg KE, Van De Putte A, Ollevier F (2005) A spatially explicit, individual- based model to assess the role of estuarine nurseries in the early life history of North Sea herring, Clupea harengus. Fisheries Oceanography 14:17-31

Maneja R, Frommel A, Geffen A, Folkvord A, Piatkowski U, Chang M, Clemmesen C (2013) Effects of ocean acidification on the calcification of otoliths of larval Atlantic cod Gadus morhua. Marine Ecology Progress Series 477:251-258

Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology and Evolution 18:189-197

Martin GB, Thorrold SR (2005) Temperature and salinity effects on magnesium, manganese, and barium incorporation in otoliths of larval and early juvenile spot Leiostomus xanthurus. Marine Ecology Progress Series 293:223-232

Martin GB, Thorrold SR, Jones CM (2004) Temperature and salinity effects on strontium incorporation in otoliths of larval spot (Leiostomus xanthurus). Canadian Journal of Fisheries and Aquatic Sciences 61:34-42

McCulloch M, Cappo M, Aumend J, Müller W (2005) Tracing the life history of individual barramundi using laser ablation MC-ICP-MS Sr-isotopic and Sr/Ba ratios in otoliths. Marine and Freshwater Research 56:637-644

McCusker MR, Bentzen P (2010) Positive relationships between genetic diversity and abundance in fishes. Molecular Ecology 19:4852-4862

Melling H (2002) Sea ice of the northern Canadian Arctic Archipelago. Journal of Geophysical Research-Oceans 107:3181. doi: 10.1029/2001jc001102

Melnikov IA, Chernova NV (2013) Characteristics of under-ice swarming of polar cod Boreogadus saida (Gadidae) in the Central Arctic Ocean. Journal of Ichthyology 53:7-15

Michaud J, Fortier L, Rowe P, Ramseier R (1996) Feeding success and survivorship of Arctic cod larvae, Boreogadus saida, in the Northeast Water Polynya (Greenland Sea). Fisheries Oceanography 5:120-135

Middag R, de Baar HJW, Laan P, Klunder MB (2011) Fluvial and hydrothermal input of manganese into the Arctic Ocean. Geochimica et Cosmochimica Acta 75:2393- 2408

162

Miller J, Wells B, Sogard S, Grimes C, Cailliet G (2010) Introduction to proceedings of the 4th International Otolith Symposium. Environmental Biology of Fishes 89:203-207

Miller JM, Crowder LB, Moser ML (1985) Migration and utilization of estuarine nurseries by juvenile ﬁshes: an evolutionary perspective. In: Rankin MA (ed) Migration: mechanisms and adaptive signiﬁcance. Contribution to Marine Science Supplement 27, Port Aransas TX, p. 338-352

Miller T, Crowder L, Rice J, Marschall E (1988) Larval size and recruitment mechanisms in fishes: toward a conceptual framework. Canadian Journal of Fisheries and Aquatic Sciences 45:1657-1670

Møller PR, Jordan AD, Gravlund P, Steffensen JF (2002) Phylogenetic position of the cryopelagic codfish genus Arctogadus Drjagin, 1932 based on partial mitochondrial cytochrome b sequences. Polar Biology 25 342-349

Moore SE, Huntington HP (2008) Arctic marine mammals and climate change: impacts and resilience. Ecological Applications 18:S157-S165

Mousseau L, Fortier L, Legendre L (1998) Annual production of fish larvae and their prey in relation to size-fractionated primary production (Scotian Shelf, NW Atlantic). ICES Journal of Marine Science 55:44-57

Mueter FJ, Broms C, Drinkwater KF, Friedland KD, Hare JA, Hunt Jr GL, Melle W, Taylor M (2009) Ecosystem responses to recent oceanographic variability in high- latitude Northern Hemisphere ecosystems. Progress in Oceanography 81:93-110

Mundy CJ, Barber DG (2001) On the relationship between spatial patterns of sea-ice type and the mechanisms which create and maintain the North Water (NOW) Polynya. Atmosphere-Ocean 39:327-341

Nelson RJ, Bouchard C (2013) Arctic Cod (Boreogadus saida) population structure and connectivity as examined with molecular genetics. In: North Pacific Research Board Project 1125 Final Report

Nelson RJ, Bouchard C, Madsen M, Praebel K, Rondeau E, Schalburg K, Leong J, Jantzen S, Sandwith Z, Puckett S, Messmer A, Fevolden S-E, Koop B (2013) Microsatellite loci for genetic analysis of the arctic gadids Boreogadus saida and Arctogadus glacialis. Conservation Genetics Resources 5:445-448

Nesbø CL, Rueness EK, Iversen SA, Skagen DW, Jakobsen KS (2000) Phylogeography and population history of Atlantic mackerel (Scomber scombrus L.): a genealogical approach reveals genetic structuring among the Eastern Atlantic stocks. Proceedings of the Royal Society B: Biological Sciences 267:281-292

Nielsen JG, Jensen JM (1967) Revision of the Arctic cod genus Arctogadus (Pisces, Gadidae). Meddr Grønland 184:28 p

163

NPFMC (2009) Fishery management plan for fish resources of the Arctic management area, Anchorage, USA, 158 p.

O'Connell M, Wright JM (1997) Microsatellite DNA in fishes. Reviews in Fish Biology and Fisheries 7:331-363

O'Reilly PT, Canino MF, Bailey KM, Bentzen P (2004) Inverse relationship between FST and microsatellite polymorphism in the marine fish, walleye pollock (Theragra chalcogramma): implications for resolving weak population structure. Molecular Ecology 13:1799-1814

Olofsson H, Ripa J, Jonzén N (2009) Bet-hedging as an evolutionary game: the trade-off between egg size and number. Proceedings of the Royal Society B: Biological Sciences

Osuga DT, Feeney RE (1978) Antifreeze glycoproteins from Arctic fish. The Journal of Biological Chemistry 253:5338-5343

Overland JE, Stabeno PJ (2004) Is the climate of the Bering Sea warming and affecting the ecosystem? EOS, Transactions, American Geophysical Union 85:309-316

Pálsson S, Källman T, Paulsen J, Árnason E (2009) An assessment of mitochondrial variation in Arctic gadoids. Polar Biology 32:471-479

Pannella G (1971) Fish otoliths: daily growth layers and periodical patterns. Science 173:1124-1127

Parker-Stetter S, Horne J, Weingartner T (2011) Distribution of polar cod and age-0 fish in the U.S. Beaufort Sea. Polar Biology:1-15

Parmesan C (2006) Ecological and evolutionary responses to recent climate change. Annual Review of Ecology Evolution and Systematics 37:637–669

Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature 421:37-42

Perry AL, Low PJ, Ellis JR, Reynolds JD (2005) Climate change and distribution shifts in marine fishes. Science 308:1912-1915

Peterson BJ, Holmes RM, McClelland JW, Vörösmarty CJ, Lammers RB, Shiklomanov AI, Shiklomanov IA, Rahmstorf S (2002) Increasing river discharge to the Arctic Ocean. Science 298:2171

Peterson BJ, McClelland J, Curry R, Holmes RM, Walsh JE, Aagaard K (2006) Trajectory shifts in the Arctic and subarctic freshwater cycle. Science 313:1061-1066

Piatt JF, Nettleship DN (1985) Diving depths of four alcids. Auk 102:293-297

164

Pineault S, Tremblay JE, Gosselin M, Thomas H, Shadwick E (2013) The isotopic signature of particulate organic C and N in bottom ice: Key influencing factors and applications for tracing the fate of ice-algae in the Arctic Ocean. Journal of Geophysical Research-Oceans 118:287-300

Pogson GH, Taggart CT, Mesa KA, Boutilier RG (2001) Isolation by distance in the Atlantic cod, Gadus morhua, at large and small geographic scales. Evolution 55:131-146

Polyakov IV, Alexeev VA, Belchansky GI, Dmitrenko IA, Ivanov VV, Kirillov SA, Korablev AA, Steele M, Timokhov LA, Yashayaev I (2008) Arctic Ocean freshwater changes over the past 100 years and their causes. Journal of Climate 21:364-384

Polyakov IV, Beszczynska A, Carmack EC, Dmitrenko IA, Fahrbach E, Frolov IE, Gerdes Rd, Hansen E, Holfort Jr, Ivanov VV, Johnson MA, Karcher M, Kauker F, Morison J, Orvik KA, Schauer U, Simmons HL, Skagseth Ø, Sokolov VT, Steele M, Timokhov LA, Walsh D, Walsh JE (2005) One more step toward a warmer Arctic. Geophysical Research Letters 32:L17605

Ponomarenko VP (1967) Feeding of the larvae and fry of the Arctic cod (Boreogadus saida Lepechin) in the Barents and Kara seas. PINRO 10:20-27. Translated by Narragansett Marine Game Fish Research Laboratory Bureau of Sport Fisheries and Wildlife, Narragansett, RI, 1968.

Ponomarenko VP (1968) Some data on the distribution and migrations of polar cod in the seas of the Soviet Arctic. Rapports et Procès-Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 158:131-135

Ponomarenko VP (2000) Eggs, larvae, and juveniles of polar cod Boreogadus saida in the Barents, Kara, and White Seas. Journal of Ichthyology 40:165-173

Ponton D, Fortier L (1992) Vertical distribution and foraging of marine fish larvae under the ice cover of southeastern Hudson Bay. Marine Ecology Progress Series 81:215- 227

Potter IC, Beckley LE, Whitfield AK, Lenanton RCJ (1990) Comparisons between the roles played by estuaries in the life cycles of fishes in temperate Western Australia and Southern Africa. Environmental Biology of Fishes 28:143-178

Provencher JF, Gaston AJ, O’Hara PD, Gilchrist HG (2012) Seabird diet indicates changing Arctic marine communities in eastern Canada. Marine Ecology Progress Series 454:171-182

Rampal P, Weiss J, Dubois C, Campin JM (2011) IPCC climate models do not capture Arctic sea ice drift acceleration: Consequences in terms of projected sea ice thinning and decline. Journal of Geophysical Research-Oceans 116:C00D07

165

Rand K, Logerwell E (2011) The first demersal trawl survey of benthic fish and invertebrates in the Beaufort Sea since the late 1970s. Polar Biology 34:475-488

Rass T (1968) Spawning and development of polar cod. Rapports et Procès-Verbaux des Réunions du Conseil International pour l’Exploration de la Mer 158:135-137

Raventos N, Macpherson E (2005) Effect of pelagic larval growth and size-at-hatching on post-settlement survivorship in two temperate labrid fish of the genus Symphodus. Marine Ecology Progress Series 285:205-211

Réale D, McAdam AG, Boutin S, Berteaux D (2003) Genetic and plastic responses of a northern mammal to climate change. Proceedings of the Royal Society B: Biological Sciences 270:591-596

Renaud P, Berge J, Varpe Ø, Lønne O, Nahrgang J, Ottesen C, Hallanger I (2012) Is the poleward expansion by Atlantic cod and haddock threatening native polar cod Boreogadus saida. Polar Biology 35:401-412

Rico C, Rico I, Hewitt G (1996) 470 million years of conservation of microsatellite loci among fish species. Proceedings of the Royal Society B: Biological Sciences 263:549-557

Rignot E, Box JE, Burgess E, Hanna E (2008) Mass balance of the Greenland ice sheet from 1958 to 2007. Geophysical Research Letters 35:L20502

Robert D, Murphy H, Jenkins G, Fortier L (in press) Poor taxonomical knowledge of larval fish diet composition is impeding our ability to assess the existence of a ‘critical period’ driving year-class strength. ICES Journal of Marine Science. ICES Journal of Marine Science

Rose KA (2000) Why are quantitative relationships between environmental quality and fish populations so elusive? Ecological Applications 10:367-385

Rothrock DA, Zhang J (2005) Arctic Ocean sea ice volume: What explains its recent depletion? Journal of Geophysical Research-Oceans 110:C01002

Runge JA, Ingram RG (1991) Under-ice feeding and diel migration by the planktonic copepods Calanus glacialis and Pseudocalanus minutus in relation to the ice algal production cycle in southeastern Hudson Bay, Canada. Marine Biology 108:217- 225

Ruttenberg BI, Hamilton SL, Hickford MJH, Paradis GL, Sheehy MS, Standish JD, Ben- Tzvi O, Warner RR (2005) Elevated levels of trace elements in cores of otoliths and their potential for use as natural tags. Marine Ecology Progress Series 297:273-281

166

Saucier FJ, Senneville S, Prinsenberg S, Roy F, Smith G, Gachon P, Caya D, Laprise R (2004) Modelling the sea ice-ocean seasonal cycle in Hudson Bay, Foxe Basin and Hudson Strait, Canada. Climate Dynamics 23:303-326

Scherrer B (2009) Biostatistique, 2e édition, Volume 2. Gaëtan Morin, 592 p.

Schiermeier Q (2007) Polar research: The new face of the Arctic. Nature 446:133-135

Schlitzer R (2009) Ocean Data View software, http://odv.awi.de

Schoener TW (1970) Nonsynchronous spatial overlap of lizards in patchy habitats. Ecology 51:408-418

Schweiger A, Lindsay R, Zhang J, Steele M, Stern H (2011) Uncertainty in modeled arctic sea ice volume. Journal of Geophysical Research-Oceans 116:C00D06. doi: 10.1029/2011JC007084

Scott WB, Scott MG (1988) Atlantic fishes of Canada. Canadian Bulletin of Fisheries and Aquatic Sciences, Volume 219, University of Toronto Press, Toronto, 731 p.

Secor DH, Dean JM, Campana SE (1995) Recent developments in fish otolith research. University of South Carolina Press, Columbia SC, 730 p.

Secor DH, Dean JM, Laban EH (1992) Otolith removal and preparation for microstructural examination. In: Stevenson DK, Campana SE (eds) Otolith microstructure examination and analysis. Canadian Special Publication of Fisheries and Aquatic Sciences 117, p. 19-57

Secor DH, Rooker JR (2000) Is otolith strontium a useful scalar of life cycles in estuarine fishes? Fisheries Research 46:359-371

Sekerak AD (1982) Young-of-the-year cod (Boreogadus) in Lancaster Sound and Western Baffin Bay. Arctic 35:75-97

Selkoe KA, Henzler CM, Gaines SD (2008) Seascape genetics and the spatial ecology of marine populations. Fish and Fisheries 9:363-377

Shannon LJ, Cury PM, Jarre A (2000) Modelling effects of fishing in the Southern Benguela ecosystem. ICES Journal of Marine Science 57:720-722

Shaw, Turan, Wright, O’Connell, Carvalho (1999) Microsatellite DNA analysis of population structure in Atlantic herring (Clupea harengus), with direct comparison to allozyme and mtDNA RFLP analyses. Heredity 83:490-499

Shiklomanov IA, Shiklomanov AI (2003) Climatic Change and the Dynamics of River Runoff into the Arctic Ocean. Water Resources 30:593-601

167

Shimada K, Kamoshida T, Itoh M, Nishino S, Carmack E, McLaughlin F, Zimmermann S, Proshutinsky A (2006) Pacific Ocean inflow: influence on catastrophic reduction of sea ice cover in the Arctic Ocean. Geophysical Research Letters 33:L08605

Sinclair DJ, Kinsley LPJ, McCulloch MT (1998) High resolution analysis of trace elements in corals by laser ablation ICP-MS. Geochimica et Cosmochimica Acta 62:1889- 1901

Smith ADM, Brown CJ, Bulman CM, Fulton EA, Johnson P, Kaplan IC, Lozano-Montes H, Mackinson S, Marzloff M, Shannon LJ, Shin Y-J, Tam J (2011) Impacts of fishing low–trophic level species on marine ecosystems. Science 333:1147-1150

Smith SD, Muench RD, Pease CH (1990) Polynyas and leads: an overview of physical processes and environment. Journal of Geophysical Research-Oceans 95:9461-9479

Smith SJ, Campana SE (2010) Integrated stock mixture analysis for continous and categorical data, with application to genetic–otolith combinations. Canadian Journal of Fisheries and Aquatic Sciences 67:1533-1548

Sogard SM (1997) Size-selective mortality in the juvenile stage of teleost fishes: a review. Bulletin of Marine Science 60:1129-1157

Sol Y, Horness BH, Collier TK (2000) Relationship between oil exposure and reproductive parameters in fish collected following the Exxon Valdez oil spill. Marine Pollution Bulletin 40:1139-1147

Sork VL, Waits L (2010) Contributions of landscape genetics – approaches, insights, and future potential. Molecular Ecology 19:3489-3495

Standish JD, Sheehy M, Warner RR (2008) Use of otolith natal elemental signatures as natural tags to evaluate connectivity among open-coast fish populations. Marine Ecology Progress Series 356:259-268

Steele M, Ermold W (2004) Salinity trends on the Siberian shelves. Geophysical Research Letters 31:L24308

Storfer A, Murphy MA, Evans JS, Goldberg CS, Robinson S, Spear SF, Dezzani R, Delmelle E, Vierling L, Waits LP (2007) Putting the ‘landscape’ in landscape genetics. Heredity 98:128-142

Storfer A, Murphy MA, Spear SF, Holderegger R, Waits LP (2010) Landscape genetics: where are we now? Molecular Ecology 19:3496-3514

Stroeve J, Holland MM, Meier W, Scambos T, Serreze M (2007) Arctic sea ice decline: faster than forecast. Geophysical Research Letters 34:L09501

168

Stroeve J, Serreze M, Drobot S, Gearheard S, Holland M, Maslanik J, Meier W, Scambos T (2008) Arctic sea ice extent plummets in 2007. EOS, Transactions, American Geophysical Union 89:13-20

Sturgeon RE, Willie SN, Yang L, Greenberg R, Spatz RO, Chen Z, Scriver C, Clancy V, Lam JW, Thorrold S (2005) Certification of a fish otolith reference material in support of quality assurance for trace element analysis. Journal of Analytical Atomic Spectrometry 20:1067-1071

Süfke L, Piepenburg D, von Dorrien CF (1998) Body size, sex ratio and diet composition of Arctogadus glacialis (Peters, 1874) (Pisces: Gadidae) in the Northeast Water Polynya (Greenland). Polar Biology 20:357-363

Suzuki KW (2013) Identification guide to Arctic ichthyoplankton. http://www.arcticnet.ulaval.ca/pdf/identification_guide_arctic_ichthyoplankton.pdf.

Svedang H, Stal J, Sterner T, Cardinale M (2010) Consequences of subpopulation structure on fisheries management: cod (Gadus morhua) in the Kattegat and Oresund (North Sea). Reviews in Fisheries Science 18:139-150

Thanassekos S, Fortier L (2012) An Individual Based Model of Arctic cod (Boreogadus saida) early life in Arctic polynyas: I. Simulated growth in relation to hatch date in the Northeast Water (Greenland Sea) and the North Water (Baffin Bay). Journal of Marine Systems 93:25-38

Thanassekos S, Robert D, Fortier L (2012) An Individual Based Model of Arctic cod (Boreogadus saida) early life in Arctic polynyas: II. Length-dependent and growth- dependent mortality. Journal of Marine Systems 93:39-46

Thomas H, Shadwick E, Dehairs F, Lansard B, Mucci A, Navez J, Gratton Y, Prowe F, Chierici M, Fransson A, Papakyriakou TN, Sternberg E, Miller LA, Tremblay J-É, Monnin C (2011) Barium and carbon fluxes in the Canadian Arctic Archipelago. Journal of Geophysical Research-Oceans 116:C00G08

Thresher RE (1999) Elemental composition of otoliths as a stock delineator in fishes. Fisheries Research 43:165-204

Tóth G, Gáspári Z, Jurka J (2000) Microsatellites in different eukaryotic genomes: survey and analysis. Genome Research 10:967-981

Tourangeau S, Runge JA (1991) Reproduction of Calanus glacialis under ice in spring in southern Hudson Bay, Canada. Marine Biology 108:227-233

Tremblay JÉ, Bélanger S, Barber DG, Asplin M, Martin J, Darnis G, Fortier L, Gratton Y, Link H, Archambault P, Sallon A, Michel C, Williams WJ, Philippe B, Gosselin M (2011) Climate forcing multiplies biological productivity in the coastal Arctic Ocean. Geophysical Research Letters 38:L18604

169

Tynan CT, DeMaster DP (1997) Observations and predictions of Arctic climatic change potential effects on marine mammals. Arctic 50:308– 322

Umezawa A, Tsukamoto K (1991) Factors influencing otolith increment formation in Japanese eel, Anguilla japonica T. and S., elvers. Journal of Fish Biology 39:211- 223

Valerio PF, Goddard SV, Kao MH, Fletcher GL (1992) Survival of northern Atlantic cod (Gadus morhua) eggs and larvae when exposed to ice and low temperature Canadian Journal of Fisheries and Aquatic Sciences 48:2588-2595 van den Broeke M, Bamber J, Ettema J, Rignot E, Schrama E, van de Berg WJ, van Meijgaard E, Velicogna I, Wouters B (2009) Partitioning recent Greenland mass loss. Science 326:984-986

Varsamos S, Nebel C, Charmantier G (2005) Ontogeny of osmoregulation in postembryonic fish: a review. Comparative Biochemistry and Physiology Part A- Molecular and Integrative Physiology 141:401-429

Velicogna I (2009) Increasing rates of ice mass loss from the Greenland and Antarctic ice sheets revealed by GRACE. Geophysical Research Letters 36:L19503

Vigliola L, Meekan MG (2002) Size at hatching and planktonic growth determine post- settlement survivorship of a coral reef fish. Oecologia 131:89-93

Vilas R, Bouza C, Vera M, Millán A, Martínez P (2010) Variation in anonymous and EST- microsatellites suggests adaptive population divergence in turbot. Marine Ecology Progress Series 420:231-239

Wadhams P (1981) The ice cover of the Greenland and Norwegian seas. Reviews of Geophysics 19:345 -393

Walkusz W, Paulic JE, Williams WJ, Kwasniewski S, Papst MH (2011) Distribution and diet of larval and juvenile Arctic cod (Boreogadus saida) in the shallow Canadian Beaufort Sea. Journal of Marine Systems 84:78-84

Walther BD, Thorrold SR (2006) Water, not food, contributes the majority of strontium and barium deposited in the otoliths of a marine fish. Marine Ecology Progress Series 311:125-130

Walther G-R, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC, Fromentin J-M, Hoegh-Guldberg O, Bairlein F (2002) Ecological responses to recent climate change. Nature 416:389-395

Wang M, Overland JE (2009) A sea ice free summer Arctic within 30 years? Geophysical Research Letters 36:L07502

170

Ward RD (2000) Genetics in fisheries management. Hydrobiologia 420:191-201

Ward RD, Woodwark M, Skibinski DOF (1994) A comparison of genetic diversity levels in marine, freshwater, and anadromous fishes. Journal of Fish Biology 44:213-232

Was A, Gosling E, McCrann K, Mork J (2008) Evidence for population structuring of blue whiting (Micromesistius poutassou) in the Northeast Atlantic. ICES Journal of Marine Science 65:216-225

Wassmann P, Duarte CM, AgustÍ S, Sejr MK (2011) Footprints of climate change in the Arctic marine ecosystem. Global Change Biology 17:1235-1249

Wassmann P, Reigstad M (2011) Future Arctic Ocean seasonal ice zones and implications for pelagic-benthic coupling. Oceanography 24:220-231

Watts PC, Kay SM, Wolfenden D, Fox CJ, Geffen AJ, Kemp SJ, Nash RDM (2010) Temporal patterns of spatial genetic structure and effective population size in European plaice (Pleuronectes platessa) along the west coast of Scotland and in the Irish Sea. ICES Journal of Marine Science 67:607-616

Welch HE, Bergmann MA, Siferd TD, Martin KA, Curtis MF, Crawford RE, Conover RJ, Hop H (1992) Energy flow through the marine ecosystem of the Lancaster Sound region, Arctic Canada. Arctic 45:343-357

Welch HE, Crawford RE, Hop H (1993) Occurrence of Arctic cod (Boreogadus saida) schools and their vulnerability to predation in the Canadian High Arctic. Arctic 46:331-339

Weslawski JM, Ryg M, Smith TG, Oritsland NA (1994) Diet of ringed seals (Phoca hispida) in a fjord of West Svalbard. Arctic 47:109-114

Wong S, Walkusz W, Hanson M, Papst MH (2013) The influence of the Mackenzie River plume on distribution and diversity of marine larval fish assemblages on the Canadian Beaufort Shelf. Journal of Marine Systems 127:36-45

Wright JM, Bentzen P (1994) Microsatellites: genetic markers for the future. Reviews in Fish Biology and Fisheries 4:384-388

Wright S (1943) Isolation by distance. Genetics 28:114-138

Yoklavich MM, Bailey KM (1990) Hatching period, growth and survival of young walleye pollock Theragra chalcogramma as determined from otolith analysis. Marine Ecology Progress Series 64:13-23

Zakharov VF (1997) Sea ice in the climate system. World Climate Research Programme/Arctic Climate System Study, WMO/TD 782, World Meteorological Organization, Geneva, 80 p.

171

Zhang X, He J, Zhang J, Polyakov I, Gerdes R, Inoue J, Wu P (2013) Enhanced poleward moisture transport and amplified northern high-latitude wetting trend. Nature Climate Change 3:47-51

172