Respiration and bacterial carbon dynamics in the Amundsen Gulf, western Canadian Arctic


  • Dan Nguyen,

    1. Groupe de Recherche Interuniversitaire en Limnologie et en Environnement Aquatique (GRIL), Département des Sciences Biologiques, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Roxane Maranger,

    Corresponding author
    1. Groupe de Recherche Interuniversitaire en Limnologie et en Environnement Aquatique (GRIL), Département des Sciences Biologiques, Université de Montréal, Montréal, Québec, Canada
    • Corresponding author: R. Maranger, Groupe de Recherche Interuniversitaire en Limnologie et en Environnement Aquatique (GRIL), Département des Sciences Biologiques, Université de Montréal, CP 6128, Succ. Centre-ville, Montréal, QC H3C 3J7, Canada. (

    Search for more papers by this author
  • Jean-Éric Tremblay,

    1. Québec-Océan et Takuvik, Département de Biologie, Université Laval, Québec, Québec, Canada
    Search for more papers by this author
  • Michel Gosselin

    1. Institut des Sciences de la Mer (ISMER), Université du Québec à Rimouski, Rimouski, Québec, Canada
    Search for more papers by this author


[1] Respiration rates are fundamental to understanding ecosystem C flux; however, respiration remains poorly characterized in polar oceans. The Circumpolar Flaw Lead (CFL) study provided a unique opportunity to sample the Amundsen Gulf, from November 2007 to July 2008 and follow microbial C dynamics. This study shows that bacterial production (BP) was highly variable, ranging from 0.01 to 2.14 μg C L−1 d−1 (CV = 192%), whereas the range in community respiration (CR) was more conservative from 3.8 to 44.2 μg C L−1 d−1 (CV = 55%), with measurable rates throughout the year. The spring-summer peak in BP preceded the peak in CR suggesting differential predominant control. From May until July, BP was more related to chlorophyll a concentration (r = 0.68) whereas CR was not. Given the observed high variability, BP was the main driver of bacterial growth efficiency (BGE) (r2 = 0.86). The overall average BGE was low at 4.6%, ranging from 0.20 in winter to a peak of 18.6% during the spring bloom. This study confirms that respiration is an important fate for C in the Amundsen Gulf, and our rate-based estimates of ecosystem scale CR suggests that considerably more C is respired than could be accounted for by gross primary production (GPP). One of the most plausible explanations for this observed discrepancy is that regenerated primary production is currently underestimated.

1. Introduction

[2] One of the dominant fates of organic carbon (OC) and a fundamental process in aquatic ecosystem metabolism is the transformation of OC to inorganic C (CO2) via the respiration of organisms. From a trophodynamic perspective, respiration is considered a C sink as it represents an important loss of OC from the ecosystem [Jahnke and Craven, 1995; del Giorgio and Duarte, 2002]. In most aquatic systems, microbes, primarily heterotrophic bacteria, are responsible for the majority of respiratory losses: 45%, on average, for the global ocean [Robinson, 2008] with a range of 25 to 90% of community respiration (CR) being attributed to bacterial respiration (BR) alone in polar and sub-polar regions [Sherr and Sherr, 1996; Robinson et al., 1999; Rivkin and Legendre, 2001; Kirchman et al., 2009a]. Despite this important functional role, there are considerably fewer estimates of respiration by comparison to rates of primary production (PP) in the literature [Williams and del Giorgio, 2005], particularly in polar oceans, representing an important knowledge gap in the biogeochemical cycling of C.

[3] Another important functional role of bacteria is bacterial production (BP), the conversion of dissolved organic C (DOC) into biomass for subsequent transfer to the food web [Pomeroy, 1974; Azam et al., 1983]. The relative amount of C used for biomass production versus the total amount consumed for metabolism (both production and respiration) is referred to as bacterial growth efficiency (BGE, where BGE = BP/(BP + BR) [del Giorgio and Cole, 1998]. BGE is highly variable ranging from less than 1% to 65% across marine ecosystems, with an average C conversion efficiency of approximately 15% for open oceans [del Giorgio and Cole, 2000]. As resource limitation often results in decreased BP while having only negligible effects on BR, BGE typically increases with higher productivity and inorganic nutrient availability [Robinson, 2008]. BP has also been more commonly measured in the global oceans and the total amount of C estimated to flow through bacteria is typically converted using either fixed estimates of BGE or BR estimates empirically derived from measured BP [del Giorgio and Cole, 1998, 2000; Rivkin and Legendre, 2001].

[4] One region where there is very little information available on the amount of C lost through microbial respiration or the characterization of BGE is the Arctic Ocean, where only a handful of respiration measurements from different regions exist [Sherr and Sherr, 2003; Cottrell et al., 2006; Kirchman et al., 2009a; Kritzberg et al., 2010b; R. Vaquer-Sunyer et al., Seasonal patterns in Arctic planktonic metabolism (Fram Strait–Svalbard region), submitted to Biogeosciences, 2012]. Working in the Arctic poses a great logistical challenge not only in terms of access to sites on a regular basis but also of the technical challenge of making accurate microbial rate measurements. Conditions also vary widely in polar seas on both a temporal and spatial basis: from complete darkness to full sunlight, from little autochthonous production to very high rates of primary production (PP) resulting in large pulses of labile C, a wide range of surface temperatures, changing sea ice conditions and areas with and without significant riverine OC inputs. Thus it is difficult to elucidate how microbial respiration, and the relative fate of OC, will change over space and time.

[5] The Arctic Ocean is one of the most rapidly changing ecosystems on earth [Wassmann et al., 2011]. Increased temperatures as a function of climate change have resulted in a dramatic loss of permanent sea-ice cover and the thinning of sea ice [Maslanik et al., 2007; Comiso et al., 2008]. This reduction in sea ice coverage has increased the relative amount of open water and has extended the phytoplankton growing season, resulting potentially in a greater amount of C inputs via primary production [Wang et al., 2005; Arrigo and van Dijken, 2011]. Riverine input of terrestrial OC is expected to increase due to the melting of the permafrost [Peterson et al., 2006] potentially increasing the terrestrial OC load. Sea surface temperatures during the summer are also rising [Comiso, 2003]. Increased autochthonous and allochthonous C inputs and rising water temperatures should also impact the rates of microbial respiration, and the relative conversion efficiency of bacteria (BGE) [del Giorgio and Cole, 1998]. In contrast to PP [e.g., Arrigo and van Dijken, 2011], there is no comparable baseline for respiration and bacterial C conversion efficiency in Arctic waters. C cycling throughout the Arctic in general will be greatly altered with climate change [Kirchman et al., 2009b; McGuire et al., 2009].

[6] The Circumpolar Flaw Lead (CFL) system study provided a unique opportunity to characterize the seasonal patterns of bacterial production, respiration and BGE over a nine-month period in the Amundsen Gulf in the southeastern Beaufort Sea [Barber et al., 2010, Deming and Fortier, 2011]. This study enabled us to report the seasonal patterns of measured rates of BP and CR and BGE, and elucidate some of the main controlling factors. Furthermore it allowed us to estimate how much C was fluxing through microbial communities, thus providing novel information on the seasonal respiratory demands and microbial C cycling dynamics of this particular region of the Arctic Ocean.

2. Material and Methods

2.1. Study Site

[7] From November 2007 to July 2008, sampling was carried out onboard the CCGS Amundsen in the Amundsen Gulf of the southeastern Beaufort Sea in the western Arctic Ocean (Figure 1), as part of the International Polar Year–Circumpolar Flaw Lead system study (IPY–CFL). A total of 45 stations were sampled (Figure 1), for a total of 50 discrete measurements of respiration and 270 of BP. This study includes different periods of the year: (1) the fall sea ice freeze-up, (2) the winter maximum ice cover, (3) the vernal sea ice break-up and (4) the summer open water season. For a comprehensive review of the basis of the CFL project and general physical conditions during the study, see Barber et al. [2010].

Figure 1.

Map of the area sampled in the Amundsen Gulf, eastern Beaufort Sea and M'Clure Strait. Filled circles denote stations sampled for bacterial production only, and open circles denote stations sampled for both bacterial production and respiration. FB and DB denote Franklin and Darnley Bay, respectively. Inset shows map of Canada for spatial reference.

2.2. Sample Collection and Processing

[8] Every 7 to 14 days, water samples were collected at four to six depths using a Carousel Rosette equipped with twenty-four 12 L Niskin-type bottles. Depths of interest were the surface (<12 m), nitracline (15–90 m), chlorophyll maximum when present (10–71 m), O2 minimum (120–253 m) and occasionally at bottom depths (125–967 m). BP was measured at all 6 depths per cast, and 2 to 3 were selected for respiration measurements, with 1 depth being the surface and the other most often at the nitracline and chlorophyll max, however rarely at depths below 100 m. Water was gravity-filtered directly from Niskin-type bottles using 53 μm Nitex mesh to remove large zooplankton. Water was kept in the dark in isothermal containers until processing onboard.

2.3. Physical and Chemical Variables

[9] Profiles of temperature and salinity were obtained using a SeaBird 911 + CTD mounted on the rosette, with additional probes for oxygen and chlorophyll fluorescence. Concentrations of phosphate (PO4), silicic acid (Si(OH)4) and nitrate + nitrite (NO3 + NO2) and nitrite (NO2) were determined using an Bran and Luebbe Autoanalyzer 3 with routine colorimetric methods [Grasshoff et al., 1999]. NO3 + NO2 is hereafter referred to as NO3, as NO2 concentrations were minimal throughout the entire period. Detailed analytical methods for nutrient analyses are reported elsewhere by Simpson et al. [2008] and Martin et al. [2010].

2.4. Bacterial Abundance, Cell Biovolume and Bacterial Biomass

[10] We use the term “bacteria” to refer to both archaea and bacteria, as the latter is usually the dominant prokaryotic group in Arctic waters [Kirchman et al., 2007; Garneau et al., 2008;] and in order to lighten the text. Bacterial abundance (BA, cell mL−1) was determined using epifluorescence microscopy of DAPI (4′6′-Diamidino-2-phenylindole dihydrochloride) stained cells [Porter and Feig, 1980]. Briefly, 20 mL water samples were fixed with 1 mL formaldehyde (37%) and stored in the dark at 2–4°C until processing. Samples (10–20 mL) were filtered on 0.2 μm polycarbonate black filters (25 mm, Millipore) to which 50 μL of DAPI (0.5 mg mL−1) stain was added. The stain was left on the filter for 5 min before completing the filtration. Towers were rinsed with 5 mL of ultrapure water to minimize bacterial adhesion to tower walls and distribute cells evenly on the filter. Filters were mounted on slides and stored at −20°C until photographed using a digital camera (Canon Canada) mounted on a Leica epifluorescence microscope. Counts were done by image analysis using the ImageJ software [Abramoff et al., 2004]. For each slide, 7 fields or more were photographed and analyzed, and a minimum of 200 cells was counted for each slide. From these images, cell biovolume (V, μm3) was calculated using the equation V = 4πr3/3 + (l − 2r)πr2, assuming a cylinder shaped cell with two hemispherical caps [Smith and Prairie, 2004]. Spherical cells occur when l = r. Presented values are the mean cell volume in a given sample. All size measurements were calibrated against fluorescent microspheres (ThermoFisher Scientific) to account for halo effect. Cells with a volume lower than 0.00418 μm3 were excluded (corresponding to a pore size of 0.2 μm) and large cells with a volume greater than 0.344 μm3 were not considered bacterial [Gasol et al., 1995; Straza et al., 2009] The mean C content per cell (CC, fg cell−1) was calculated using a conversion factor of 148 fg C μm−3 [Kirchman et al., 2009a].

2.5. Bacterial Production

[11] Bacterial production was measured using the 3H-leucine incorporation method [Smith and Azam, 1992]. Water samples (1.2 mL) were dispensed, in triplicate, into clean 2 mL microcentrifuge tubes pre-loaded with 50 μL 3H-leucine (115.4 Ci mmol−1, Amersham) to produce a final leucine concentration of 10 nM [Garneau et al., 2008]. Samples were incubated in the dark for approximately 4 h at in situ temperature. Leucine incorporated into cell protein was collected after precipitation with 250 μL 50% trichloroacetic acid (TCA) and centrifugation at 14000 RPM where the protein adhered to tube walls. The liquid was aspirated and samples were rinsed with 5% TCA with a second centrifugation and aspiration. Tubes were filled with 1.25 mL liquid scintillation cocktail (ScintiVerse, Fisher Scientific) and radioactivity was measured using a Tri-Carb 2900 TR Packard Liquid Scintillation Analyzer. Rates of leucine incorporation were corrected for radioactivity adsorption using TCA killed controls and converted to bacterial C production (BP) using a conversion factor of 1.5 kg C per mol 3H-leucine [Garneau et al., 2008]. Cell specific BP (BPsp) was calculated as the ratio of BP to BA.

2.6. Respiration Rate and Potential Bacterial Growth Efficiency

[12] Samples for community respiration (CR) measurements were allowed to equalize with incubation temperature in the ship's cold room prior to transfer in 500 mL gas-tight glass Erlenmeyers. CR was calculated by measuring the change in dissolved O2 consumption with optical fiber optodes (Fibox, PreSens, Germany) adapted from methods described by Kragh et al. [2008] and Marchand et al. [2009]. Briefly, an O2 sensitive sensor spot (pst3, 5 mm) is fixed to the internal wall of a 500 mL Erlenmeyer flask in which the samples are incubated. O2 concentrations are measured by linking a light emitting (600–660 nm) optical fiber to the sensor spot (from the outside of the bottle), and the sensor emits more or less luminescence depending on O2 concentrations in the sample. Measurements account for salinity and O2 concentrations are corrected for sample temperature. Unlike other methods, the use of optodes has the advantage of leaving the incubation unperturbed with almost real time O2 measurements, while being relatively space efficient and not producing toxic waste. The limit of detection of the method is 0.03% O2 sat with a relative accuracy of +/−0.4% at 20.9%O2 sat.

[13] Incubations for CR were carried out in duplicate for 5 to 10 days in a temperature controlled cold room and Erlenmeyers were kept in the dark submerged under water in isothermal containers to minimize temperature variations and prevent gas exchange. O2 concentrations were measured at 24 to 48 h intervals. For each station, a 500 mL Erlenmeyer flask containing only ultrapure water was used as control. In all incubations, O2 concentrations in the control were either stable or increased slightly, suggesting possible underestimation of CR rates.

[14] Temperature of the chamber was kept at 2–4°C during all incubations and remained stable over the course of the incubation period. The difference between incubation and in situ temperature was +4.3°C on average, with a range of ±0–6°C with 70% of the samples incubated within a narrow range of 4–5.4°C warmer than in situ (see Table S1 and Figure S1 in the auxiliary material). Only 4 samples were incubated at temperatures lower than in situ. Although it is preferable to keep incubations as close as possible to in situ temperature, logistical and material constraints did not permit this.

[15] To correct for temperature deviations from in situ, we applied a metabolic conversion factor commonly referred to as Q10 to derive more plausible in situ CR rates. We used a Q10 value of 4 based on average Q10 measurements made specifically in the polar environments and under cold-water conditions for CR [Martínez, 1996; Yager and Deming, 1999; Apple et al., 2006; Kritzberg et al., 2010b; Vaquer-Sunyer et al., 2010]. Although higher than the canonical Q10 conversion factor of 2, given that our incubation temperatures were higher than in situ over 90% of cases, the use of a Q10 of 4 provided a more conservative estimation of CR. Values were converted to C respired (μg C L−1 d−1) assuming a respiratory quotient of 0.8 [Robinson et al., 1999; Williams and del Giorgio, 2005]. Linear rates were observed throughout the incubation period and support a limited bottle effect, as suggested in a recent review [Robinson, 2008]. For quality control, any incubation with an r2 inferior to 0.60 was rejected and the mean r2 during incubation periods was 0.90.

[16] Other than a 53 μm Nitex mesh gravity screening at the rosette, no additional filtration was performed therefore whole community respiration (CR) rates were determined. Previous studies have shown that polar bacterial communities can be strongly associated to particles [Garneau et al., 2009; Kellogg et al., 2011]. Given the anticipated low respiration rates in Arctic waters and the potential importance of particle-associated bacteria, we omitted a pre-filtration step as that may have compromised our ability to measure any rate of respiration.

[17] Bacterial respiration (BR) was therefore estimated using an empirically derived equation reported by Robinson [2008], where BR = 0.45CR0.93. This relationship is based on successful fractionation experiments carried out in various locations in the global ocean. This equation allows for variations in the CR:BR ratio, and results in a lower relative proportion of BR at high CR rates. Using this equation, the mean BR:CR ratio during this study was 44% and ranged from 41 to 49%. Kirchman et al. [2009a] observed lower proportions in the Chukchi Sea with an average of BR:CR of 25% when he excluded high values where BR>CR. However, when we include these high estimates, assuming a 100% value, the average BR measured by Kirchman et al. [2009a] increases to 44% of CR. While we acknowledge that in some cases our BR may be overestimated (or underestimated), the empirical conversion model of Robinson [2008] is currently the most robust option. BGE estimates were calculated as the ratio of BP to (BP + BR). Several conversion factors were used to estimate CR, BR and therefore BGE. For this reason we consider these all to be potential estimates, but we will refer to them simply as CR, BR and BGE for the remainder of the text.

2.7. Phytoplankton Biomass

[18] Samples for phytoplankton chlorophyll a (Chl a) biomass and pheopigment concentration were filtered on 25 mm diameter Whatman GF/F filters (nominal pore size 0.7 μm). Filters were then placed in 90% acetone over 24 h at 5°C in the dark for pigment extraction. Fluorescence of the extracted pigments was measured using a Turner Designs fluorometer model 10-AU (acidification method [Parsons et al., 1984]). Chl a and pheopigment concentrations were then calculated using equations of Holm-Hansen et al. [1965].

2.8. Statistical Analysis

[19] Type 1 linear regressions and nonlinear relationships were calculated using SigmaPlot 10 (Systat Software Inc) for Windows. All linear regressions excluded samples collected at coastal sites during the spring, namely Franklin and Darnley bays. These sites were characterized by significant allochthonous inputs from rivers and coastal runoff in spring and lower water depths (<100 m), and were typically outliers in our general trends. Student's t-test for difference of means and descriptive statistics were computed with PASW 18.0 for MacOSX (SPSS Inc). Volumetric data were converted to areal data using standard trapezoidal integration over the top 80 m of the water column. This depth interval was chosen based on mean euphotic zone depth and covered most of the phytoplankton production realized in the water column [Forest et al., 2011]. When necessary, data were log-transformed to meet normality and homoscedasticity assumptions of parametric tests.

3. Results

3.1. Seasonal Patterns in Bacterial Dynamics

[20] Seasonal patterns for BP, CR, BA, and BGE are presented in Figures 2a–2d. BP rates were very low and stable from November until mid-April (Figure 2a). Rates of BP became much higher and slightly more variable from mid-May to July (Figure 2a), with a coefficient of variation (CV) from the late spring-summer of 123% as compared to 99% from winter-early spring (Table 1). Rates of CR were measured throughout the entire sampling period (Figure 2b), and were on average 210 times higher than BP from the winter-early spring period and 52 times higher in the late spring-summer. Compared to winter-early spring, average CR rates increased in late spring-summer by a factor of 1.7 but this was markedly less than BP that increased by a factor of 6.8 on average. BA also followed a similar pattern to both BP and CR with low constant abundance throughout the winter and higher more variable numbers in the spring-summer (Figure 2c). Overall biomass increased as well, but this was solely as a function of an increase in abundance, as we observed no measurable difference in cell biovolume between cells sampled in winter and in summer (Table 1).

Figure 2.

Temporal pattern of measured rates of (a) bacterial production (BP), (b) community respiration (CR), (c) bacterial abundance (BA), and (d) bacterial growth efficiency (BGE) sampled from November 2007 until July 2008.

Table 1. Seasonal Means and Ranges of Variables Measured in Amundsen Gulf for All Samples Collected at Variable Depths in the Water Columna
 November to AprilMay to JulyOverall
  • a

    Values from November to April, May to July and the overall study period are presented. Depth range was concentrated from 0 to 100 m. The asterisk denotes significant differences between periods (p ≤ 0.05) according to Student's t-test for differences of means.

  • b

    Q10 corrected values.

  • c

    Based on the equation BR = 0.45CR0.93 [Robinson, 2008].

BP (μg C L−1 d−1)0.053*0–0.2420.051820.358*0.004–2.680.44870.1460.29
CR (μg C L−1 d−1)b11.1*3.8–22.55.02118.7*6.7–44.29.02915.58.43
BR (μg C L−1 d−1)b,c4.941.84–9.642.1218.053.14–18.13.6296.743.41
BGE (%)b,c1.82*0.20–5.631.6216.60*0.44–18.65.3294.594.82
BA (×105 cells mL−1)1.63*0.77–2.480.40393.9*0.72–12.32.7372.732.20
Biovolume (μm−3 cell−1)0.1070.055–0.1560.020390.1030.064–0.1500.021370.1050.020
BB (μg C L−1)2.6*0.85–4.90.87395.5*1.1–16.93.4374.022.85
BPsp (fg C cell−1 d−1)0.39*0.02–1.10.33331.59*0.186–5.421.20291.021.08
BRsp (fg C cell−1 d−1)b,c32.89.6–60.022.02131.26.81–95.031.22931.822.0
Chl a (μg L−1)0.09*0–4.40.411161.46*0.001–10.62.31590.541.50
Pheopigments (μg L−1)0.06*0.01–0.270.041160.33*0–2.20.41540.150.26
NO3 (μmol L−1)10.30–20.16.521538.880–22.66.42819.826.51
PO43− (μmol L−1)1.150.44–1.90.39521.030–1.690.44581.090.42

[21] Converting CR into BR using the equation of Robinson [2008] enabled us to look at cell specific respiration rates and allowed us to compare how this varied over the sampling period. The relative increase in abundance of 2.4 was higher than that of BR, making for an interesting contrast in seasonal cell specific rates. Cell specific rates of bacterial respiration (BRsp) were not significantly different between the sampling periods at around 30 fg C cell−1 d−1, whereas cell-specific rates of BP (BPsp) were significantly higher in late spring-summer (Table 1). This stability in average respiration per cell between the seasons was a surprise, however the variability in the BRsp in late spring-summer was much greater with values increasing toward the end of our sampling period.

[22] The BR estimates allowed for the calculation of BGE, which was lower and more stable in winter-early spring as compared to the late spring-summer (Figure 2d). Given that BP was more variable over the whole sampling period, with a CV of 192% as compared to a more constrained annual variability in BR (CV = 55%), changes in BGE were driven by changes in BP. Indeed the overall relationship between BP and BGE, best described using a positive hyperbolic function, was very strong (Table 2 and Figure 3a; r2 = 0.86). No significant relationship was observed between BGE and BR (Figure 3b).

Table 2. Parameter Estimates and Statistics of Regression Models for Community Respiration (CR), Bacterial Production (BP) and BGE Modeled With Other Variablesa
VariableRegression Model Functionnpr
  • a

    Relationships for the overall study period and late spring-summer (May–July) are shown. Bacterial respiration, BR; bacterial production, BP; bacterial abundance, BA; bacterial growth efficiency, BGE; chlorophyll a, Chl a; temperature, T; nitrate, NO3; number of samples, n; p values, p; Pearson correlation coefficient, r; not statistically significant, NS.

CR (μg C L−1 d−1)    
BA (cells mL−1)CR = 2.5E-05BA + 9.51410.00530.51
Chl a (μg L−1)NS33  
Temperature (°C)CR = 2.9 T + 16.7420.00000.77
Nitrate (μmol L−1)LOG(CR) = −0.20LOG(NO3) + 1.18340.0015−0.53
BP (μg C L−1 d−1)    
BA (cells mL−1)LOG(BP) = 1.21LOG(BA) − 7.49580.00020.47
Chl a (μg L−1)LOG(BP) = 0.44LOG(Chl a) − 0.651590.00010.66
Temperature (°C)NS233  
Nitrate (μmol L−1)LOG(BP) = −0.35LOG(NO3) − 2.591900.0001−0.38
BGE (%)    
BP (μg C L−1d−1)BGE = 0.36 + (40.9BP/(2.52 + BP))510.00000.93
Chla (μg L−1)BGE = 2.52Chl a + 5.5330.00330.50
Temperature (°C)NS42  
Nitrate (μmol L−1)NS40  
May to July
CR (μg C L−1d−1)    
BA (cells mL−1)NS20  
Chl a (μg L−1)NS18  
Temperature (°C)CR = 2.7T + 18.0200.00000.79
Nitrate (μmol L−1)LOG(CR) = −0.22LOG(NO3) + 1.27160.0029−0.71
BP (μg C L−1d−1)    
BA (cells mL−1)LOG(BP) = 0.72LOG(BA) − 4.5230.0170.49
Chl a (μg L−1)LOG(BP) = 0.47LOG(Chl a) − 0.61490.00010.68
Temperature (°C)NS64  
Nitrate (μmol L−1)LOG(BP) = −0.56LOG(NO3) − 0.51510.0001−0.59
BGE (%)    
BP (μg C L−1d−1)BGE = −0.32 + (39.8BP/(2.2 + BP))290.00000.92
Chl a (μg L−1)NS18  
Temperature (°C)NS20  
Nitrate (μmol L−1)NS20  
Figure 3.

Overall relationship of BGE modeled (a) as a hyperbolic function of bacterial production (BP) (r2 = 0.86); modeled equation reported in Table 2 and (b) as a function of bacterial respiration (BR) (non-significant).

3.2. Factors Controlling BP and CR

[23] Despite an apparent similar seasonal trend, no significant relationship was observed between BP and CR estimates, suggesting that these rates were controlled, at least primarily, by different factors. Volumetric rates of BP could be predicted from Chl a concentrations throughout the entire sampling period by a moderately strong positive relationship explaining 43% of the variation in BP (Table 2 and Figure 4a). No significant relationship was observed between CR and Chl a (Table 2 and Figure 4b).

Figure 4.

Overall log linear relationships between chlorophyll a (Chl a) concentration and (a) bacterial production (BP) (r = 0.66; equation presented in Table 2) and (b) community respiration (CR) (non-significant). The November–April period is represented by open triangles and the May–July period by full circles.

[24] In the overall relationships, CR was significantly and linearly related to water temperature and to NO3, the latter as a log-log function, explaining 57% and 26% of the variance, respectively (Table 2). BP, however, was not significantly related to temperature but was weakly related to NO3 (r = −0.38), also in a log-log relationship. When the sampling period was restricted to late spring-summer only, relationships with CR were slightly stronger. The relationship with temperature explained 61% of the variability in CR (Table 2) but the latter should be interpreted with caution given the Q10 conversion factors used to derive estimates. In the case of NO3, relationships with BP and CR were stronger during the summer explaining 33% and 47% of the variation respectively (Figures 5a and 5b and Table 2). The summer relationship between BP and Chl a remained similar to the overall data (r = 0.68) and even when restricted to this period, there was still no observed significant relationship with CR.

Figure 5.

Negative log linear relationship between (a) bacterial production (BP) and (b) community respiration (CR) and nitrate (N-NO3) concentration (μmol L−1) for the May–July period. Details are presented in Table 2.

[25] The uncoupling of BP and CR could also be inferred by a time lag as observed in our time series and in integrated measured rate values, averaged by seasons. Rates of CR were on average 1.8 times higher in late June–July as compared with late March to early June, whereas for BP, rates were 1.4 times lower. BP responded earlier and more strongly to the accumulation of algal biomass and the concomitant labile C pulse. Increase in CR however occurred later in the slightly warmer waters in late June and July, when BA was highest. This was observed after the peak in production and when NO3 was depleted in the post-bloom period.

4. Discussion

4.1. Seasonal Patterns and Controls of CR, BP, and BGE

[26] The seasonal estimates of CR, BP, and BGE in this study represent a unique contribution to our understanding of microbial C metabolism for the Amundsen Gulf specifically and the Arctic Ocean in general. Few studies have attempted to measure respiration in the Arctic and generally considered short seasonal observation windows and only a subset of relevant bacterial activity parameters (Table 3). In this study, the concomitant estimates of CR, BP, and BGE obtained over nine consecutive months thus represent a unique contribution to our understanding of microbial C metabolism for the Amundsen Gulf specifically and the Arctic Ocean in general.

Table 3. Compilation of Volumetric Microbial Metabolic Rates and Bacterial Growth Efficiencies From Published Reports for the Arctic Oceana
   BA (105 cells mL−1)CR (μg C L−1 d−1)BR (μg C L−1 d−1)BP (μg C L−1 d−1)BGE (%) 
  • a

    Potential bacterial respiration, BR; bacterial production; BP, bacterial abundance, BA; community respiration, CR; potential bacterial growth efficiency, BGE. WC = water column. When necessary, O2 respiration rates were converted to C units using a respiratory quotient of 0.8 for consistency with the present study.

  • b

    Values based on the equation BR = 0.45CR0.93 [Robinson, 2008].

  • c

    Values presented are medians.

  • d

    Estimated using BP and BR means from this table.

Amundsen Gulf (Canadian Arctic)Whole WCAnnual2.730.72–12.315.53.8–44.26.74b1.84–18.1b0.1460–2.684.590.20–18.6This study
Chukchi Sea (Western Arctic)Photic LayerSpring–Summer7.42.7–10      6.90.6–44Kirchman et al. [2009a]
  Summer 2004  42.60––75.8     
Central ArcticUpper 240 mWinter1.50.72–2.92.90–15.91.4b0–7.0b0.0120.005–0.060.85d Sherr and Sherr [2003]
  Summer2.20.72–6.79.33–28.34.2b1.5–12b0.1300.043–0.363.0d BA in Sherr et al. [2003]
Franklin Bay (Canadian Arctic)Whole WCAutumn1.2     0.045c   Garneau et al. [2008]
  Winter3.1     0.016c    
  Spring2.0     0.015c    
  Summer6.8     0.274c    
Fram StraitPhotic LayerAll  34.10.096–28014.2b0.06–101b    Vaquer-Sunyer et al. (submitted manuscript, 2012)
  Spring  9.3 4.2b      
  Summer  36.3 15b      
  Winter  8.1 3.7b      
Chukchi Sea (Western Arctic)Photic LayerSummer 2002  14.3 6.32     Cottrell et al. [2006]
  Spring 2004  39.7 16.3      
  Summer 2004  26.1 11.1      
Kara SeaWhole WCAugust–Sept.3.52.3–4.7  9.53.3–212.240.67–6.916.410–31Meon and Amon [2004]

[27] Over the course of this study, BP was the most variable rate ranging over 2 orders of magnitude. Our average values are similar to those reported in other Arctic studies but our extremes covered a slightly wider range (Table 3). The seasonal pattern we observed was similar to the one observed in Franklin Bay [Garneau et al., 2008], adjacent to the Amundsen Gulf during another overwintering program in 2004, where low levels of BP were sustained in the winter and rates increased with the spring bloom. We also observed a strong relationship with Chl a, as reported by others for several regions of the Arctic [Garneau et al., 2008; Kirchman et al., 2009a]. We saw no statistically significant effect of water temperature on BP, possibly as a result of the limited range of temperatures observed during this study. This may also suggest a co-limitation with organic matter availability [Pomeroy and Wiebe, 2001; Kirchman et al., 2009b]. The latter could be a function of the earlier ice breakout that occurred during the CFL study [Barber et al., 2010] that resulted in a slightly earlier peak in algal production [Forest et al., 2011]. Algal blooms do not necessarily coincide with warmer surface temperatures in the Arctic [Tremblay et al., 2006], although in some years they might. Alternatively, the response of BP to additional substrates may have been so strong that any temperature effects were masked. Regardless, it is clear that the growth of Arctic bacterial communities responds almost unilaterally to substrate additions [Yager and Deming, 1999; Meon and Amon, 2004; Cuevas et al., 2011], while the effect of temperature in stimulating bacterial growth without the addition of substrates in the Arctic remains unclear [Kirchman et al., 2005, 2009b].

[28] CR followed a similar pattern to BP: there were low but measurable levels in the winter, increasing in the spring and summer. However, the following important differences were observed. First, measured CR rates and estimated BR were considerably higher than BP and a very high proportion of the total C demand. This is also evidenced by the low BGEs in the winter, within the lower range of published values, but similar to other winter values reported from the Arctic. This suggests that a far greater amount of OC is respired in the Amundsen Gulf than previously thought [Garneau et al., 2008]. The high CR rate estimates in our study are well within the range of published respiration reports for other Arctic regions (Table 3), further substantiating our values and supporting the notion that respiration is an important C sink in the polar regions.

[29] In addition, the peak in respiration followed the peak in BP. This temporal uncoupling resulted in a lack of a relationship of CR with BP and Chl a. This uncoupling between BP and CR is not surprising and has often been observed between BR and BP [del Giorgio et al., 1997; del Giorgio and Cole, 1998; Maranger et al., 2005], while the temporal shifting of peaks in time series is not commonly reported. We did however see a correspondence of CR with temperature in summer, however our relationship must be interpreted with caution, because of the conversion factors used to derive CR estimates. Nevertheless a strong effect of temperature on microbial respiration has certainly been observed in the Arctic and elsewhere [Vosjan and Olanczukneyman, 1991; Yager and Deming, 1999; Rivkin and Legendre, 2001; Apple et al., 2006; Hoppe et al., 2008; Kritzberg et al., 2010b; Vaquer-Sunyer et al., 2010]. Moreover, although some studies have observed an increase in both BP and CR with temperature, typically respiration responds more strongly [Rivkin and Legendre, 2001; Kritzberg et al., 2010a; Vaquer-Sunyer et al., 2010] which will have serious implications for the metabolic balance of the system in a warming climate [Kirchman et al., 2009b; Wohlers et al., 2009; Vaquer-Sunyer et al., submitted manuscript, 2012]. This remains an important line of future research.

[30] CR rates were by comparison less variable than rates of BP. The relatively more conservative range in CR as compared to BP could suggest a more limited effect of change in substrate availability on CR [López-Urrutia and Moran, 2007; Robinson, 2008]. However, the comparison of Arctic rate estimates in Table 3 and published relationships linking respiration to PP across systems [del Giorgio et al., 1997] suggest that bulk C availability and quality obviously plays a role in determining respiration. Therefore it is difficult to imagine that respiration was not in some way related to C availability in this time series. The negative relationship of CR with NO3 concentration most likely reflects the higher rates of C availability associated with NO3 draw down by phytoplankton and not a specific link to NO3 itself. Nonetheless, it is possible that there was a combined effect, an inter-play between post-bloom OC quality, reduced lability and inorganic nutrient availability, that resulted in higher CR, with concomitant lower BP.

[31] BR was derived empirically [Robinson, 2008] and was over the course of the time series estimated to be from 41 to 49% of CR. Therefore patterns in estimated BR mimicked patterns in CR quite closely. Given the seasonal variability of BP and the more constrained BR estimates, the pattern in BGE and its predictability were predominantly driven by changes in BP. BGE was low in the Amundsen Gulf, averaging 4.6% throughout the entire period with very low winter rates of slightly less than 1%, a maximum of 18.6%, and an average late spring-summer rate at 6.6%. Despite the conversions made in this study to derive seasonal BGE estimates, our values are similar to those reported for other Arctic regions [Kirchman et al., 2009a; Kritzberg et al., 2010b]. These lower BGE estimates support the idea that we are underestimating BR from BP when extrapolating it using an average BGE of 15% or greater in polar oceans [Robinson et al., 1999; Garneau et al., 2008]. It has been suggested that low BGEs reflect very hostile environments and that perhaps the average oligotrophic oceanic BGE hovers more around 8% rather than the more commonly referred to 15% [Carlson et al., 2007]. Results from Arctic regions certainly support this suggestion.

[32] Cell-specific rates of BP and BR give additional insight into the dynamics of microbial metabolism and their controlling factors. Although BP and BR explained an equal amount of variability in BA, BPsp rates were significantly higher during the summer period, whereas the average winter and summer BRsp rates were not significantly different. Thus, unlike BP, the peak in BR is likely a function of greater abundance rather than increased rates at the cellular level. Looking at temperature corrected cell-specific BR rates, López-Urrutia and Moran [2007] observed a strong positive relationship between Chl a and both BPsp and BGE, while no relationship was observed with BRsp, suggesting a stronger impact of substrate limitation on BPsp and its predominant role in driving BGE, similarly to our study. This could explain the observed uncoupling of BP and respiration peaks: BP would increase rapidly with substrate availability, whereas the peak in respiration would occur later when microbial abundance was high but substrate availability low during post bloom conditions. Our findings were supported independently using a dissolved inorganic carbon (DIC) mass balance approach in 2008 in the Amundsen Gulf, where a one-month lag between peak PP and peak respiration was also observed [Shadwick et al., 2011].

4.2. Implications for C Cycling in Amundsen Gulf

[33] The high respiration rates measured throughout most of the year in this study, suggest high microbial C demand in the Amundsen Gulf. Our spring-summer averages of CR integrated over the top 80 m of the water column were equal to 225 (±93) g C m−2. If we consider the GPP of 52.4 g C m−2 for this same period, estimated by dividing NO3-based new PP by the average f-ratio during this study [Forest et al., 2011], the system is in a significant C deficit. Large C deficits of 121 [Kirchman et al., 2009a] to 737 mg C m−2 d−1 [Cota et al., 1996] have also been observed in the Chukchi Sea, where measured respiration was much higher than 14C estimates of PP during the spring and summer. However, 14C estimates are known to underestimate GPP [Howarth and Michaels, 2000] and may explain some of the deficit observed in these studies. Several possibilities could explain this missing C: 1) respiration is overestimated, 2) allochthonous sources of OC are critical to supply microbial demands, 3) GPP is underestimated and/or 4) the system is heavily reliant on C recycling.

[34] There are a number of reasons why CR could be overestimated. Accurate measures of respiration are difficult to acquire and subject to methodological limitations [Robinson and Williams, 2005], particularly in cold environments. Relatively long incubations could artificially increase CR rates, in part due to bottle effects. The latter however is minimized when rates are linear over the course of the incubation [Robinson, 2008], as was the case for the measurements made in this study. One of the more important constraints of our reported CR rates is the choice of the Q10 conversion. Several studies have shown that when cold-water communities are subject to temperature deviations, the Q10 is often higher than the canonical value of 2 [Martínez, 1996; Yager and Deming, 1999; Apple et al., 2006; Kritzberg et al., 2010b; Vaquer-Sunyer et al., 2010]. This issue is believed to be especially significant when these deviations occur at low in situ temperatures [Pomeroy and Wiebe, 2001; Kirchman et al., 2009b], as in the present study. The average Q10s for CR from these studies was 4, with bacterial Q10's averaging 6, however some extreme values reported for the Arctic are >15. In our case, most of our incubations were higher than in situ temperature, meaning that the use of a canonical Q10 of 2 would provide an even higher respiration estimate. Therefore we opted for the more realistic and ultimately more conservative Q10 of 4 for our Arctic samples. However, if we double the average cold temperature Q10 of 4 to 8 in order to present a more conservative estimate of respiration, the value would be 144 gC m−2 and would still exceed PP. Nevertheless, our respiration estimates are on par with all other reports published for the Arctic (Table 3) regardless of the methods used, supporting a generally high and consistent respiratory demand in the Arctic. There is obviously a great need to better elucidate the impact of warming on the Q10 of metabolic rates in the Arctic given the rapid changes occurring in the system as a function of climate forcing [Wassmann, 2011], as warming may result in a greater than anticipated change in system C demand.

[35] Allochthonous inputs of OC could help compensate potential C deficits in Arctic Ocean. This system receives relatively high terrestrial inputs from rivers [Peterson et al., 2002; Dittmar and Kattner, 2003] and is subject to important coastal and seabed erosion [Carmack and Macdonald, 2002; O'Brien et al., 2006; McGuire et al., 2009], all of which are expected to increase C loading to the Arctic with regional climate warming. Atmospheric deposition may also be a relatively important external source of OC to the system [Macdonald et al., 2004]. With turnover times on the scale of years, terrestrial C is usually considered recalcitrant [Hansell et al., 2004], and is thought to be a minor source of OC available to organisms relative to C derived from primary production, at least seasonally for productive surface waters. Fresher material loading to the system during periods of peak run-off may provide some regional relief from C deficits, either through direct loading or from advection to sites further from shore. The extent and significance of the biological processing of terrestrial C in the Arctic Ocean remains poorly characterized [Benner et al., 2004] and merits further consideration.

[36] Another explanation for the discrepancy between CR and GPP is that GPP is underestimated. However, while incubation-based estimates of GPP are subject to bottle effects [Quay et al., 2010], this does not apply to the non-incubation based estimate from Forest et al. [2011], where GPP is back-calculated by dividing NO3-based new PP by the average f-ratio. In the aforementioned study, the background DOC pool for the upper 100 m remained high and relatively constant (∼80–95 mg C m−2) from February to August 2008, with a 133 mg C m−2 peak in July that rapidly declined to background averages [Forest et al., 2011]. The latter suggests a rapid production and consumption of ∼50 mg C m−2 on a time scale of days and represents an important pulse in C to the system. Such pulses could be fuelled by short local upwelling events which re-supply nitrate in the euphotic layer. These events are not always accounted for in nitrate drawdown inventories, and could underestimate GPP when derived from new PP and f-ratios. These short-lived events could help fill the gap of observed C deficits [Karl et al., 2003]. Furthermore pulse events may become more important in coastal areas, where deep-water upwelling as a function of climate forcing in a rapidly changing Arctic, may result in overall higher productivity [Tremblay et al., 2011]. Alternative C pathways, such as dark CO2 fixation by heterotrophic [Alonso-Sáez et al., 2010] or autotrophic prokaryotes may also be important autochthonous C inputs to the Arctic, but remain poorly constrained. These may be particularly critical in supplying C during the winter months and at depth.

[37] In a high-resolution study of the 2007–2008 DIC fluxes in Amundsen Gulf, Shadwick et al. [2011] conclude that the region is globally heterotrophic with an annual net-autotrophic surface layer with brief periods of net-heterotrophy occurring in winter. DIC mass balance models are based on the deviation of measured from expected DIC values as a function of physical properties and represent the endpoint results of integrated biological processes. DIC-based estimates however use net community production (NCP) in their models and these values will be much lower than GPP under conditions of high remineralization, where part of the DIC used by primary producers originates from in situ respiration [Mathis et al., 2009]. In fact if we apply an f-ratio more representative of systems subject to elevated rates of recycling, such as 0.2 as compared to the 0.6 estimated by Forest et al. [2011] at the peak of the productive season, estimated GPP would be around 166 g C m−2 thus closing the gap with our most conservative respiration estimates. Annual time series of f-ratios for the Arctic are rare, but values ranging from 0.05 to 0.38 have been observed in the Chukchi Sea [Cota et al., 1996]. The variability in the f-ratio is likely to influence final estimates as these have been shown to rapidly decrease to values closer to 0.1–0.4 upon nitrate depletion in the euphotic layer in the North Water Polynya [Tremblay et al., 2006]. Future investigations are required to better constrain f-ratios and their variability in the Arctic Ocean to derive improved estimates of GPP. Furthermore recycling from the food web seems to be of prime importance in the Arctic (Forest et al., 2011) and the potential for active nitrification in the photic layer [Yool et al., 2007; Christman et al., 2011] may mask the true contribution of regenerated PP to GPP.

[38] Interestingly there are multiple lines of evidence to support that this system is strongly supported by internal recycling processes. First, there is consistently low export of OC from the surface to the deep layers in this region [Forest et al., 2010; Sallon et al., 2011] where high extracellular enzyme activity has been measured on sinking particles [Kellogg et al., 2011]. Second there is sustained microbial activity observed throughout the winter [Alonso-Sáez et al., 2008; Garneau et al., 2008; this study], along with active grazer dependence on recycled materials [Sampei et al., 2009]. While we acknowledge the shortcomings of rate-based measurements, our study and others seem to support regional annual heterotrophy in the Amundsen Gulf [Garneau et al., 2008; Shadwick et al., 2011]. We suggest that the most plausible explanation to reconcile some of the observed discrepancies is that the system is heavily dependent on recycling of C and nutrients through microbes and the rest of the food web.

5. Conclusion

[39] This study presents a unique time series of microbial C dynamics in the Arctic Ocean where we were able to measure relatively constant rates of respiration throughout the winter, with slight increases in rates during the spring and summer. Although subject to methodological constraints, the measured high rates of respiration and estimated low BGEs in the Amundsen Gulf are similar to other published reports for the Arctic. Our findings confirm that a large amount of OC is being channeled toward respiration in the Amundsen Gulf, and would be indicative of a high level of recycling within the system. How warming influences the metabolic balance of the Arctic remains an exciting line of inquiry. Although climate warming may result in increased productivity [Tremblay et al., 2011], some studies suggest relatively greater respiratory losses [Wohlers et al., 2009; Kritzberg et al., 2010b; Vaquer-Sunyer et al., 2010;], altering system C demand and dynamics. Given the apparently large amounts of C respired in the Arctic, there is an urgent need to refine the parameterization of metabolic estimates as a function of warming given the impending ecosystem scale alterations anticipated in future climate.


[40] We sincerely thank the captains L. Marchand and S. Julien and crew of the CCGS Amundsen and all of the technical and administrative support team of the CFL project. Thanks to C. Robinson for her help with CR to BR conversions. We thank all of our CFL collaborators and colleagues, especially L. Delaney, G. Maltais-Landry, C. J. Mundy, C. Pedrós-Alió, and participating members from ICM for logistic and technical support. Comments from two anonymous reviewers greatly improved the manuscript. Research was supported by a CFL-IPY-Team grant (R.M, Team 7, team lead J-É Tremblay; overall project lead D. Barber, co-PIs J. Deming and G. Stern) and by an NSERC discovery grant (R.M.). D.N. was supported by a FQRNT and NSERC Ph.D. student scholarships. This is a contribution from the Groupe de recherche interuniversitaire en limnologie et en environnement aquatique (GRIL) and Québec-Océan.