4.1. Comparison With Previous Studies
 Published studies on DOC distributions in first-year sea ice are rare and exclusively focused on the lowermost 4 cm of landfast ice [Smith et al., 1997; Riedel et al., 2008]. Our finding of co-variation between [DOC] and [Chl a] in the bottom layer is consistent with the previous studies. Our peak-bloom (HCP2) DOC concentrations ([DOC]s) in the bottom layer of the landfast sea ice in the western Canadian Arctic (893–1230 μmol L−1) were lower than those in the Resolute area, eastern Canadian Arctic under thin snow cover (1000–3358 μmol L−1) [Smith et al., 1997]. This could be due partly to the higher ice algal biomass observed in the Resolute area ([Chl a]: 700–2480 μg L−1) than in the present study ([Chl a]: 854–1260 μg L−1). Furthermore, our thicker ice sampling layer (10 versus 4 cm) should also substantially contributed to this difference, given that ice algae and DOC are enriched predominantly within the very thin bottom layer. Riedel et al.  reported a mean [DOC] of ∼800 μmol L−1 in the bottom 4 cm of landfast sea ice in Franklin Bay, western Canadian Arctic under thin snow cover during the vernal ice algal bloom season. Corresponding mean [DOC] from our study, 1070 μmol L−1, was higher even without taking into account the difference in the ice sampling thickness, in accordance with more abundant Chl a observed in the present study (854–1260 versus < 872 μg L−1).
 Vertical CO profiles collected during the HCPs resembled those previously documented in Franklin Bay [Xie and Gosselin, 2005]. However, in the present study, we observed a substantially higher mean [CO] in the bottom 10 cm layer (179 nmol L−1) than the earlier study (81.4 nmol L−1 in the bottom 4 cm layer). Accordingly, the mean [CO]dw and [CO]cb from the present study, 23.8 nmol L−1 and 33.6 μmol m−2, are greater than the values of 15.3 nmol L−1 and 29.4 μmol m−2 reported by Xie and Gosselin . The smaller difference in [CO]cb is due to thicker sea ice during the previous survey (mean ice thickness: 134 versus 192 cm). The cruise-mean [CO]cb in sea ice is approximately 3 times and 21% of the CO inventories observed in the open water of the Amundsen Gulf in fall 2003 and late spring 2004, respectively [Xie et al., 2009a].
4.2. Factors Controlling [DOC] and [CO] in Sea Ice
 Significant relationships between [DOC] and [Chl a] (Table 2) suggested that primary production was the main control on DOC accumulation in bottom sea ice during the HCPs, in accordance with previous studies [Smith et al., 1997; Riedel et al., 2008], but not within the upper ice matrix and during the LCPs. Other processes influencing DOC distribution in sea ice include trapping of riverine and resuspended organic matter during ice formation [Rachold et al., 2004], aeolian organic matter deposition, chemical enrichment in the quasi-liquid layer at the surface [Grannas et al., 2007], brine drainage [Giannelli et al., 2001; Amon, 2004], and microbial [Amon et al., 2001] and photochemical remineralization [Belzile et al., 2000; Xie and Gosselin, 2005]. Inclusion of resuspended or riverine organic matter into ice could be attributable to the marked near-bottom [DOC] peak observed at Stn FB07 (Figure 5f). Similar CDOM peaks have been observed in Franklin Bay [Xie and Gosselin, 2005]. Atmospheric deposition and organic matter accumulation in the quasi-liquid layer might be responsible for the frequent DOC enrichment at the surface (Figures 5a and 5e). Brine drainage most likely resulted in the rapid [DOC] drawdown in bottom ice at the onset of ice melting (Figures 4b, 5c, 5e, and 5f). To further elucidate this point, we calculated brine volume fractions (Bf) according to Cox and Weeks  and assumed that significant fluid transport occurs in first year sea ice at Bf ≥ 5% [Golden et al., 1998]. Bf in the top 10 cm increased by ca. 4 times from the pre-melting (before 13 May; 4.4% ± 2%) to the melting season (after 13 May; 16% ± 7%). Similarly, Bf within the interior ice increased by ca. 3 times from 4.4% (±3%) to 13% (±6%) over the same period of time. Bf in the bottom 10 cm ranged from 7.2% to 32% and on average was only moderately higher during the melting season (18% ± 8%) than during the preceding period (14% ± 7%). Bottom ice was thus permeable to fluids over the entire sampling period while surface and interior ice were permeable only during the period of ice melt. Brine drainage, which requires compensation by melting in the upper layers [Tison et al., 2008], should thus take place primarily during the melting season when the entire ice column was fluid-permeable.
 Notably, exchange of interior ice brine with the atmosphere and ocean can be impeded by the formation of superimposed ice as fresh ice melt comes in contact with interior ice temperature below its freezing temperature near the surface and bottom of the ice cover [Eicken, 1992; Ehn et al., 2011]. This process has been suggested to be responsible for the development of a distinct interior ice community during advanced stages of ice melt [Mundy et al., 2011]. This formation of superimposed ice contributes to the explanation of why vertical distributions of [DOC] during the melting season were more erratic than during the previous periods, particularly within the interior ice (Figures 5c and 5e).
 The significant relationships between [CO] and [Chl a] suggest that [CO] distribution in the bottom ice prior to the melting season was dictated by biological processes and chemical reactions involving algal particles and algae-derived DOM. Based on controlled laboratory experiments and field observations, Xie and Gosselin  proposed that CO photoproduction from CDOM is the principal driving force creating the distinct sea-ice [CO] vertical structure during the ice algal bloom season: lowest near the middle, highly elevated at the bottom, and moderately enriched at the surface (Figures 5b and 5d). Solar-simulated irradiations of sea ice and brine samples collected from the IPY-CFL project further confirmed the CDOM photochemical source (H. Xie, unpublished data, 2008). Moreover, Xie and Zafiriou  discovered that photodegradation of particulate organic matter (POM) in seawater is also an important CO production pathway. A later study further suggests that marine POM is more efficient than marine CDOM at CO photoproduction, particularly at visible wavelengths (G. Song and H. Xie, unpublished data, 2009). As relatively more visible radiation reaches the bottom ice due to stronger reflection of UV radiation by snow and ice [Wiscombe and Warren, 1980; Winther et al., 2004], this evidence is of great significance to the high Chl a phases when POM can be more abundant than DOM in the bottom ice [Smith et al., 1997; Riedel et al., 2008]. Additionally, prolific oxygen production due to photosynthesis by ice algae [Gleitz et al., 1995; Delille et al., 2007] promotes CO photoproduction since high oxygen concentrations accelerate organic matter photooxidation [Gao and Zepp, 1998; Xie et al., 2004]. If CDOM and POM photochemistry were the primary CO sources, then the pre-HCP [CO] in sea ice should decrease with depth as a consequence of progressive light attenuation through ice and of weak organic matter enrichment at the bottom, as was observed in the pre-HCP vertical CO profiles (Figure 5a).
 Besides photochemistry, organic matter thermal (dark) reactions also produce CO with its formation rate rising quickly above pH 8 [Zhang et al., 2008]. Although low temperatures in sea ice are not favorable for dark CO production, the highly elevated levels of organic matter and pH (up to 10) [Gleitz et al., 1995] associated with the ice algal bloom suggest that this process can be potentially important to bottom ice CO production. Moreover, Gros et al.  observed significant CO production in laboratory cultures of marine phytoplankton, particularly certain diatoms. The strong correlation between [CO] and [Chl a] found in the present study during the early stage of the ice algal accumulation (LCP1) offers circumstantial field evidence for this argument. Further investigations, particularly laboratory incubations, are needed to elucidate whether ice algae, among which diatoms often dominate in the Arctic first-year sea ice [Horner, 1985; Riedel et al., 2008; Różańska et al., 2009], are important CO producers.
 The nature of multiple formation pathways implies that CO production in sea ice commenced during freeze up around mid-October of the preceding year and sustained all the way toward the end of the melting season in summer. The strength and mechanism of CO production, however, varied in response primarily to the seasonal periodicity of solar radiation. The production is expected to be moderate and mainly CDOM photochemistry-based in fall, minimal and virtually exclusively thermal reaction-induced in winter, and maximal and largely photochemically and biologically driven in spring and summer. The continued increase in [CO] in the middle layer during the melting season (data not shown), where CO loss was weaker than at the upper and lower interfaces (see below), was indicative of continued CO production after the disappearance of the high ice algal biomass. Vanishing snow cover and thawing surface sea ice in the melting season allow more UV radiation to enter the ice, thereby facilitating CO photoproduction.
 Loss of CO in sea ice results from microbial uptake, upward release into the atmosphere, and downward transport into the water column. Microbial CO uptake, though well-documented in various marine water bodies [Conrad et al., 1982; Tolli and Taylor, 2005; Xie et al., 2005, 2009a, 2009b], has been little studied in sea ice. Limited published data suggests that this process in sea ice is relatively slow but significant with turnover times from ∼10 d in the high-biomass bottom layer to several tens of days in the much less biologically active upper layer [Xie and Gosselin, 2005]. Surface ice temperature was close to or above −10°C from 6 April onward (Figure 2c), suggesting that the ice matrix was diffusive to gas transport [Gosink et al., 1976; Loose et al., 2009, 2011]. The supersaturated state of CO (see section 3.2) hence led to an egress of this gas from ice to the atmosphere. Warmer surface ice temperatures during the melting season (range: −0.1 to −4°C; mean: −1.5°C) greatly enhanced gas diffusion and brine movement (see above), thereby boosting ice-to-air gas exchange and lowering [CO] within the surface layer (Figure 4c). Similar to DOC, the abrupt loss of CO at the start of ice melt in the bottom ice (Figure 4c) most likely resulted from brine drainage and the irregularity of [CO] vertical distributions during the melting season, particularly within the interior ice (Figures 5c and 5e), was plausibly linked to the formation of superimposed ice.
4.3. Net Production of DOC and CO
 Biological DOC production in the ice took place predominantly during the HCPs and within the lowermost 30-cm layer (see sections 3.2 and 3.3). Depth-integrated [DOC] in that layer averaged over the HCP was 127 ± 22 mmol m−2 in landfast ice and 57 ± 19 mmol m−2 in drifting ice excluding data from 6 to 11 April when [DOC] remained low (Figure 4b). Net biological DOC production was estimated as the difference between [DOC] in the HCP and that in LCP1, taking the mean depth-integrated [DOC] in the lowermost 30-cm layer for LCP1 (13 ± 5 mmol m−2) as the background [DOC]. This gave a net DOC production of 114 ± 22 mmol m−2 in landfast ice and 44 ± 19 mmol m−2 in drifting ice. The net biological DOC production averaged for both the landfast and drifting ice (75 ± 29 mmol m−2 or 1.9 ± 0.7 mmol m−2 d−1 over a period of 40 d) is comparable to the ice algal DOC release rate in the Chukchi Sea (1.6 ± 2 mmol m−2 d−1) estimated by Gosselin et al.  using deck incubations. Our value is also similar in magnitude to the annual ice algal primary production (∼80 mmol m−2 a−1) modeled by Lavoie et al.  for the Mackenzie Shelf and corresponds to 45–90% of the spring primary production of ice algae on the shelves of the Chukchi and Beaufort Seas (83–167 mmol m−2) [Gradinger, 2009]. The majority of this newly produced DOC would be released into the water column during the melting season as inferred from the precipitous [DOC] drop at the onset of bottom ice melt in mid-May (Figure 4b). Notably, the net DOC production estimated here omits any biological DOC formation occurring during the melting season and does not take into account DOC loss processes as elaborated above. Therefore, our estimate is believed to be conservative and thus underestimates the gross DOC production. In contrast, all our ice cores were taken under relatively low snow covers (≤10 cm), which reportedly favor biological DOC production over higher snow covers [Riedel et al., 2008]. This potentially leads to upward biases in our estimates. However, we note that snow depths of greater than 15 cm were not common during the present study, except near deformed ice (C. J. Mundy, unpublished data, 2008) and that maximum Chl a concentrations during HCP2 were observed under a medium snow cover (10–15 cm), reaching > 3000 μg L−1 (M. Gosselin, unpublished data, 2008).
 Net production of CO in the bottom 30-cm layer during the HCPs, assessed with an approach similar to that for DOC, was 6.5 ± 2 μmol m−2 in drifting ice, 21.5 ± 9 μmol m−2 in landfast ice, and 13.2 ± 9 μmol m−2 averaged for the two. For comparison, the parallel CO net production in the whole ice column was 7.2, 27.5, and 15.3 μmol m−2, respectively, demonstrating that CO production primarily took place within the bottom layer. Furthermore, we estimated the CO net production from the start of ice formation to the end of the HCPs as 11.3 μmol m−2 in drifting ice, 33.4 μmol m−2 in landfast ice, and 18.7 μmol m−2 averaged for the two, assuming negligible inclusion of CO from seawater during ice formation [Xie and Gosselin, 2005]. These estimates are only moderately higher than those for the HCPs due apparently to low CO production over the extended dark winter period. It is again noted that all above estimates were lower limits of the corresponding gross production rates since various loss processes were not taken into account.
4.4. Contribution of Sea Ice to Atmospheric CO
 As sea ice in our study area was supersaturated with CO, it acted as a source of CO to the atmosphere. Only the Amundsen Gulf was evaluated with regard to the CO flux to the atmosphere since the majority of the sampling stations were located in this region (Figure 1). The CO flux can be roughly estimated by multiplying a sea-ice CO transfer velocity of 0.70 m d−1 [Xie and Gosselin, 2005] by the cruise-mean surface ice [CO] (11.7 ± 8 nmol L−1) in the Amundsen Gulf. Here the sea-ice CO transfer velocity is based on the study by Fanning and Torres  indicating that ice cover reduces the 222Rn transfer velocity by ∼80%. The resultant ice-to-air flux estimate was 8.2 ± 5 μmol m−2 d−1, which is 45% of the CO flux from the open seawater to the atmosphere in the Amundsen Gulf in spring 2004 [Xie et al., 2009a]. This value translates to an annual area-integrated flux of 7.4 × 107 moles of CO in the Amundsen Gulf by applying it to the period from mid-March to mid-July and taking a mean sea ice area of 7.4 × 104 km2 (Canadian Ice Service, 2009, http://ice-glaces.ec.gc.ca/IceGraph103/page1.jsf). Therefore, the annual sea ice CO flux is > 74% of that from the open water in the same region [Xie et al., 2009a]. Note that the flux estimates made here omit the period from the start of ice formation (mid-October 2007) to mid-March 2008 over which sea ice [CO] is unknown. This underestimates the annual flux, though fluxes during wintertime are likely low due to low CO production and low ice permeability. It must be mentioned that there could be potentially large uncertainties in our CO flux estimates associated with using a constant transfer velocity. Unlike air-sea gas exchange, which has been relatively well parameterized by wind speed [e.g., Wanninkhof, 1992], there are currently no acceptable parameterizations for air-sea ice gas exchange. Gas transfer through sea ice should be dominated by diffusive processes controlled by the porosity of sea ice [Gosink et al., 1976; Loose et al., 2011]. As porosity is a function of sea ice temperature [Cox and Weeks, 1983], gas permeability is expected to change seasonally. A recent laboratory simulation study, conducted at ice surface temperatures of −4 to −12°C, observed significant gas diffusion but was unable to find a consistent relationship between gas diffusion and sea ice porosity within the porosity range encountered (0.061–0.079) [Loose et al., 2011]. Therefore, our CO flux estimates represent a first-approximation and will likely need to be refined upon the advent of a quantitative understanding of gas transfer across the air-sea ice interface.
 In addition to the direct emission of CO from ice, part of the CO released from ice into the water column is also exchanged to the atmosphere when the ice cover breaks up or through leads and polynyas. [CO]cb lacked a consistent drawdown during LCP2 (data not shown), implying that CO production during that time was no less than losses caused by melting, brine drainage, microbial uptake, and emission to air. The sea-ice CO stock in LCP2 (12.3 ± 7 μmol m−2) can, therefore, be considered the lower limit transferred to and trapped in surface seawater through melting. According to Xie et al. [2009a], microbial uptake and outgassing each accounts for half of the CO loss in the mixed layer of the Amundsen Gulf in spring. Consequently, 50% of the sea-ice CO released to the surface seawater ended up in the atmosphere, i.e., 6.2 μmol m−2 or 3.5 × 105 moles of CO, taking the mean sea ice area of 5.7 × 104 km2 during LCP2 (Canadian Ice Service, 2009, http://ice-glaces.ec.gc.ca/IceGraph103/page1.jsf). This represents only 0.5% of the direct ice-to-air flux assessed above.