Introduction
The extent and thickness of sea ice in the Arctic Ocean have been
declining in recent decades (Comiso, 2012; Stroeve et al., 2012a, b); these
declines are widely considered a consequence of climate change resulting from
the emissions of anthropogenic greenhouse gases. The average monthly extent
of sea ice in September in the Arctic Ocean decreased by about 12.4 % per
decade from 1979 to 2010 (Stroeve et al., 2012b). In September 2012, the area
of sea ice in the Arctic Ocean was less than 4 × 106 km2,
about 50 % of the average in the 1980s (Parkinson and Comiso, 2013).
Because of this decline in the extent of sea ice, the air–sea CO2 flux
in the Arctic Ocean is also thought to be dramatically changing. Currently,
the Arctic CO2 sink has been estimated as 66–199 Tg C yr-1,
with a large uncertainty (Bates and Mathis, 2009; Yasunaka et al., 2016).
This value is equivalent to 3–8 % of the net CO2 sink of the global
ocean (2.6 ± 0.5 Pg C yr-1 in the period 2006–2015; Le
Quéré et al., 2016). A recent modeling study suggests that the
CO2 sink in the Arctic Ocean is increasing (Manizza et al., 2013).
The reasons for an increasing CO2 sink in the Arctic Ocean include the
increase in the area and duration of ice-free conditions and the enhanced net
primary production they induce. However, the effect of increasing meltwater
input that accompanies the sea-ice decline should also be taken into
consideration (Yamamoto-Kawai et al., 2009; Rabe et al., 2014). The large
input of ice melt enhances stratification in the upper layer and forms a thin
surface mixed layer with a distinct halocline below. Although dilution of the
surface water with meltwater lowers the partial pressure of CO2(pCO2sea), shoaling of the surface mixed layer would
accelerate equilibration of the surface water with the overlying air. The
input of meltwater is also likely to influence carbonate chemistry by
altering the ratio of dissolved inorganic carbon (DIC) to total alkalinity
(TA), although it is still unclear whether the addition of meltwater
increases or decreases the DIC / TA ratio (Rysgaard et al., 2007; Bates et al.,
2014). Cai et al. (2010) reported unprecedented high pCO2sea
(∼ 370 µatm) in the Canada Basin in summer. They ascribed
this high pCO2sea to low net primary production and rapid
equilibration with atmospheric CO2 in the shallow mixed layer derived
from meltwater input. The low nutrient concentration in meltwater limits the
biological drawdown of pCO2sea. Else et al. (2013) found
that surface warming also contributed significantly to
pCO2sea increase in a shallow mixed layer in the Canada
Basin. Both studies concluded that an increase in meltwater lowers CO2
absorbing capacity in the Canada Basin.
A notable feature of the Canada Basin in summer is a complex water-column
structure. Because of the strong salinity gradient, there are several maxima
and minima of temperature within 150 m of the surface. This water-column
structure does not always remain stable in the rapidly changing Arctic Ocean.
Although Cai et al. (2010) and Else et al. (2013) reported relatively high
surface pCO2sea in the Canada Basin, they did not fully
explain the CO2 chemistry below the surface mixed layer there. We
studied the water-column CO2 variation in the western Arctic Ocean and
the processes that cause it.
In late summer 2013, we made shipboard observations in the Chukchi Sea and in
the Canada Basin of the western Arctic Ocean. While underway or at
hydrographic stations, or both, we measured temperature, salinity, dissolved
oxygen, and the carbonate system variables pCO2sea, DIC, and
TA. From the salinity–TA relationship, we also mapped the mixing ratio of
sea-ice meltwater, riverine outflow (RRO), and water of Pacific origin that entered
into the western Arctic Ocean through the Bering Strait. The results
demonstrate the importance of large net primary production in reducing the
pCO2sea and increasing air-to-sea CO2 flux in the
Chukchi Sea. Although this low pCO2sea water is advected
into the Canada Basin, air-to-sea CO2 flux there is blocked by a
stratified shallow surface layer that is formed by a large ice melt input.
Measurements and data
Oceanographic measurements in the Chukchi Sea and the Canada Basin were made
during cruise MR13-06 of the R/V Mirai conducted by the Japan Agency
for Marine-Earth Science and Technology from 28 August to 6 October 2013
(Nishino et al., 2015). The port of departure and arrival was Dutch Harbor,
Alaska, USA. Most of the data used in this paper are available from the JAMSTEC
Data Site for Research Cruises
(http://www.godac.jamstec.go.jp/darwin/cruise/mirai/mr13-06_leg1/e).
The rest of the data will be made available as soon as they are ready.
We made underway measurements of pCO2sea and DIC together
with temperature (T) and salinity (S) in seawater pumped continuously
from an intake located 4.5 m below surface. For the measurement of
pCO2sea, the pumped water was continuously introduced into a
shower-type equilibrator with 1.4 dm3 headspace at a rate of
4 dm3 min-1. A wavelength-scanned cavity ring-down spectrometer
(G2301, Picarro Inc., USA) was used to measure the concentrations of
CO2 in the headspace of the equilibrator and in the atmosphere sampled
from the foremast. The instrument was stable and calibrated once a day
against a set of three working standard gases of CO2 in air (Japan Fine
Products, 206.34 to 489.28 ppmv) that had been standardized on the WMO X2007
scale (Zhao and Tans, 2006). Response of the instrument to the CO2
concentration was linear. The residual of each standard gas from liner
regression was less than 0.03 ppmv. According to the manufacturer's report,
precision of CO2 measurement above 500 ppmv is 0.1 %.
For the underway measurement of DIC, a portion of pumped water was
automatically taken every 15 min and filled into glass bottles (SCHOTT
DURAN®; 300 cm3) that have been capped
with a screw-type lid. Filling, transport, and discharge of the samples were
all done through high-density PFA tubes mounted through the lid. DIC was
measured after the temperature of the sample seawater was adjusted to 20.0 ∘C in
a thermostated water bath for 1 h, using a CO2
extraction–coulometric titration system (Nippon ANS, Inc., Japan). This
system was comprised of seawater dispensing unit, a CO2 extraction unit,
and a coulometer (Model 3000, Nippon ANS, Inc., Japan). This unit dispenses
the seawater from sample bottle to a water-jacketed glass pipette of nominal
15 cm3 volume. The temperature of seawater samples in the pipette was
kept at 20.0 ∘C by a water jacket. The sample was then transferred
into a glass stripping chamber and stripped of DIC by adding 2 cm3 of
phosphoric acid (10 % v/v). The emerged CO2 was extracted into the
stream of nitrogen gas (130 cm3 min-1) and transferred to the
coulometer. The system was standardized with Certified Reference Material
(CRM; batch no. 113) supplied by A. G. Dickson (Scripps Institution of
Oceanography); underway measurement of DIC was interrupted for several hours
once a day for calibration. The precision of measurement determined by
repeatability of CRMs was ±2.2 µmol kg-1. The values
of TA in the surface were calculated from measured pCO2sea,
DIC, temperature, and salinity using dissociation constants of carbonic acid
given by Lueker et al. (2000) and a macro package of CO2SYS program for
Microsoft Excel (Pierrot et al., 2006). Uncertainty in surface TA was
estimated as ±3 µmol kg-1 by taking the uncertainties
of DIC and pCO2sea into account.
At hydrographic stations, seawater-column profiles of temperature, salinity,
and dissolved oxygen were obtained with a CTD (SBE 9 plus, Sea-Bird
Scientific, USA) rosette sampler equipped with sensors for dissolved oxygen
(SBE 43, Sea-Bird Scientific, USA) and Niskin bottles (12 dm3). In
addition to the CTD casts, some expendable CTDs (XCTD) were used to obtain
water-column profiles of temperature and salinity. Discrete water samples
were taken in Niskin bottles at depths of 0, 5, 10, 20, 30, 40, 50, 75, 100,
125, 150, and 200 m along with CTD measurements. Samples were also collected
at chlorophyll maximum layers that ranged from 12 to 92 m. Measurements of
dissolved oxygen were made by the Winkler titration method following
Dickson (1994) and used to correct for the bias of the data from the oxygen
sensor SBE 43. Apparent oxygen utilization (AOU), i.e., the difference
between the measured concentration of dissolved oxygen and its saturation
concentration under the same potential temperature and salinity conditions,
was calculated using the oxygen solubility constant given by Garcia and
Gordon (1992). Water samples for chlorophyll a were vacuum-filtered
(< 0.02 MPa) through a 25 mm diameter Whatman grade GF/F filter,
and fluorescence was measured for each sample with a fluorometer (10-AU-005,
Turner Designs, USA). The fluorometer was calibrated against pure
chlorophyll a (Sigma-Aldrich, USA). The concentration of nutrients
(nitrate, nitrite, silicate, phosphate, and ammonia) was determined using a
continuous flow analyzer (QuAAtro 2-HR, BLTEC, Japan) according to the
GO-SHIP Repeat Hydrography Manual (Hydes et al., 2010). Reference materials
for nutrients of seawater (Aoyama and Hydes, 2010) were used.
Subsamples for DIC and TA measurements in the discrete water samples were
drawn into borosilicate glass bottles (300 cm3 for DIC and
125 cm3 for TA) using the protocol of Dickson et al. (2007).
Measurements of DIC at depths were also made with the extraction–coulometric
system (Nippon Ans., Japan). Saturated solution of HgCl2 (0.1 cm3)
was added to each of the samples to inhibit any biological activity.
Measurements of TA were made with a spectrophotometric system (Nippon Ans.,
Japan) based on a single-point pH determination using bromocresol green as an
indicator dye (Yao and Byrne, 1998). Seawater sampled in the glass bottles
was transferred to an optical cell via dispensing unit. The length and volume
of the cell were 8 cm and 13 cm3, respectively. Temperature of the
cell was kept at 25.0 ∘C. The sample was mixed for 8.5 min after the
injecting the indicator dye solution and hydrochlonic acid (0.05 M). TA was
calculated from absorbance ratio at 444 and 616 nm (A616/A444).
Replicate measurements yielded an average and standard deviation of
differences of 1.0 ± 1.1 µmol kg-1 for DIC and
0.9 ± 0.8 µmol kg-1 for TA. Values of
pCO2sea in discrete water samples were calculated from DIC,
TA, temperature, and salinity using dissociation constants of carbonic acid
given by Lueker et al. (2000).
Satellite-derived net primary production (NPP) was used to evaluate
biological production in a broad area. NPP is estimated from empirical
equations using chlorophyll concentration, sea surface temperature (SST),
photosynthetically active radiation, and length of the daytime as variables
in the Vertical Generalized Production Model (Behrenfeld and Falkowski,
1997). The data of NPP were downloaded from the website of Oregon State
University
(http://www.science.oregonstate.edu/ocean.productivity/standard.product.php).
Monthly averaged wind speed data from the Japanese 55-year reanalysis
(JRA-55) product were used to calculate air–sea CO2 flux. JRA-55 has a
spatial resolution of 1.25∘ longitude by 1.25∘ latitude
(Kobayashi et al., 2015). We applied data of reanalysis rather than those of
an anemometer mounted on the ship after considering the representativeness of
data.
Calculations
CO<inline-formula><mml:math id="M122" display="inline"><mml:msub><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msub></mml:math></inline-formula> flux and half-life of <inline-formula><mml:math id="M123" display="inline"><mml:mrow><mml:mi mathvariant="normal">Δ</mml:mi><mml:mi>p</mml:mi></mml:mrow></mml:math></inline-formula>CO<inline-formula><mml:math id="M124" display="inline"><mml:msub><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msub></mml:math></inline-formula>
Surface pCO2sea was calculated from the mole fraction of
CO2 in the air by taking the water vapor pressure and atmospheric
pressure into account. The temperature and salinity of the pumped water at
the intake were continuously measured with sensors SBE 38 and SBE 45,
respectively (Sea-Bird Electronics, USA). PT100 thermometer was equipped with
the equilibrator. An increase in temperature between in situ seawater
(Tin) and the equilibrator (Teq) was typically about 0.2 ∘C. Equation (1)
in Takahashi et al. (2009) was applied to convert
pCO2(Teq) to pCO2(Tin). Uncertainty in the value of
pCO2sea measured using the same type of equilibrator has
been estimated to be ±3 µatm (Midorikawa et al., 2006).
Air–sea CO2 flux (FCO2) was calculated using the following
equation:
ΔpCO2=pCO2sea-pCO2air,FCO2=kαΔpCO2,
where α denotes the solubility of CO2 in seawater (Weiss, 1974).
We used a gas-transfer piston velocity k given by Wanninkhof (2014):
k=0.251×U102×(Sc/660)-0.5.
U10 denotes wind speed at 10 m a.s.l. (above sea
level). Sc indicates the Schmidt number (Wanninkhof, 2014).
Half-life of CO2 gas exchange (τ1/2) was calculated for
quantitative discussion about temporal variation in ΔpCO2. At
first, initial conditions of temperature, salinity, DIC, and TA were set.
Initial pCO2 (pCO20) was calculated from these values using
dissociation constants of carbonic acid given by Lueker et
al. (2000).
pCO20=f(T,S,DIC,TA)
All parameters but DIC were fixed during the calculation; i.e., temperature,
salinity, and TA were assumed unchanged. FCO2 in each time step was
calculated using Eqs. (2), (3), and (4). Increase in DIC was calculated
from FCO2. The time step was set to 1 day.
ΔDIC=FCO2MLD⋅ρ(T,S)DICt+1=DICt+ΔDIC
Here, MLD and T,S are fixed mixed layer depth (m) and density of seawater
in surface mixed layer (kg m-3), respectively. After each time step,
ΔpCO2t and pCO2t were calculated from DIC. The time
required to reduce ΔpCO2t to half of ΔpCO20
was defined as τ1/2.
Water properties around the temperature minimum layer in the
Chukchi Sea and the Canada Basin in samples collected during cruise MR13-06
from 3 September to 1 October 2013. Potential temperature
versus (a) salinity and (b) total alkalinity (TA).
![]()
Determination of freshwater fraction
In the western Arctic Ocean, the water at the temperature minimum layer
(∼ 150 dbar) is known to originate in the North Pacific and be advected
into the Arctic through the Bering Strait (POW: Pacific-origin water; Shimada
et al., 2005), and the water in the layer above the temperature minimum is
thought to be a mixture of this POW with sea-ice melt (SIM) and RRO.
To determine the fractions (f) of the three different source waters –
POW, SIM, and RRO – in the upper-layer
waters of the Chukchi Sea and the Canada Basin, we used the relationship
between salinity and TA and the following mass balance equations.
TA=fPOW⋅TAPOW+fSIM⋅TASIM+fRRO⋅TARROS=fPOW⋅SPOW+fSIM⋅SSIM+fRRO⋅SRRO1=fPOW+fSIM+fRRO
We chose the data of TA and S from 38 sampling layers/locations in the
temperature minimum layer, where T < -1.5 ∘C in the
Chukchi Sea and the Canada Basin during the cruise, and defined their means
(2264.2 ± 12.6 µmol kg-1 and 32.89 ± 0.22) as the
values of TAPOW and SPOW, respectively (Fig. 1).
TA versus salinity in the western Arctic Ocean above the
temperature minimum layer from 4 to 11 September 2013 color-coded for (a) the
fraction of sea-ice melt (fSIM) and (b) the fraction of riverine outflow
(fRRO). Circles and diamonds denote surface water and CTD samples,
respectively. Star denotes the salinity and TA of Pacific-origin water (POW:
S= 32.89 and TA = 2264.2 µmol kg-1). Broken lines
extend to the two endmembers, sea-ice melt (SIM: S= 5 and TA = 349 µmol kg-1) and riverine
output (RRO: S= 0 and TA = 793 µmol kg-1).
![]()
There are several studies of TA of riverine outflow in the Arctic. Cooper et
al. (2008) directly measured TA in six major rivers in the Arctic: they
concluded that flow-weighted average of TA of these six rivers was
1048 µmol kg-1. Yamamoto-Kawai et al. (2009) made linear
regression analysis of salinity and TA and reported that the intercept (S= 0) was 793 µmol kg-1 for the whole Canada Basin.
Data of TA and salinity taken during our cruise indicate that the upper limit
of distribution in salinity–TA plots (Fig. 2) is consistent with the line
extended to this intercept deduced by Yamamoto-Kawai et al. (2009);
consequently, we regarded this value as TARRO. In consideration of the
spatial and temporal fluctuation of riverine TA, we assumed that the
uncertainty of TARRO is ±100 µmol kg-1
(Yamamoto-Kawai et al., 2005). Accordingly, the estimated errors of fSIM
and fRRO are as large as ±0.02.
Conversely, values of SSIM and TASIM reported so far fall within a
relatively narrow range. We applied SSIM= 5 and TASIM= 349 µmol kg-1 following Fransson et al. (2009).
Differences in fSIM and fRRO are not larger than ±0.008 when
applying other values suggested by Anderson et al. (2004) (SSIM= 4,
TASIM= 263 µmol kg-1). Cumulative error in
fSIM and fRRO associated with the selection of the endmember
salinity and TA are within ±0.03. As shown in Fig. 2, S–TA plots
for the Chukchi Sea and Canada Basin fall among the three S–TA endmembers
of POW, SIM, and RRO. Surface water in the Chukchi Sea and Canada Basin
consists mainly of POW but includes sizable fSIM up to 0.16 and
fRRO up to 0.18.
Surface water properties along the track of cruise MR13-06 from 4
to 11 September 2013. (a) Sea surface temperature (SST); (b) sea surface
salinity (SSS); (c) BCW (Barrow coastal water), CSW (Chukchi Sea water), and
CBW (Canada Basin water) water types according to SST and SSS. The numbers
over
triangles indicate distance sailed from Dutch harbor, Alaska, USA
(km). (d) fSIM. (e) fRRO.
(f) pCO2sea. (g) nDIC32= DIC / S⋅32, where DIC normalized is to S= 32. (h) nTA32= TA / S⋅32, where
TA is normalized to S= 32. Dotted lines indicate 50 and
200 m isodepths.
![]()
Results and discussion
Variations in temperature and salinity in the surface layer
Cruise MR13-06 occupied a wide area of the Chukchi Sea and the Canada Basin.
General variations in surface pCO2sea in these regions have
already been well investigated (Bates, 2006; Cai et al., 2010). The results
from our cruise were not much different from these reports. Therefore, we
highlighted below the differences of water mass characteristic and CO2
dynamics in these regions. In the period from 4 to 11 September 2013,
SST, sea surface salinity (SSS), pCO2sea, and DIC in
surface water were highly variable in the western Arctic Ocean (Fig. 3),
particularly around the continental slope (200 m isodepth in Fig. 3) of the
Chukchi Sea. Average pCO2air measured on board was
385.0 µatm, which is consistent with the value observed at Point
Barrow, Alaska (http://ds.data.jma.go.jp/gmd/wdcgg/). According to JRA-55,
average U10 in the region north of 70∘ N was 4–5 m s-1
in September 2013.
Variation in SST and SSS on the cruise track was abrupt rather than gradual
(Fig. 3a and b). Therefore, we defined three subregions: (1) Barrow coastal
water (BCW), (2) Canada Basin water (CBW), and (3) Chukchi Sea water (CSW).
The boundary between BCW and CBW was 2 ∘C isotherm at
72.5∘ N, 154.8∘ W. CBW and CSW were separated by the 28 psu
isohaline at 73.3∘ N, 168.3∘ W (Fig. 3c).
Summary of three water types (BCW, Barrow coastal water; CBW, Canada
Basin water; and CSW, Chukchi Sea water) at the surface in the western Arctic
Ocean. All samples were pumped up underway from an intake at ship's bottom.
Values are averages for samples collected from 4 to 11 September 2013. N
denotes the number of samples. nDIC32 and nTA32 are DIC and TA
normalized to S = 32, respectively
(nDIC32 = DIC / S⋅32; nTA32 = TA / S⋅32). Standard deviation (SD) was listed below each value.
Water
N
T
S
DIC
nDIC32
pCO2
TA
nTA32
DIC / TA
fRRO
fSIM
Type
(∘C)
(µmol kg-1)
(µmol kg-1)
(µatm)
(µmol kg-1)
(µmol kg-1)
BCW
109
2.88
27.01
1827
2166
274
1948
2309
0.938
0.11
0.08
SD
±0.30
±1.37
±72
±34
±13
±85
±25
±0.006
±0.02
±0.04
CBW
118
0.66
26.19
1803
2203
332
1882
2299
0.958
0.10
0.12
SD
±0.58
±0.24
±19
±16
±19
±16
±9
±0.004
±0.01
±0.01
CSW
54
3.03
31.06
1923
1982
198
2131
2196
0.903
-0.01
0.08
SD
±0.23
±0.19
±13
±6
±19
±12
±3
±0.002
±0.00
±0.01
The fraction of freshwater had distinct spatial variations among the three
subregions (Fig. 3d and e; summarized in Table 1). Low salinity in BCW was
mainly due to riverine outflow: in this subregion, fRRO was as large as
0.18, presumably because the Alaskan coastal current which flows northward
along the Alaskan coast toward Point Barrow contains a considerable fraction
of Yukon River outflow (Steele et al., 2004). In contrast, the lower
salinity in the CBW was primarily due to the input of meltwater from sea
ice, although it also contained significant riverine outflow. In the
northernmost region of the Canada Basin (north of 74∘ N),
fSIM was as large as 0.16, whereas fRRO was no more than 0.10 and
was almost always lower than fSIM. CSW was largely composed of Pacific
water and rarely contained riverine outflow as it flowed directly from the
Bering Strait.
Variations in carbonate chemistry in the surface layer
Remarkable differences in pCO2sea, DIC, and TA were observed
among the three subregions (Table 1, Fig. 3f, g, and h). We attributed the
low DIC / TA ratio and the low pCO2sea (< 200 µatm)
in CSW to the massive biological activity there in early summer. In this
region, net primary production decreases pCO2sea to 200 µatm or
less in early summer (Bates, 2006). According to the analysis
of satellite imagery, NPP in July 2013 was as high
as 1000 mgC m-2 day-1 in the majority of the Chukchi Sea. Even
though NPP had decreased to ∼ 500 mgC m-2 day-1 in September
2013 during our measurements, pCO2sea had been notably lower
than pCO2air for months after the massive primary production in
early summer. This was because of both a still relatively high biological
production and slow net CO2 exchange. Under typical summer conditions
(T= 3 ∘C, S= 32, DIC = 2000 µmol kg-1, TA = 2220 µmol kg-1, mixed
layer depth = 15 m, and U10= 5.0 m s-1), τ1/2 is
considered to be longer than 100 days.
The DIC / TA ratio in CBW was higher than that in CSW (Table 1). The value of
pCO2sea in CBW ranged from 300 to 360 µatm (Fig. 3f).
Although the level of pCO2sea in CBW was still lower than
the pCO2air (385 µatm), it was much higher than that in CSW.
The primary cause of the pCO2sea being nearly as high as
pCO2air is that the addition of meltwater to the surface layer
shoals the mixed layer (Cai et al., 2010; Else et al., 2013), thereby
reducing the time for surface water to reach air–sea CO2 equilibrium.
An additional cause of higher pCO2sea in the high fSIM
region is probably low net primary production, because the concentrations of
nutrients in meltwater are low; e.g., Lee et al. (2012) reported that the
concentration of nitrate in a melt pond being formed on the top of sea ice in
the Canada Basin was low (< 0.5 µM) and that the low nitrate
concentration limited biological production in the pond. Our results
corroborate previous reports by Cai et al. (2010) and Else et al. (2013) that
the overspreading of the surface layer by sea-ice melt inhibits CO2
uptake by the ocean. Near equilibrium pCO2sea conditions
after seasonal sea-ice retreat is likely to be common in the Canada Basin.
The impact of sea-ice melt itself on pCO2sea was difficult
to resolve only from our observations. Bates et al. (2014) found both basic
(i.e., DIC / TA < 1) and relatively acidic (i.e.,
DIC / TA > 1) melt ponds in the Canada Basin. To study the impact of
meltwater on carbonate chemistry, direct sampling of sea ice into gastight
bags (Fransson et al., 2013) will be required.
In BCW, pCO2sea was about 270 µatm on average,
between that in CSW and CBW (Fig. 3f). The fraction of freshwater indicates
that surface freshening in BCW is mainly caused by riverine outflow (fRRO= 0.11) rather than sea-ice melt (fSIM= 0.08). Riverine
outflow had a higher TA / S ratio than sea-ice melt (Fig. 2). It also has
larger content of DIC (Ulfsbo et al., 2014). In our measurements, surface
chlorophyll a was higher in BCW (0.4 to 2.0 mg dm-3) than in CBW
(0.1 to 0.3 mg dm-3), implying that biological drawdown of DIC was
greater in BCW. Consequently, both DIC / TA and pCO2sea in BCW
were lower than those in CBW. At the time of our observation, BCW still could
absorb more CO2 from the atmosphere than offshore CBW. This is an
important finding because river water inflow into the Arctic Ocean is
considered highly likely to increase with climate change (McClelland et al.,
2006; Déry et al., 2009).
Variations in the water column
Water properties differed not only in the surface but also in the water
column among these three. T-S diagrams obtained by CTD for each subregion
are shown in Fig. 4. The surface around Point Barrow was fresh and warm; as
depth increases, the water column gradually cooled to the coldest water
around S= 33 (Fig. 4a). A similar decrease in temperature from near the
surface to bottom was observed in the Chukchi Sea (Fig. 4b). In contrast, the
water column in the Canada Basin was more complex, with a number of
temperature maxima and minima (Fig. 4c). Jackson et al. (2010) classified the
water column in the Canada Basin from the top to the bottom into a surface
mixed layer, a near-surface temperature maximum (NSTM), a remnant of the
winter mixed layer (rWML), Pacific summer water (PSW), and Pacific winter
water (PWW). The surface mixed layer had the lowest salinity (S < 27) because almost all sea-ice melt is trapped in this layer
during summer. The NSTM is separated from the surface mixed layer by
stratification and warmed by the input of solar radiation. The depths of NSTM
ranged between 15 and 26 m during our observations. Below the NSTM, the rWML
corresponded to the temperature minimum (T≈-1∘,
S≈29.3), which was formed in the Canada Basin during the
previous winter. Another temperature maximum around S= 30.5
corresponded to PSW, which was advected and modified in the Chukchi Sea
during summer. The lowest temperature observed was near the freezing point in
PWW at around S= 33.1.
Column salinity and potential temperature in (a) BCW (Barrow
coastal water), (b) CSW (Chukchi Sea water), and (c) CBW (Canada Basin
water). Water pressure is indicated by color. Gray contours indicate
potential density (σθ= {density - 1} × 1000 [kg m-3]).
![]()
Water-column profiles of (a) temperature, (b) salinity, (c) fSIM,
(d) fRRO, (e) pCO2sea, and (f) apparent
oxygen utilization (AOU) along the cruise track in the period 4–11 September
2013. Data were obtained by CTD and XCTD in (a) and (b), by oxygen sensor
SBE43 on CTD in (f) and by discrete bottle samples in (c), (d), and (e).
Water types BCW (Barrow coastal water), CBW (Canada Basin water), and CSW
(Chukchi Sea water) are indicated at the top of the figure.
![]()
Temperature and salinity were frequently measured along the cruise track by
CTD and XCTD sensors. Water-column profiles showed a distinct halocline from
10 to 20 dbar in BCW and CBW (Fig. 5a and b). In these two subregions, the
difference in salinity between above and below the halocline was up to 2.
Unlike the other two subregions, thermocline was more prominent than
halocline in CSW. Column variation of fSIM and fRRO indicate that
both sea-ice meltwater and river outflow greatly contributed to the formation
of the halocline (Fig. 5c and d). In the Canada Basin, fSIM was as high
as 0.12 ± 0.01 (±standard deviation) in the top layer down to 10 dbar,
but decreased abruptly with depth to practically zero (0.01 ± 0.01) in
the 29–50 dbar layer. Likewise, fRRO was also quite high (0.09 ± 0.01)
in the top layer down to 10 dbar and decreased gradually to
0.06 ± 0.02 in the 29–50 dbar layer. This vertical structure indicates
that the input of sea-ice melt occurred shortly before the measurement (at
least in summer 2013) and contributed to the formation of a discrete layer
in the surface over the main water mass in the Canada Basin, whereas river
outflow had undergone vertical mixing in the course of advection before it
reached the Canada Basin.
Among these three subregions, differences were also evident in column
pCO2sea. In the upper layer (above 10 dbar) of CSW and BCW,
average pCO2sea was 195 ± 11 µatm and 258 ± 14 µatm,
respectively. As mentioned in Sect. 4.2, these low
pCO2sea values were the result of net primary production. In
these subregions, pCO2sea increased with depth below the
halocline. The water-column profile of AOU indicates that the increase in
pCO2sea was due to the input of CO2 associated with the
degradation of organic matter (Fig. 5e and f).
Variation in several parameters of Canada Basin water against
salinity. (a) pCO2sea in discrete bottle
samples; (b) apparent oxygen utilization (AOU) from CTD cast data; (c) preformed
nDIC32 (= {DIC - AOU ⋅117/170}/S⋅32) in discrete bottle samples and dissolved nitrate (NO3+ NO2+ NH4) in logarithmic scale in discrete samples. Salinity
of rWML (S= 29.3) and PWW (S= 33.1) is indicated as gray
dotted lines.
![]()
Unlike the water-column profiles of pCO2sea and AOU in CSW
and BCW, those in CBW were distinctive in that they had subsurface minima. In
the top 10 dbar of CBW, pCO2sea reached 322 ± 20 µatm,
a value still lower than pCO2air
(∼ 385 µatm) but the highest among the three subregions.
However, pCO2sea decreased with depth below the halocline
and reached 271 ± 31 µatm in the range of
29.3 < S < 31.3 (30 to 50 dbar layer; Figs. 5e and 6a). Below the halocline in
CBW, AOU was significantly negative
(< -20 µmol kg-1) like that in the CSW, where net
primary production was large (Figs. 5f and 6b). Subsurface maximum of
chlorophyll a and dissolved oxygen have also been found in the Canadian
Archipelago (Martin et al., 2010). Here, our frequent observations
facilitated classification of the water masses and their origins in the
Canada Basin. According to the salinity–AOU profile (Fig. 6b) in the
Canada Basin, AOU was largely negative
(< -20 µmol kg-1) in the salinity range of 28 to
31.5, which corresponds to NSTM, rWML, and PSW. NSTM and rWML were formed in the
Canada Basin. Nitrates were almost depleted
(< 0.2 µmol kg-1) in these two layers during our
observations (Fig. 6c). Even in spring, concentration of nitrate was low
(< 2 µmol kg-1) in surface water of the Canada Basin
(Codispoti et al., 2005). Sunlight surely reaches the 50 m depth in the
Canada Basin although it is not strong (Jackson et al., 2010). Negative AOU
in NSTM and rWML indicated the large biological production utilizing the
nutrients and the sunlight. Excess oxygen produced by the biological
production remained in subsurface as it was isolated from surface by strong
halocline. Results of our measurements also showed that significant nutrients
remained in PSW, where S > 29.3 (Fig. 6c). Oversaturation of
oxygen in PSW was due to the remnant of massive biological production in
early summer and/or the production undersea during advection from the Chukchi
Sea to the Canada Basin.
To compare the water properties among layers, we calculated preformed
nDIC32 as defined by the following equation.
PreformednDIC32=DIC-AOU×rC:OS×32
Here, rC:O denotes stoichiometric ratio of DIC to AOU being set to
117/170 (Anderson and Sarminento, 1994). As shown in Fig. 6d, preformed
nDIC32 was almost constant and the highest in the water column from
surface to the depth where salinity was 29.3. This indicates that the water
above the depth where salinity was 29.3 had the same origin in the Canada
Basin. There was no clear minimum of pCO2sea in NSTM and
rWML in spite of the negative AOU and biological production. This suggests
that the pCO2sea minimum in the PSW below was explained by
the drawdown of DIC due to biological production when the PSW was in the
surface of the Chukchi Sea, in addition to the undersea DIC drawdown. In
fact, preformed nDIC32 in the salinity range of 29.3 to 33.1 that
corresponded to PSW and PWW from the Chukchi Sea was about lower by about 100 µmol kg-1
than in the upper layers from Canada Basin (Fig. 6d). However, we have to note that the conventional salinity normalization
like those Eq. (10) overestimates the nDIC in a source water when it
is diluted with river runoff and/or sea-ice melt that contain DIC (Friis et
al., 2013). In a stricter sense, preformed DIC in the subsurface waters of
the Canada Basin is approximated by
preformedDIC=fPOW⋅preformedDICPOW+fSIM⋅DICSIM+fRRO⋅DICRRO.
Preformed DICPOW was determined to be 2170 µmol kg-1
from salinity, DIC, and AOU in temperature minimum layer (PWW). DICSIM
and DICRRO were assumed to be 300 µmol kg-1 (Fransson et
al., 2013) and 800 µmol kg-1 (Tank et al., 2012),
respectively. For the PSW in which salinity ranged from 29.3 to 31.5, fSIM= 0.01, and fRRO= 0.06, preformed DIC observed was 2027 µmol kg-1.
This was lower by 34 µmol kg-1
than that calculated from Eq. (11) (2061 µmol kg-1),
which suggests the biological DIC drawdown in the preformed condition of the
PSW. However, for rWML and shallower in which S < 29.3, fSIM= 0.12, and fRRO= 0.09, preformed DIC observed was 1850 µmol kg-1.
This was consistent with that calculated from Eq. (11) (1858 µmol kg-1). These results allowed us to
support to conclude that pCO2sea minimum in subsurface in
the Canada Basin was attributable not only to large biological drawdown of
DIC but also to lower DIC in PSW as compared with Canada Basin origin water
lying above.
Future direction of hidden CO<inline-formula><mml:math id="M513" display="inline"><mml:msub><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msub></mml:math></inline-formula> sink in the Canada Basin
How the long-term retreat of sea-ice changes the air–sea CO2 flux in
the Arctic Ocean is a controversial matter. Manizza et al. (2013) argued
that increasing SST will enhance biological primary production and drawdown
of CO2 in seawater. Laruelle et al. (2014) asserted that larger ice-free
areas and longer ice-free periods will provide greater occasion for oceanic
CO2 uptake. In contrast, Cai et al. (2010) and Else et al. (2013)
insisted that the increase in sea-ice melt results in the formation of thin
surface mixed layers and limits further uptake of CO2 from the
atmosphere by this layer.
As a result of our observations that a subsurface minimum of
pCO2sea existed in the Canada Basin, it is necessary to
study whether the surface mixed layer there will deepen under a warming
climate. If the surface layer is stirred by strong wind and mixed with the
subsurface low pCO2sea layer, the surface will act as a
further CO2 sink. Several reports indicate that the strong wind
associated with the passage of low pressure systems deepens the surface mixed
layer and has impacts on the biogeochemistry (Wada et al., 2011; Rumyantseva
et al., 2015). Simmonds and Keay (2009) reported that the strength of
cyclones in the Arctic Ocean is increasing with the long-term reduction of
sea-ice cover. However, we also have to consider the strength of
stratification in the Canada Basin. In a comprehensive analysis of mixed
layer depth in the Arctic Ocean, Peralta-Ferriz et al. (2015) found a
significant positive correlation between the mixed layer depth and the
maximum wind speed in the preceding 5 days (4.6 ± 0.8 m per m s-1) in the case that the differences in density between the mixed
layer and 20 m below (Δρ) is smaller than 0.5 kg m-3.
However, in the case of Δρ > 0.5 kg m-3,
deepening of the mixed layer is much less sensitive to the increase in wind
speed (0.77 ± 0.52 m per m s-1). In our observations, Δρ exceeded 2.0 kg m-3 at all CTD stations in the Canada Basin
(Fig. 4c). Hence, we suggest that additional CO2 uptake in the Canada
Basin by wind mixing is unlikely because stratification was strong even in
2013 and will be further strengthened by the additional input of sea-ice melt
in the future.
Climate change also affects the subsurface layer in the Canada Basin, where
low pCO2sea is caused by net primary production. McLaughlin
and Carmack (2010) reported that increase in sea-ice melt and the
strengthening of Ekman pumping deepened the nutricline and the depth of
chlorophyll maximum in the Canada Basin. Nishino et al. (2013) also observed
decreases in nitrate and chlorophyll in the 0 to 50 m depth layer in the
Canada Basin during 2002–2010; they attributed these decreases to the
decrease in inflow of nutrient-rich water from the East Siberia Sea. In
either case, biological production below the halocline of the Canada Basin is
likely to decrease in the long term. In this regard, it seems unlikely that
the subsurface low pCO2sea layer in the Canada Basin will
act as another CO2 sink.
Conclusions
A wide range of surface pCO2sea was observed in the western
Arctic Ocean in September 2013. The value was as low as 180 µatm in
the Chukchi Sea, where biological activity was high in early summer. In
contrast, pCO2sea in the Canada Basin in September reached
360 µatm, the value comparable to pCO2air. Based on the
relationship between salinity and TA, we attributed the low-salinity water in
the Canada Basin mainly to the input of sea-ice melt. Large input of
oligotrophic sea-ice melt not only inhibits biological activity but also
facilitates to form a thin surface mixed layer that is easier to reach
equilibrium with respect to the atmospheric CO2. In the area where
mixing with riverine output was more dominant than with sea-ice melt, the
increase in pCO2sea was indistinct due to the input of
riverine nutrients and TA.
In the Canada Basin, pCO2sea was the lowest
(∼ 250 µatm) under the surface mixed layer below a strong
halocline (difference in density is larger than 2.0 kg m-3). This
differs from other regions where the lowest pCO2sea was
observed in the surface. This subsurface pCO2sea minimum
corresponds to PSW and is attributable to the lager net primary production
and originally lower DIC of PSW compared to those water of Canada
Basin origin. The subsurface low pCO2sea layer in the Canada
Basin has a potential to absorb CO2 from the atmosphere in the case when
it mixes with the surface by a strong turbulence. However, such an increase
of CO2 absorption is unlikely because this stratification is strong
enough to resist vertical mixing by wind. Additionally, long-term
observations in the Canada Basin suggest that subsurface biological activity
has been declining in recent decades.
Our observations could reveal only a part of the complex carbon cycle in the
Arctic Ocean. The subsurface pCO2sea minimum is specific to
the Canada Basin where circulation of waters generates a complicated
water-column structure; the results here are unlikely to be applicable to the
entire Arctic Ocean. In the changing Arctic Ocean, in spite of the finding
that these subregional variations and processes are essential for better
projections of the future carbon cycle, they are not adequately reflected in
current models. The areas we can observe in the Arctic Ocean are expected to
expand along with the long-term sea-ice retreat. Comprehensive observations
are essential especially in such areas because sea-ice melt may cause other
effects that are unknown to date.