Institut des Sciences de la Mer (ISMER), Université du Québec à Rimouski, Rimouski, Québec, Canada
Corresponding author: C. B. Brunelle, Institut des Sciences de la Mer (ISMER), Université du Québec à Rimouski, 310 Allée des Ursulines, Rimouski, QC G5L 3A1, Canada. (firstname.lastname@example.org)
 Phytoplankton light absorption spectra (aϕ(λ)) were measured in the Canadian Arctic (i.e., the Amundsen Gulf, Canadian Arctic Archipelago, northern Baffin Bay and the Hudson Bay system) to improve algorithms used in remote-sensing models of primary production. The absorption by algae, dominated by picophytoplankton (<5 μm), was not the major light absorption factor in the four provinces; the colored dissolved organic matter (CDOM) contributed up to 70% of total light absorption. During the fall, the low total chlorophyll a-specific aϕ*(443) (aϕ(443)/TChl a) coefficients of the Canadian High Arctic were associated with photoacclimation processes (i.e., the package effect) occurring in light-limited environments. Low light availability and high proportion of CDOM (absorbing strongly the ultraviolet) seem to allow the growth of phytoplankton with accessory pigments absorbing light at longer wavelengths. The ratio of photoprotective and photosynthetic carotenoids (PPC:PSC) was inversely proportional with the salinity and the cell size, and mostly decreases throughout the Canadian High Arctic during fall. In return, the highest TChla-specific phytoplankton light absorption coefficients at the blue peak (aϕ*(443)) were observed in the Hudson Bay system from September to October (i.e., fall) as well as in the Amundsen Gulf from May to July (i.e., spring/summer). These results will ultimately allow the accurate monitoring of phytoplankton biomass and productivity evolution that is likely to take place as a result of the fast-changing Arctic environment.
 The objectives of this study are 1) to determine the main variability sources of aϕ*(λ) and 2) characterize spatial and temporal variations of phytoplankton absorption spectra, including the light budget, in the different oceanographic provinces of the Canadian Arctic. A large data set of absorption spectra has been gathered and analyzed regarding available physical (i.e., water stratification, absorption by nonalgal material (ana) and colored dissolved organic matter (acdom)) and biological (i.e., cell sizes and algal pigments) parameters. Hypotheses are that 1) spatial variability in phytoplankton light absorption spectra is explained by differences in phytoplankton cell size and water column stratification and 2) the low of aϕ*(λ) values are related to the decrease of solar elevation and incident irradiance.
2. Materials and Methods
2.1. Data Sampling
 Sampling was conducted in the Hudson Bay system from 22 September to 13 October 2005 and in northern Baffin Bay, Canadian Arctic Archipelago and in the Amundsen Gulf from 19 October to 15 November 2007, onboard the icebreaker CCGS Amundsen as part of the Canadian research program ArcticNet (Figure 1). Those three aforementioned oceanographic provinces are identified as the Canadian High Arctic. Additional data were obtained in the Amundsen Gulf from 8 May to 6 July 2008 during the International Polar Year-Circumpolar-Flaw Lead system study (IPY-CFL). In the present study, spring/summer and fall are defined as the time period between 8 May and 6 July and between 22 September and 15 November, respectively. At each station, the depth of the euphotic zone was determined with a Secchi disk and vertical profiles of temperature (°C), salinity (psu) andin vivofluorescence were performed using a Sea-Bird 911plusCTD probe equipped with a SeaPoint chlorophyll fluorometer. Water samples were collected with 12 L Niskin-type bottles (OceanTest Equipment) at three optical depths (50%, 10% and 1% of surface irradiance), at depth of the subsurface chlorophyll fluorescence maximum (SCM) and at 60 m. In addition, surface samples (100% surface irradiance) were also collected using a clean bucket. This strategy was used for all cruises except for Hudson Bay system where samples were only taken at the surface. The vertical light attenuation coefficient (k in m−1) was calculated using the formulation of Holmes  expressed as k = 1.44/ZSD, where ZSD is the Secchi disk depth.
2.2. Pigment Composition and Taxonomy
 Samples (1.5 to 2.0 L) for pigment composition were filtered onto 25 mm Whatman GF/F glass-fiber filters (nominal porosity of 0.7 μm), stored in cryovials and frozen in liquid nitrogen until measurement by reverse-phase high-performance liquid chromatography. Pigments were extracted from the filter in ice-cold 95% methanol using a sonicator (Ultrasonic Processor XL 2010). The extracts were cleared from the filter by centrifugation and further filtered on a 0.22 μm PTFE syringe filter. A 50 μL extract was then injected in a reversed phase C8 Waters Symmetry column (150 × 4.6 mm, 3.5 μm). Gradient elution was controlled by a Thermo Separation P4000 pump. The gradient elution method used was developed by Zapata and Garrido  (mobile phases A and B1, flow rate 1 ml min−1). Pigments were detected using a Thermo Separation FL 3000 fluorescence detector in series with a photodiode array detector (Thermo Separation UV-6000). Pigments were identified using their retention time, visible spectrum and comparison with standards from DHI Water and Environment (Hørsholm, Denmark). This method allowed us to determine the concentrations of the most abundant pigments in each sample with a detection limit of 0.03 mg m−3 for Chl a. The distribution of major and taxonomically significant pigments in algal divisions/classes of Jeffrey et al.  was used to describe phytoplankton communities. As already observed in the Canadian Arctic [Tremblay et al., 2009], divinyl chlorophyll a and b (i.e., the two major pigments of Prochlorococcus) were not detected in our samples. In this study, total Chl a concentration measured by HPLC method (hereafter denoted as TChl a) was defined as the sum of Chl a, chlorophyllide a, and pheophorbide a (Table 1). Samples for the identification and enumeration of eukaryotic cells were collected at the surface and at the bottom of the euphotic zone, preserved in acidic Lugol's solution [Parsons et al., 1984] and then stored in the dark at 4°C until analysis. Cells >2 μm were identified to the lowest possible taxonomic rank and enumerated under an inverted microscope (Wild Heerbrugg) equipped with phase contrast optics [Lund et al., 1958].
Table 1. Symbols and Abbreviations
Chlorophyll a concentration measured by fluorometry
Chlorophyll a concentration for parameter i
〈Chl aFluo 〉Zeu
Integrated Chl aFluo over the euphotic zone
Total chlorophyll a concentration measured by HPLC (i.e., chlorophyll a (Chl a) + pheophorbide (Phide a) + chlorophyllide (Chlide a))
Chlorophyll b concentration (HPLC)
Total chlorophyll c concentration (HPLC) (i.e., chlorophyll c1 (Chl c1) + chlorophyll c2 (Chl c2) + chlorophyll c3 (Chl c3))
Constants in the regression between aϕ(443) and TChl a
Mixed layer depth
Z ≤ Z50%
Depth over or equal to 50% of surface irradiance (surface samples)
Z > Z50%
Depth between 1 and 50% of surface irradiance
2.3. Phytoplankton Size Structure
 Samples were filtered onto 25 mm Whatman GF/F glass-fiber filters, 5 μm Nuclepore polycarbonate membranes and 20 μm Nitex screens for the determination of the Chl abiomass of pico- (<5 μm), nano- (5–20 μm) and microphytoplankton (>20 μm). After 24 h extraction in 90% acetone at 4°C in the dark, Chl aconcentrations were determined on a 10-AU Turner Designs fluorometer (acidification method [Parsons et al., 1984]). In addition, fluorometric measurements of Chl a retained on the GF/F filters (hereafter denoted as Chl aFluo) were used to provide a global view of the spatial distribution of phytoplankton biomass. The relative biomass of each size class determined above was compared with HPLC pigment-derived class methods [Vidussi et al., 2001; Uitz et al., 2006]. The relative biomass of small phytoplankton cells (0.2–20 μm) was underestimated by the pigment composition method since zeaxanthin (a tracer of picophytoplankton) and alloxanthin (a tracer of nanophytoplankton) were not detected in our samples. Hence, the phytoplankton size structure was determined from the fluorometric method.
2.4. Particulate and Algal Absorption Measurements
 At selected stations, samples (1.5 to 2.0 L) for the measurement of the spectral absorption of particulate matter were filtered under low vacuum onto 25 mm Whatman GF/F glass-fiber filters. Blank filters were made regularly by filtering distilled water onto GF/F filters. Filters were stored in Petri dishes and kept frozen in liquid nitrogen until laboratory analysis. The transmittance-reflectance (T-R) method was used to measurein vivo light absorption by aquatic particles retained on the filter [Tassan and Ferrari, 2002]. This method is recommended in coastal regions (Case 2 waters) containing highly scattering matter. The optical density (OD, dimensionless) of these filters, before (ODp(λ)) and after (ODna(λ)) methanol extraction [Kishino et al., 1985], was measured using a dual beam Perkin-Elmer Lambda 2 spectrophotometer equipped with a 50 mm integrating sphere (Labsphere RSA-PE-20). Algal pigments from the filters were extracted using 95% methanol [Kishino et al., 1985] since (1) no unextractable pigments were detected and (2) no significant difference between pigment extraction solvents (i.e., methanol, acetone [Bricaud and Stramski, 1990] and sodium hypochlorite [Tassan and Ferrari, 2002]) has been reported in the literature. Scans were conducted at 1 nm intervals from 300 to 800 nm at a speed of 240 nm min−1. Baseline and null corrections were performed by subtracting the ODf of a fully hydrated blank filter and the averaged ODf values between 790 and 800 nm from ODp(λ) and ODna(λ) [Babin and Stramski, 2002]. The ODp and ODna coefficients were transformed into their equivalent OD value in suspension (ODsus(λ)). The ODsus(λ) is a general and empirical relationship between the optical density of particles retained on filters (ODs(λ)) and particles in suspension (ODsus(λ)) for mixed cultures and various filter types [Mitchell, 1990]. The particulate ap(λ) and nonalgal ana(λ) absorptions were obtained by equations given by Tassan and Ferrari  using ODsus(λ), filtered volume of seawater and the area of the filter. The phytoplankton light absorption coefficient (aϕ(λ), m−1) was determined using the following equation: aϕ(λ) = ap(λ) – ana(λ) (m−1). The TChl a-specific absorption coefficient of phytoplankton (aϕ*(λ), m2 mg TChl a−1) was calculated as: aϕ*(λ) = aϕ(λ)/TChl a, where TChl a is the total chlorophyll a concentration measured by HPLC (mg m−3; see Table 1). When TChl a values were not available, they were estimated from a linear regression between TChl a and Chl aFluo. For Chl aFluo concentrations ranging from 0.059 to 5.0 mg m−3, the regression equation is: TChl a = 0.825(Chl aFluo) (r2 = 0.90, n = 62) setting intercept equals to zero.
 At selected stations, water samples were filtered using 0.2 μm Anotop® filters. The filtrates were kept in HCl-cleaned bottles at −20°C in the dark. The OD of CDOM (ODcdom(λ)) was measured from 200 to 800 nm at 1 nm intervals using a Perkin-Elmer Lambda 2 spectrophotometer in a 10 cm quartz cell. Scan speed was 240 nm min−1. The CDOM absorption coefficient (acdom(λ)) was calculated as: acdom(λ) = 2.303(ODcdom(λ) – ODcdom(600))/0.1 (m−1), where ODcdom(600) is the averaged OD value between 590 and 600 nm [Fargion and Mueller, 2000].
2.6. Data Processing and Statistical Analyses
 The relative absorption contributions of phytoplankton (aϕ), nonalgal (ana) and colored dissolved organic matter (acdom) to the total nonwater absorption (at-w) were calculated at five SeaWiFS wavelengths (Figure 2). The at-w is defined as: at-w(λ) = aϕ(λ) + ana(λ) + acdom(λ) (m−1). The Brunt-Väisälä frequency (N2 in s−2), a water stratification index, was computed according to Pond and Pickard : N2 = g ρ−1 (dρ/dz), where g (m s−2) is the gravitational acceleration, ρ (kg m−3) is the seawater density [Fofonoff and Millard, 1983] and z (m) is the depth. The mixed layer depth (ZMLD) corresponds to the maximum value of N2. The depth interval for these calculations is 1 m. Interquartile ranges, extents and medians of aϕ*(λ) samples were calculated at eight SeaWiFS wavelengths (Figure 3). The outliers (i.e., values that are more than 1.5 times the interquartile range away from the interquartile range itself) that do not correspond to observed values of literature [Bricaud et al., 1995, 1998; Babin et al., 1996; Allali et al., 1997; Matsuoka et al., 2007; Roy et al., 2008] were removed. The relationships between aϕ(λ) and TChl a were calculated according to the following equation: aϕ(λ) = Aϕ(λ) (TChl a)Bϕ(λ) (m−1) [Bricaud et al., 1998], where Aϕ and Bϕ are regression's constants. TChl a and aϕ(443) regressions were base 10 log-transformed. Thepackage effect was estimated (hereafter denoted as Q*(675)) using the ratio between Chl a-specific phytoplankton light absorption coefficients (aϕ*(675)) and specific absorption coefficient of Chl a in solution (0.033 m2 mg Chl a−1) at the wavelength 675 nm [Johnsen and Sakshaug, 2007; Roy et al., 2008]. This ratio assumed that Chl a is the main light absorbing pigment at 675 nm. The ratio Q*(675) (dimensionless) ranging from 1 (unpackaged pigments) to 0.10 (strong packaging), where values above 1 indicate missing absorption terms [Bricaud et al., 2004, Roy et al., 2008]. The total concentration of pigments per cell also provided an estimation of package effect level.
 Before performing parametric tests, the normality of distribution of the data was verified by the Lilliefors test. A one-way analysis of variance (ANOVA) and a multiple comparison of means using Tukey's HSD criterion were then conducted to find any significant differences (p < 0.05) between the four oceanographic provinces. Cluster analysis, the single method using the nearest neighbor by Euclidean distance, and multiple linear regressions (y = a1x1 + a2x2 + b) were used to determine the relationships between biological and/or physical variables. If the normality of a distribution was not confirmed, a nonparametric Kruskal-Wallis test was performed instead of the ANOVA. The p value is only mentioned for comparison test. Statistical analyses were conducted using MATLAB software (version 7.1). Abbreviations are listed inTable 1. The following sections were determined by the data availability; vertical structures were only studied in the Canadian High Arctic and the seasonality variations only in the Amundsen Gulf.
3. Results and Discussion
3.1. Absorption Budget
 The light absorption budget can be used to characterize the water color. This budget is composed of the relative absorption of aϕ(λ), ana(λ) and acdom(λ) (i.e., divided by the total nonwater absorption at-w(λ)). Table 2 shows that near the phytoplankton absorption maximum, at 443 nm, the CDOM is generally the major light absorbing compound in surface waters of all regions, with aϕ(443) only contributing to 11% of total nonwater absorption in the Amundsen Gulf, 21% in the Canadian Arctic Archipelago and 28% in northern Baffin Bay. During the fall, the absorption by nonalgal (ana(443)) matter was relatively weak except in the Amundsen Gulf (21%). In return, the contribution of phytoplankton to light absorption was higher (21%) than nonalgal matter (7%) during spring/summer in this region. In the Canadian High Arctic, the major contribution of CDOM (the Amundsen Gulf = 70%, Canadian Arctic Archipelago = 72% and northern Baffin Bay = 60%) reflects the global influence of freshwater pathways from Arctic rivers through the area [Macdonald et al., 2005]. In the northern Baffin Bay, surface currents bring saltier Atlantic waters along the Greenland coast while there is a sustained southerly flow of fresher water along Ellesmere Island on the western side [Melling et al., 2001] characterized by higher CDOM concentrations. Near the Amundsen Gulf, the principal source of CDOM is the Mackenzie River outflow located 200 km to the west (418 km3 y−1) [Carmack et al., 2004; Macdonald et al., 2005; Lammers et al., 2001]. The highest CDOM proportion in percentage was observed in the Hudson Bay system (80%), a large estuarine-like inland sea that receives 760 km3 of freshwater per year from its tributaries [Déry et al., 2011]. The aϕ(443) coefficients only contributed to 13% of absorption budget in Hudson Bay (except for station AN01) and up to 65% in Hudson Strait (Table 2). This spatial variability reflects the pattern of freshwater inflows located mostly in the southern and eastern portions of the Bay while the Hudson Strait region is more influenced by inputs coming from Atlantic and Arctic [Granskog et al., 2011]. Even though the CDOM absorption (0.013–0.29 m−1) in the surface waters are in the range of European coastal waters (0.004–0.7 m−1) [Babin et al., 2003], the contributions in percentage are higher. These high proportions of CDOM are also observed at other wavelengths (412, 443, 490, 510 and 560 nm), used in the development of ocean color remote sensing algorithms, and also at all depths (Figure 2).
Table 2. Phytoplankton (aϕ), Nonalgal Material (ana), and Colored Dissolved Organic Matter (acdom) Absorption Coefficients at 443 nm, Their Relative Contributions to the Total Nonwater Absorption (at-w), and Chlorophyll a Concentration Measured by Fluorometry (Chl aFluo) in Surface Waters (Z ≤ Z50%) of the Four Oceanographic Provincesa
Year and Season
aϕ(443)/ at-w(443) (%)
ana(443)/ at-w(443) (%)
acdom(443)/ at-w(443) (%)
Chl aFluo Range (mg m−3)
Average, SD (in parentheses), and range are shown; n = number of observations.
 Spatial variability was assessed using only the data taken at the surface during the fall period in order to include the Hudson Bay. Figure 3 shows the mean specific phytoplankton absorption coefficient spectra for each province with descriptive statistics at eight SeaWiFS wavelengths. The two absorption maxima of Chl a around 440 and 675 nm are easy to identify in the aϕ*(λ) spectra. Absorption maxima of Chl b at 465 nm and TChl c at 461 nm [Hoepffner and Sathyendranath, 1991; Bricaud et al., 2004] combined with photoprotective carotenoids (PPC) absorption maximum at 460 nm and the photosynthetic carotenoids (PSC) absorption maximum at 490 nm [Bricaud et al., 2004] also contribute to the observed spectral shapes. The aϕ*(443) average was lower (p < 0.05) in the Canadian High Arctic than in the Hudson Bay system. In the Hudson Bay system the observed values are similar to the aϕ*(443) coefficients measured in 1) the Black Sea [Dmitriev et al., 2009], 2) the North Atlantic waters dominated by pico- and nanophytoplankton [Bricaud et al., 2004] and 3) the Labrador Sea [Cota et al., 2003]. The spectra measured are relatively flattened in the Canadian High Arctic. Thus, the blue-to-red ratio averages aϕ(443):aϕ(675) are also generally low (Figure 4a) [Bricaud et al., 1995; Babin et al., 2003], especially in the Amundsen Gulf (1.8, SD = 0.3) and northern Baffin Bay (1.5, SD = 0.3) (Figure 4b). Those ratios are similar to those observed in the Atlantic sector of the Southern Ocean (1.6–2.2) [Bracher and Tilzer, 2001]. The highest averaged aϕ(443):aϕ(675) ratios have been observed in the Canadian Arctic Archipelago (2.0, SD = 0.3) and the Hudson Bay system (2.5, SD = 0.7) (Figure 5b), but the spatial variability was very high in the last province (Figure 4b). Consequently, the aϕ(443) and aϕ(675) coefficients are well correlated in the Canadian High Arctic (r2 = 0.93) but not in the Hudson Bay system (r2 = 0.17).
 The spectral flattening could be caused by the package effect, closely associated with the light availability and the cell size [Bricaud and Stramski, 1990; Babin et al., 2003]. The averaged Q*(675) values, decreasing as the package effect level increases, were lower in the Amundsen Gulf (0.70) and northern Baffin Bay (0.73) than in the Canadian Arctic Archipelago (0.82) and Hudson Bay system (0.77). More precisely, the total concentration of pigments (i.e., the sum of all different pigment concentrations) per cell was higher (p ≪ 0.05) in the Canadian High Arctic (0.32–3.8 pg cell−1) than in the Hudson Bay system (0.21–1.44 pg cell−1). This indicates that the cells are potentially shade-adapted in the Canadian High Arctic during fall producing more pigments per cell, creating thepackage effect and reducing the aϕ*(443) coefficients. The proportions of accessory pigments are also a little higher in this region (Figure 6a). This could explain why the values of aϕ*(443) are directly proportional (r2 = 0.73) to Q*(675) in the Canadian High Arctic but are not (r2 = 0.11) in the Hudson Bay system. For the Hudson Bay system, the best linear regression fit for aϕ*(443) used not only the index of the package effect Q*(675) but also the concentrations of accessory pigments (AccP* = AccP/TChl a) as independent variables (r2 = 0.74). The small cell size experiencing a weak package effect level, the high proportions of small cells (<5 μm) in the Canadian Arctic Archipelago (82%) and in the Hudson Bay system (67%), compared to the Amundsen Gulf (60%) and northern Baffin Bay (49%), could also explain the relatively high Q*(675) values observed in those regions (Figure 5c). The presence of larger cells and higher package effect in the Baffin Bay and the western Arctic Ocean is probably related to higher nutrient levels availability [Hill and Cota, 2005; Tremblay et al., 2006; Klein et al., 2002; Lovejoy et al., 2002].
 Assuming that the package effect is not totally overwhelming the pigment effect, the pigment composition could also explain some differences between phytoplankton specific absorption coefficients. The averages of TChl a-specific concentrations of PSC (PSC* = PSC/TChla) in the Amundsen Gulf (0.70), Canadian Arctic Archipelago (0.45) and northern Baffin Bay (0.73) (Figure 6d) were similar to the highest values measured in the North Atlantic waters during the February to May time period with a population dominated by both nano- and picophytoplankton [Bricaud et al., 2004]. Those PSC* proportions are also higher than what was measured in the Hudson Bay system (0.37). Inversely, the averages of TChl a-specific concentrations of PPC* (PPC* = PPC/TChla) were clearly smaller (p ≪ 0.05) in the Amundsen Gulf (0.09), northern Baffin Bay (0.10) and Canadian Arctic Archipelago (0.10) than in the Hudson Bay system (0.22) (Figure 5d). The PPC* values in the Hudson Bay system were in the range of observations made in coastal waters of temperate regions (Figure 6b) [Babin et al., 2003], while those of the Canadian High Arctic were similar to values measured in arctic marine phytoplankton adapted to low incident light [Matsuoka et al., 2011]. These results highlight the fact that the high aϕ*(443) observed in the Hudson Bay system are related to higher proportions of PPC (absorbing at shorter wavelengths than PSC). Moreover, the main pigment contributions to PPC in the Canadian High Arctic were, on average, the diadinoxanthin (DD/TChl a = DD* = 0.05, SD = 0.02) with no or very low concentrations of alloxanthin (Allo) and zeaxanthin (Zea). In the Hudson Bay system, the averages of DD* (0.08, SD = 0.02), Allo* (Allo/TChl a = 0.06, SD = 0.03) and Zea* (Zea/TChl a = 0.05, SD = 0.05) were relatively higher. This suggests once again that the Canadian High Arctic's phytoplankton community could be acclimated to low light conditions (i.e., epoxidization of DT into DD under low irradiance) [Demers et al., 1991; Kashino and Kudoh, 2003; Lavaud et al., 2004; Goss and Jackob, 2010]. The xanthophyll cycle (i.e., response to sudden change in irradiance) could be more active in the Hudson Bay system, containing more xanthophyll pigments [Kashino et al., 2002] and a more adequate nutrients supply [Dubinsky and Stambler, 2009; Moreno et al., 2012].
 A cluster analysis of TChl a-specific light absorption by phytoplankton, cell size and Q*(675) values divides the Canadian High Arctic data set in two major groups characterized by: 1) a relatively high proportion of picophytoplankton, lowpackage effect and high aϕ*(443) and 2) a relatively low proportion of picophytoplankton, high package effect and low aϕ*(443) (Figures 5a and 5c). Using the most abundant and quantifiable pigments determined by HPLC, we also divided the phytoplankton community into two groups. The first group includes prasinophyceae and chlorophycea (Chl b + Prasino) and the second group includes bacillariophyta, dinophyta, prymnesiophycea and chrysophycea (TChl c + Fuco + Diadino) [Jeffrey et al., 1997; Vidussi et al., 2004]. These two groups basically represent green (i.e., containing Chl b) and red algae (i.e., containing Chl c) and are correlated with the salinity (Figure 5f). The second cluster analysis showed that the group 1 is dominated by algae containing a high proportion of Chl b* (Chl b/TChl a) + Prasino* (Prasino/TChl a) while group 2 is dominated by algae containing high proportion of Fuco* (Fuco/TChl a) + TChl c* (TChl c/TChl a) (Figure 7). The first group includes stations in northeastern Amundsen Gulf (stns. 405, 1100, 1200, 1902 and 1908), Canadian Arctic Archipelago (stns. 301, 302, 308, 309 and 310) and northwestern Baffin Bay (stns. 101 and 134). The second group includes stations in the middle and southwest of the Amundsen Gulf (stns. 407, 408, 1110, 1116 and 1216) as well as those in the northeastern Baffin Bay (stns. 108, 111 and 115). The observed spatial differences of optical properties in the Amundsen Gulf could be related to the general anticyclonic circulation pattern in the area with water entering the gulf along Banks Island and exiting at the Cape Bathurst [Lanos, 2009]. The significantly highest Fuco* average (0.50, SD = 0.1) was correlated with the presence of diatoms (microscopy counts) in northern Baffin Bay (r2 = 0.76). These results agree with previous works which showed that diatoms dominated the community along the Greenland coast while waters along the Canadian coast mostly contained flagellates [Vidussi et al., 2004] and smaller cells. In summary, our results show that there exists spatial variability of the phytoplankton optical properties between the different oceanographic provinces and that the provinces cannot be considered as spatially homogeneous.
3.3. Vertical Variability
 Due to data availability, the vertical structures during fall were studied in the Canadian High Arctic only. Previous work conducted in Arctic seas have shown that the subsurface chlorophyll maximum (SCM) depth varies generally with the vertical water column stratification and nutrient supply rather than with the light availability [Tremblay et al., 2002, 2008; Martin et al., 2010]. Our results confirm this statements as there was also no clear relationship between the SCM depths and light attenuation coefficients (k) (r2 = 0.12). Over the three Canadian High Arctic provinces, the averaged pycnocline depth (24 m, SD = 16) was shallower (44 m, SD = 15) than the averaged euphotic depth (Zeu = 1% of surface irradiance). The SCMs were located below (40%) or above (60%) the pycnocline and ranged from 5 to 62 m with a mean value of 25 m (SD = 16), which is similar to the averaged pycnocline depth. This high variability in the vertical location of the SCM is mostly explained by the relatively deep SCMs observed in the Amundsen Gulf and northern Baffin Bay, which develop during the summer near the bottom of the mixed layer as nutrients become depleted [Carmack et al., 2004, Martin et al., 2010], and the presence of a near surface SCMs in the Canadian Arctic Archipelago (i.e., located above the pycnocline and above 10 m). In the temperate oceans, Uitz et al. proposed phytoplankton biomass vertical distribution models for stratified-oceanic waters, with euphotic zone depth (Zeu) thicker than the mixed layer depth (ZMLD) (Zeu/ZMLD > 1), and for mixed waters. Our results indicated that the stratified waters model doesn't work in the Canadian High Arctic (not shown). However, the mixed waters model provides a better estimation of integrated Chl a, when the high phytoplankton biomasses (〈Chl aFluo 〉Zeu) measured in the northern Baffin Bay were included. The best regression obtained for the three provinces of the Canadian High Arctic was: 〈Chl aFluo 〉Zeu = 37.75[Chl aFluo]surf(0.893) (r2 = 0.74), which is similar to the Uitz et al.  mixed waters model.
 The presence of the pycnocline within the euphotic zone with a closely associated SCM below or above it could lead to different phytoplankton communities with different photoacclimation properties along the water column [Babin et al., 2003; Dubinsky and Stambler, 2009]. During the fall, the aϕ*(443) averaged values were higher in surface waters (above 50% of surface irradiance) than deeper waters (Figure 3), except in the Amundsen Gulf. The regressions between aϕ(443) and TChl a (Table 3 and Figures 4c and 4d) also show that specific light absorption tends to decrease with depth. Those regressions between variables were generally similar to those observed in the western and southeastern Beaufort Sea during fall [Matsuoka et al., 2007, 2009] and temperate ocean [Bricaud et al., 2004] (Figure 4c). The package effect index Q*(675) was slightly higher, but not significantly so (p > 0.05), in surface waters relative to deeper waters respectively in the Amundsen Gulf (0.70 and 0.66, p = 0.93), Canadian Arctic Archipelago (0.82 and 0.80, p = 0.19) and northern Baffin Bay (0.73 and 0.63, p = 0.17). For those provinces, the Q*(675) and aϕ*(443) are well correlated at the SCM (r2 = 0.90) while the correlation was weaker for surface waters (r2 = 0.73). The smallest Q*(675) values were observed for the highest TChl a values measured at the SCMs in the northern Baffin Bay (Figure 6c). The total concentration of pigments per cell was higher, but not significantly so (p = 0.18), in surface waters (0.32–3.8 pg cell−1) relative to deeper waters (0.82–5.4 pg cell−1) of the Canadian High Arctic. This also indicates an increase of the package effectlevel with depth. Consequently, the spectra are relatively flatter in deep waters, diminishing the blue-to-red ratios aϕ(443):aϕ(675).
Table 3. Constants for the Power Law Regression aϕ(443) = Aϕ(443)[TChl a]Bϕ(443) at 443 nm for Surface (Z ≤ Z50%) and Deeper Waters (Z > Z50%)a
Period and Depth
TChl a Range (mg m−3)
Regressions including all depths are shown where r2 is the coefficient of determination, n is the number of observations. Range, mean and SD of TChl a used for the regression are presented for each province. No regression has been calculated for the Canadian Arctic Archipelago; the range of TChl a in surface waters was too short.
 Overall, the horizontal variation (see section 3.2) of averaged aϕ*(443) appears to be stronger than the vertical variation, caused here by the cell size distribution. The dominance of small cells ([Chl aFluo]pico) is particularly strong in the Canadian Arctic Archipelago's euphotic zone, both in surface and deep layers (respectively 82% and 84%). The proportions of small cells are weaker in the surface layer and deep layers of the Amundsen Gulf (respectively 60% and 49%) and northern Baffin Bay (respectively 49% and 45%). The multiple comparison test of PSC*, Chl b* and TChl c* averages has also showed that pigment compositions at the SCM and surface waters were similar, but different per region (Figure 7); respectively the Amundsen Gulf system (0.98, 0.22 and 0.44), northern Baffin Bay (0.91, 0.12 and 0.33) and Canadian Arctic Archipelago (0.47, 0.42 and 0.25) vertical averages. The proportion of Fuco* increased throughout the middle of Baffin Bay (Figure 7c) where the highest proportion of diatoms in the Canadian High Arctic was observed (64%). The grouping of stations by the cluster analysis at the SCM was similar to the grouping obtained for surface waters, except in the northern Baffin Bay where Q*(675), aϕ*(443) and proportion of [Chl aFluo]pico decreased throughout the middle of the Bay (stns. 108 and 111). Light microscopy showed that those SCMs were dominated by diatoms, as opposed to the Amundsen Gulf where flagellates were dominant. In the Amundsen Gulf, the increase of TChl c/Chl b and Fuco* ratio toward the southwest is particularly evident at the SCM (Figure 7a). In this area, the vertical stratification index (N2) was low (0.67–2.3 × 10−3 s−2) compared to northeast part of Amundsen Gulf (N2 of 1.5–5.5 × 10−3 s−2). The stratification was also stronger in the Canadian Arctic Archipelago (N2 of 1.3–5.6 × 10−3 s−2) than in the northern Baffin Bay (N2 of 0.52–1.8 × 10−3 s−2) where the greatest biomass was observed. These results show that the physical water column structure is the main factor affecting the phytoplankton biomass along the water column, having important horizontal variations (see section 3.2).
3.4. Seasonal Variability
 Seasonal variability can only be assessed using data from the Amundsen Gulf and close by stations. It has already been noted that the specific light absorption usually decreases when the spring phytoplankton bloom occurs in temperate waters [Devred et al., 2006; Roy et al., 2008] as a result of the larger cell size and package effect increases. This decrease of aϕ*(443) coefficients has also been observed in the euphotic zone of the Amundsen Gulf, from the beginning to the end of the open water season [Matsuoka et al., 2011]. The highest averaged aϕ*(443) values observed during the spring/summer period (0.068 m2 mg TChl a−1, SD = 0.028) significantly decreased (p ≪ 0.05) at all depths during the fall period (0.042 m2 mg TChl a−1, SD = 0.011). Moreover, the blue-to-red ratio during spring/summer was highly variable (Figure 4b). This indicates that aϕ(443) is probably not only influenced by the package effect at this time but also by the pigment composition. Including all depths, Q*(675) values were (0.86, SD = 0.20) higher (p < 0.05) (i.e., lower package effect) during spring/summer than during the fall (0.73, SD = 0.19) in the Amundsen Gulf. For all depths, the high proportion of small cells (72%), according to the [Chl aFluo]picovalues, as well as the high averaged blue-to-red ratio (2.3, SD = 0.4) decreased throughout the fall period; both the proportion of [ChlaFluo]pico(62%) and the blue-to-red ratio (1.7, SD = 0.3) were low during fall. Only the blue-to-red ratios significantly decreased (p ≪ 0.05) compared to proportions of small cells (p = 0.76). In the Western Arctic Ocean,Matsuoka et al.  attributed the decrease of the TChl a-specific absorption coefficient of phytoplankton during spring/summer to a strongpackage effect overwhelming the influence of the pigment composition. Under the ice cover, the average of aϕ*(443) values were almost 3 times lower (0.017 m2/mg, SD = 0.005) in the Amundsen Gulf [Palmer et al., 2011] than what was observed in the open water of this region during our study. Thus, the increase of the cell size and package effect increase the flattening of phytoplankton specific light absorption spectra as the open water season progresses.
 As HPLC data were not available for the spring/summer period, we used the location of the absorption maxima in the red portion of the spectrum as an indicator of the presence of either green algae (Chl b maxima at 650 nm) or red algae (TChl c maxima at 635 nm) [Bricaud et al., 2004]. The use of that proxy indicated that surface waters of the Amundsen Gulf were occupied by green algae (stns. 405, 2010, 1200 and 1011) and red algae (stns. 1206, 1216 and 1806). The deeper part contained a significant presence of red algae in the center and the mouth of the gulf (stns. 405, 1011, 1206 and 1806, 8010) as well as in the western side (stn. 1216). The greatest biomasses (〈Chl aFluo〉Zeu) were observed at stations 405 and 1011, corresponding to weak water stratifications (N2 = 0.48 − 0.51 × 10−3 s−2). The water stratification was similar (p > 0.05) during the spring/summer (N2 = 0.48 − 2.7 × 10−3 s−2) and fall (N2 = 0.67 − 5.5 × 10−3 s−2). The averaged biomass (〈Chl aFluo〉Zeu) was, however, three times higher during spring/summer (35 mg m−2) than during fall (12 mg m−2) and surprisingly, the averaged depth of SCMs was almost three times deeper (34 m) than during the fall (13 m). For both seasons, the concentration of Chl a in surface waters were similar (Table 3). This shows that the spatial division in the Amundsen Gulf between the Bank's Island coast and the upwelling region along the western coastline during fall still exists during spring/summer, but the total biomass is different. As noted in section 3.2, this corresponds well to the general surface circulation pattern in that province [Lanos, 2009]. Thus, the seasonal variability of the vertical structures is important and should be considered in bio-optical models retrieving the water column-integrated phytoplankton biomass.
 Our results showed that light limitation, nutrients availability, different phytoplankton communities and cell sizes driven by physical processes are the most important sources of the observed aϕ*(443) variability in the Canadian Arctic seas. We hypothesize that during the fall and winter, phytoplankton cells adapt to their light-limited environment by producing and grouping photosynthetic pigments (i.e.,package effect). During spring and early summer, phytoplankton would have lower level of package effect. Adapted phytoplankton populations thus acclimate to the changing environment (i.e., water column stratification, light and nutrient availability) by altering their cellular content and/or pigment composition. Further work is required to understand the effects of nutrients limitation on phytoplankton light absorption spectra. Ongoing environmental changes presently observed in the Arctic (loss of sea ice cover, increased freshwater fluxes, enhanced thermal stratification, etc.) are thus expected to modify phytoplankton communities with yet to be determined effects on the global Arctic food chain. These results are important as remote sensing of phytoplankton biomass, and ultimately primary production, in Arctic region is highly dependent on the use of accurate light absorption coefficients. Our results showed that the use of regional regressions for the retrieval of Chl a could improve the quantification of the phytoplankton biomass in the Canadian Arctic.
 This study was supported by the Canadian IPY Federal program office, the ArcticNet Network of Canadian Centres of Excellence and the Natural Sciences and Engineering Research Council of Canada (NSERC). CBB received a scholarship from Institut des Sciences de la mer de Rimouski (ISMER), stipends from Québec-Océan and financial support from Aboriginal Affairs and Northern Development Canada (AANDA) for field work. The authors want to thank the scientific team, the Canadian Coast Guard officers and the crew of the CCGS Amundsen for their help and support during the cruises, S. Roy, H. Xie, Y. Gratton, S. Ben Mustapha and M. Palmer for sampling and laboratory work, E. Alou and C. J. Mundy for technical assistance. We also thank the two anonymous reviewers for their constructive comments.