Ocean acidification in the three oceans surrounding northern North America
Center for Advanced Science and Technology, Tokyo University of Marine Science and Technology, Minato, Tokyo, Japan
Corresponding author: M. Yamamoto-Kawai, Center for Advanced Science and Technology, Tokyo University of Marine Science and Technology, 4-5-7 Konan, Minato, Tokyo 108-8477, Japan. (email@example.com)
 The uptake of anthropogenic CO2 drives ocean acidification, with attendant effects on the saturation state of calcium carbonate (Ω) and marine ecosystems. Here, we examine ocean acidification within the context of large-scale water mass exchange and local physical and biogeochemical processes along a section around northern North America. Waters in the North Pacific are preconditioned by the global-scale circulation to be low Ω source waters and as they move northward across the Bering and Chukchi seas they are modified by biological activity. These waters then enter the Canada Basin and Canadian Arctic Archipelago where cooling, river discharge, sea ice formation and biological activities modify water mass characteristics further. Continuing eastward into Baffin Bay, relatively low Ω waters extend from the surface to depth due to remineralization. Changes in Ω are large along this pathway and the response of local marine ecosystems will likewise be shaped by the biogeography of species occupying a given region.
 Ocean acidification is an emerging issue in global change. The uptake of anthropogenic CO2 alters the chemistry of seawater to make the world's oceans more acidic and decrease the saturation state of calcium carbonate (Ω). As calcium carbonate (CaCO3) constructs shells and exoskeletons of marine calcifying organisms such as pteropods, clams, mussels, oysters, and corals, a decrease in Ω has substantial consequences to marine ecosystems as well as fisheries [Kroeker et al., 2010; Cooley and Doney, 2009]. Accordingly, much recent attention has been given to the uptake of anthropogenic CO2 and its consequences to marine organisms [e.g., Gattuso and Hansson, 2011; Riebesell and Tortell, 2011]. Here, we examine the added effects of large-scale water mass exchange on Ω and show that local responses to climate change are also heavily dependent on local biogeochemical processes and modifications that occur along advective pathways and interbasin exchange.
 For aragonite or calcite, the two types of CaCO3 secreted by marine organisms, Ω is expressed by the product of carbonate (CO32−) and calcium (Ca2+) ion concentrations in seawater relative to the stoichiometric solubility constant (K′SP) at a given temperature (T), salinity (S), and pressure (P) [Mucci, 1983]:
 When Ω > 1 conditions are favorable to the formation of CaCO3 shells and skeletons, but when Ω < 1 conditions are corrosive, and therefore dissolution of CaCO3 shells can occur in the absence of protective mechanisms [Ries et al., 2009]. Since the industrial revolution, Ω has decreased worldwide due to anthropogenic increases in the partial pressure of atmospheric CO2 (pCO2). The global mean Ω for aragonite at the ocean surface has decreased from a preindustrial value of 3.6 to a value of 2.9 in 2007 and may further decrease to 1.8 by the end of the 21st century [Gattuso and Lavigne, 2009]. The spatial distribution of Ω, however, is not geographically uniform [Feely et al., 2009]; for example Ω is lower in high-latitude surface waters than in tropical or temperate waters because of temperature gradients (colder water dissolves more CO2), upwelling of CO2-enriched water in the Southern Ocean and northern North Pacific [Fabry et al., 2008], and massive inflows of river water into the Arctic Ocean [Yamamoto-Kawai et al., 2011; Tank et al., 2012]. Model simulations predict that surface waters will reach seasonal aragonite undersaturation (Ω < 1) by the year 2030 in the Southern Ocean and by 2100 in the subarctic North Pacific [Orr et al., 2005; McNeil and Matear, 2008]. In the Arctic Ocean, surface water was projected to reach aragonite undersaturation partly by 2016 and entirely by 2040 [Steinacher et al., 2009]. In fact, surface waters in the Canada Basin of the Arctic Ocean became undersaturated as early as 2008 due to inputs of unprecedented melting of summer sea ice [Yamamoto-Kawai et al., 2009a].
 Spatial and temporal heterogeneity of Ω is determined by an interaction of physical and biogeochemical processes associated with, for example, warming/cooling, freshening/evaporation, formation/dissolution of CaCO3, air-sea exchange of CO2, and photosynthesis/remineralization of organic matter. Due to large-scale advection, transformations that occur locally in one region affect Ω in downstream regions and these effects are additive. Therefore, the factors and processes that affect Ω must be quantitatively understood. Here, we examine the distribution of Ω along a ∼13,000 km section in the oceans around northern North American and discuss, quantitatively, various physical and biogeochemical processes that affect Ω in each hydrographic region.
 Hydrographic stations were occupied around northern North America in 2007 and 2008 as part of the Canadian IPY project Canada's Three Oceans (C3O) and the international Joint Ice Ocean Study (JOIS) aboard the CCGS Sir Wilfrid Laurier and the CCGS Louis S. St-Laurent (Figure 1). The transect, starting from Vancouver Island in the North Pacific and ending at Newfoundland in the North Atlantic, crosses five hydrographical domains, each with distinct water mass assemblies: eastern subarctic North Pacific, Bering and Chukchi shelves, Canada Basin (and Beaufort Sea), Canadian Arctic Archipelago, and Baffin Bay and the Labrador Sea, as described by Carmack and McLaughlin . The first two domains were sampled in 2008, the last three were sampled in 2007 and the Canada Basin region was sampled in both years. Observations and distributions of T, S, dissolved oxygen, nitrate, and chlorophyll-a along the C3O transect are described in Carmack and McLaughlin . Dissolved inorganic carbon (DIC) and total alkalinity (TA) were collected in the top ∼100 m of the water column, except in the Canada Basin region where sampling was extended to the bottom. DIC were analyzed using a Single-Operator Multi-Metabolic Analyzer coulometer system. Samples for TA were taken from the DIC bottle and analyzed by an open-cell titration with 0.1 N HCl. Measurements were calibrated against a certified reference material (A. G. Dickson, Scripps Institute of Oceanography). The pooled standard deviation (Sp) for duplicate samples was 3.5 μmol/kg or less for both DIC and TA in both years [cf., Giesbrecht et al., 2013]. Because our data are limited to the top ∼100 m, apart from the Canada Basin, gridded data from the Global Ocean Data Analysis Project (GLODAP) [Key et al., 2004] and World Ocean Atlas 2005 (WOA05) [Locarnini et al., 2006; Antonov et al., 2006; Garcia et al., 2006] are included in our treatment of the North Pacific, Bering, and Labrador seas in order to provide a more complete vertical distribution. Specifically, DIC and TA data from GLODAP and T, S, and nutrients data from WOA05 were chosen at a gridding point nearest to our station and combined with our data. Likewise, data at a deep station in Baffin Bay observed in September 2006 [McLaughlin et al., 2010] were also included. The Ω values were estimated from DIC and TA using the CO2SYS program [Lewis and Wallace, 1998] with constants of Lueker et al.  for K1 and K2, Dickson  for KSO4 and Mucci  for K'SP. The uncertainty in both DIC and TA measurements (3.5 μmol/kg) can lead to an error of <0.08 in the estimation of Ω for aragonite (<0.12 for calcite). The same program and constants were used to estimate carbonate parameters in section 3.
 We also measured oxygen isotope ratios (δ18O) of waters in the Canada Basin and at a few stations in the Bering Sea to estimate the mixing ratio of meteoric water (river water + precipitation), sea ice meltwater, and seawater. The mixing ratio was used to quantify the effect of freshwater input on Ω. In order to calculate an anthropogenic CO2 content in seawater, the age of the water was estimated from CFC-12 observed in the Canada Basin in 2008 [Smith et al., 2011] and in Baffin Bay in 2006 [McLaughlin et al., 2010], and obtained from the GLODAP database. It should be noted that GLODAP data represent conditions in 1990s [Key et al., 2004], ∼10 years earlier than our observations, and thus we do not include data from GLODAP/WOA05 data from the top 20 m because the influence of anthropogenic CO2 here leads to a systematic error of 0.04 in Ω.
 To interpret the distribution of Ω quantitatively, we have estimated the effects of anthropogenic CO2, biological activity, and cooling or warming on Ω. For waters in the Canada Basin, effects of mixing with meteoric water and formation and melting of sea ice were also estimated. Calculations were applied to data from the top 300 m of the water column where most of these modifications occur. Following previous studies [Gruber et al., 1996; Gruber, 1998, Sabine et al., 1999, 2002], observed DIC can be expressed as:
where CEQ280 is DIC of water in equilibrium with the preindustrial atmospheric CO2 concentration of 280 ppm at the surface, ΔCDiseq represents air-sea disequilibrium, ΔCAnth is anthropogenic carbon concentration, and ΔCBio is the change in DIC due to biological activity. Each term is calculated by;
where pCO2280 or pCO2t are pCO2 of seawater in equilibrium with the atmospheric CO2 concentration in the preindustrial period or in the year t when the water parcel was last in contact with the atmosphere. TA0 is the preformed TA at the surface, AOU (apparent oxygen utilization) is the difference between the measured oxygen concentration and its equilibrium saturation concentration, and N* is nitrogen deficiency relative to phosphorous due to denitrification (N* = 0.87 × (total inorganic nitrogen – 16 × PO4 + 2.9) [Gruber and Sarmiento, 1997]). The denitrification signal, introduced by Sabine et al.  to estimate ΔCBio, was included in the last term of equation (5) because denitrification increases TA. The year t was estimated from pCFC-12 concentrations [Doney and Bullister, 1992; Warner et al., 1996] except for waters <20 m where the year of observation was used to set the water age to 0. The value of pCO2t was calculated from T and the atmospheric CO2 mole fraction in the year t [Keeling et al., 2001]. The equation of Sabine et al.  was used to estimate TA0 from potential temperature, S and PO (PO = O2 + 170 × PO4 [Broecker, 1974]) for waters between 100 and 300 m in the North Pacific and Bering Sea regions. For waters shallower than 100 m and waters in other regions, TA0 was directly estimated from AOU and N* (TA0 = TA + 16/170 × AOU + N*) because the effect of CaCO3 dissolution should be negligible due to the deep saturation horizon (Figure 2c) and small CaCO3 production [Lee2001]. Now, Ω can be estimated without the contribution from anthropogenic CO2 (ΩNoAnth);
or without the contribution from biological activity (ΩNoBio);
 To estimate Ω without the contribution from cooling/warming (ΩNoTemp), the winter surface T of 3.5°C at the entrance to the Bering Sea (from WOA05) was chosen as the base T to investigate modification of Pacific-origin waters along their transit through the Arctic seas. Assuming that ΔCDiseq and ΔCBio remains the same,
 To estimate the effects of mixing with river water and the formation/melting of sea ice on Ω in the Canada Basin, we use the method introduced by Yamamoto-Kawai et al. . For each water sample in the Canada Basin, fractions of Pacific Water (fPW), Atlantic Water (fAW), meteoric water (fMW), and sea ice meltwater (fSIM) were estimated from S, δ18O, and the nitrogen/phosphorous relationship [Yamamoto-Kawai et al., 2008]. Formation of sea ice is expressed as a negative value of fSIM. Then, S, TA, and TA0 without the contribution from meteoric water (MW) or sea ice meltwater (SIM) were estimated;
where X represents each property (S, TA, or TA0) and the subscript FW represents the freshwater source, either MW or SIM. End-member values were set to be TAMW = 1048 μmol kg−1 using the mean observed value in Arctic rivers [Tank et al., 2012], and TASIM = 441 μmol kg−1 from sea ice observations by Rysgaard . For DIC, the effect of mixing with MW or SIM cannot be removed by equation (9) since dilution (or concentration) of DIC alters pCO2 and subsequently air-sea CO2 exchange at the surface further changes DIC. Accordingly, Ω without the contribution of MW or SIM was estimated with the assumption that ΔCDiseq and ΔCBio remain the same;
 Finally, the difference between the observed Ω and each of the ΩNO values was calculated (ΔΩ) and represents the effect of each of the processes on Ω. Note that ΔCBio was set to 0 for waters shallower than 20 m, because it is assumed that TA = TA0 and AOU = 0 in surface waters in contact with atmosphere (and thus there is no denitrification). In reality, summer photosynthesis results in negative ΔCBio and increases ΩNoBio; and therefore our assumption underestimates ΔΩBio in surface waters. Other ΔΩ are not affected by this assumption, however, because the disequilibrium in DIC at the surface caused by photosynthesis is included in ΔCDiseq (see equations (7)-(10)).
4. Evaluation of Errors
 Errors in the various ΩNo values largely depend on the uncertainty of calculating TA0, CEQ, ΔC, and XNoFW terms. For waters below 100 m in the North Pacific and Bering Sea regions, the standard error of TA0 (σTA0) is ±9 µmol kg−1 [Sabine et al., 2002]. For waters elsewhere, σTA0 is estimated to be ±4 µmol kg−1 based on uncertainties of TA (σTA) and oxygen measurements (σO), and calculations of saturated oxygen (σOsat) and N* (σN*);
where σTA = 3.5 µmol kg−1, σOsat = 4 µmol kg−1, σO = 1 µmol kg−1 [Anderson and Sarmiento, 1994; Gruber et al., 1996], and σN* ranging between 0.3 and 2.6 µmol kg−1 (cf. Gruber and Sarmiento  for estimation). The σTA0 of 9 or 4 µmol kg−1 causes an error of ± 8 or 3.5 µmol kg−1, respectively, in a calculated value of CEQ280. For ΔCAnth, most of the error (σΔCAnth) comes from σTA0, σC280, and the uncertainty in pCO2t as calculated from pCFC-12 age. Because CFC-12 at the surface may be undersaturated and the mixing of waters with differing ages may be nonlinear, the pCFC-12 age can be different from the true water age. For ages younger than 30 years old, the difference should be less than 25% [Lee et al., 2003] and thus our data has an uncertainty of ±12 µatm in pCO2t. The potential error in CEQ in the year t (the first term on the right side of the equation (3)) was estimated by changing both TA0 and pCO2t within ranges of their uncertainties. By quadratic error addition of this and the error in CEQ280, we estimate σΔCAnth to be 5–16 µmol kg−1. The error in ΔCBio is estimated by propagating through uncertainties in sampling and analysis (σTA and σO), determination of TA0 (σTA0) and N* (σN*), stoichiometric ratios (σRC:O and σRN:O), and possible disequilibrium in O2 (σOsat) when the water lost its contact with the atmosphere by following previous studies [Gruber et al., 1996; Gruber, 1998, Sabine et al., 1999, 2002].
where σTA = 3.5 µmol kg−1, σRC:O = 0.092, σRN:O = 0.0081, σOsat = 4 µmol kg−1, σO = 1 µmol kg−1 [Anderson and Sarmiento, 1994; Gruber et al., 1996]. For σTA0 and σN*, values estimated above were used. Estimated σΔCBio ranges between 4 and 27 µmol kg−1, depending on AOU as discussed in Gruber et al. . For example, σΔCBio is 6 µmol kg−1 at AOU = 50 µmol kg−1 and increases to 15 µmol kg−1 at AOU = 150 µmol kg−1. The error of ΔCDiseq (σΔCDiseq) estimated from σCEQ, σΔCBio, and σDIC (3.5 µmol kg−1) ranged between 6 and 30 µmol kg−1, again depending on AOU. Freshwater fraction of fMW or fSIM has uncertainty of ±0.03 [Yamamoto-Kawai et al., 2008]. This causes an error of as large as 1.5 to 3% in calculated values of XNoFW (equation (9)). The potential errors of ΩNo values were then estimated by changing TA0, CEQ, ΔC, and XNOFW within the range of uncertainties and are illustrated in Figures 3f–3j.
5. Results and Discussion
 Conceptually, one can visualize the water mass transit and associated modifications in Ω in a pan-northern North America section (Figure 2) that starts in the North Pacific near the terminus of the global overturning circulation (GOC) and ends in the Labrador Sea near the start of the GOC [cf. Carmack et al., 2010]. Although the majority of water upwelled in the North Pacific moves southward, a significant portion moves northward due to the difference in steric height between the Pacific and Arctic Ocean [Coachman and Aagaard, 1966]. Our section follows this pathway of Pacific-origin water and freshwater around northern North America [Wijffels et al., 1992; Yamamoto-Kawai et al., 2010; Carmack and McLaughlin, 2011; Jones et al., 2003]. First, we describe the distribution of Ω in each region and, because aragonite is more soluble than calcite, we focus on the aragonite saturation state unless noted. We then present the various estimates of ΔΩ to investigate quantitatively the various modification processes that occur along the section.
5.1. Distribution of Ω
5.1.1. Eastern Subarctic North Pacific and Bering Sea Basin
 In the northeastern North Pacific, the horizon depth of aragonite saturation is 100–300 m (Figure 2c) and of calcite saturation is 200–500 m (not shown). These depths are much shallower than aragonite and calcite horizons in the North Atlantic Ocean, ∼2500 and 4000 m, respectively [Feely et al., 2004]. The shallower depth reflects the rainout and remineralization of organic matter in the poorly ventilated old waters of the subarctic North Pacific (Figure 2d). At the surface, Ω is ∼2 and relatively low Ω values (<2) are observed near the upwelling dome of the cyclonic Alaska Gyre (distance ∼2000 km) [cf. Stabeno et al., 2004; Carmack and McLaughlin, 2011]. These low Ω Pacific waters feed the Bering Sea through the shallow and deep straits of the Aleutian Islands, and thus the chemical properties of Bering Sea water are similar. In deep waters below ∼1500 m, however, Ω is somewhat lower in the Bering Sea than in the North Pacific. This is a result of additional rainout and remineralization of organic matter that takes place in the Bering Sea and indicated by higher AOU and much higher silicate and phosphate concentrations in the Bering Sea than in the North Pacific at the same depth. The saturation horizons of aragonite and calcite in the Bering Sea Basin are at ∼200 and ∼400 m, respectively.
5.1.2. Bering and Chukchi Shelves
 Relatively nutrient-rich subsurface waters of the Bering Sea Basin upwell onto the shallow Bering Sea shelf and subsequently flow northward through Bering Strait onto the Chukchi Sea shelf, making these shelves highly productive areas [e.g., Codispoti et al., 2005]. Consequently, water entering the Arctic Ocean through Bering Strait is strongly modified by photosynthesis and remineralization that occurs on these shelves and aided by the strong pelagic/benthic coupling of the region [Grebmeier et al., 2006a]. Phytoplankton production in surface waters lowers the partial pressure of CO2 (pCO2) and increases Ω, and the vertical export of organic matter and subsequent remineralization increases pCO2 and decreases Ω in the bottom waters to <1 in some locations on these shelves [Bates et al., 2009; Mathis et al., 2011]. Our section crosses several biological hot-spots where the high flux of organic matter from the surface layer increases pCO2 at the bottom: southwest of St. Lawrence Island; in the Chirikov Basin; in the southern Chukchi Sea; and at the head of Barrow Canyon [Grebmeier et al., 2006b]. Although station coverage is sparse for DIC/TA data, we determined Ω in bottom waters at four stations and found aragonite to be undersaturated at a station southwest of St. Lawrence Island. Because the transit time for waters to cross the Bering and Chukchi shelves is approximately 1 year, cooling and brine rejection during sea ice formation yields shelf water with high pCO2, low Ω, low T, and relatively high S (∼33) during winter. This winter variety of Pacific-origin water subsequently drains into subsurface layers of the Canada Basin in the western Arctic Ocean [McLaughlin et al., 2004].
5.1.3. Canada Basin
 The Canada Basin is globally unique in its vertical profile of Ω (Figure 2c). At the surface, Ω is low (∼1) due to cooling and the dilution (Figures 2a and 2b) of DIC, TA and [Ca2+] by a large input of river water from the surrounding continents [Tank et al., 2012]. Sea ice meltwater is the secondary source of freshwater in the Canada Basin and its contribution has increased significantly with the recent extensive melting of sea ice [Yamamoto-Kawai et al., 2008, 2009b, 2011]. Both inflowing river and sea ice melt waters have low alkalinity values and thus locally reduce the buffering capacity of seawater [Broecker and Peng, 1982; Yamamoto-Kawai et al., 2011]. In 1997, surface waters in the Canada Basin were oversaturated with respect to aragonite [Jutterström and Anderson, 2005], however, by 2008 these surface waters were undersaturated [Yamamoto-Kawai et al., 2009a]. This decrease was mainly due to the addition of sea ice meltwater and the enhanced dissolution of CO2 by the reduction of sea ice coverage [Cai et al., 2010; Yamamoto et al., 2012; Yamamoto-Kawai et al., 2011]. Beneath the cold and fresh surface mixed layer lies Pacific Summer Water (PSW), characterized by a T maximum (Figure 2a). The first subsurface Ω maximum is found within the PSW layer, formed by advection of warm and low-CO2/high-Ω water from the upstream and highly productive Bering and Chukchi shelves and modified locally by photosynthesis at the top of the nutricline [McLaughlin and Carmack, 2010; Carmack and McLaughlin, 2011]. Below lies Pacific Winter Water (PWW) and these advected waters are undersaturated (Ω = 0.8) as a result of the cooling and remineralization that took place upstream on the Bering, Chukchi, and East Siberian shelves, as shown by low T and high concentrations of nutrient and AOU (Figures 2a and 2d). Below the PWW lies Atlantic water (AW), characterized by a T maximum near 400 m depth and a second Ω maximum (Figures 2a and 2c). Below 1000 m, Ω values are almost identical to those found in the northern North Atlantic source waters. Although the deep saturation horizon for aragonite is at about 2500 m [Jutterström and Anderson, 2005], the bottom water in the Canada Basin at ∼3800 m is oversaturated with respect to calcite. In summary, Ω in the Canada Basin is low at the surface (Ω = ∼1) and in the PWW layer (Ω = ∼0.8), with local maxima found in the PSW (Ω = ∼1.3) and AW (Ω = ∼1.3) layers.
5.1.4. Canadian Arctic Archipelago (CAA)
 Arctic waters from the surface and Pacific water layers exit the Canada Basin through the CAA to enter Baffin Bay and the Labrador Sea. The bathymetry and flow are complex in this domain [see McLaughlin et al., 2006; Carmack and McLaughlin, 2011 for hydrographic conditions]. In 2005, Chierici and Fransson  observed Ω in surface waters and found Ω < 1 in the vicinity of Coronation Gulf due to dilution by river runoff and sea ice melt. In our survey, surface Ω values ranged between 1.0 and 2.3 in the CAA, increased to a subsurface maximum and then decreased with depth. For the deepest sample at each station (at 43–84 m depths), Ω ranged between 0.6 and 1.2. We found aragonite undersaturation below 10 m at a station in Coronation Gulf (Ω = 1.0 at 10 m and decreased to 0.6 at 57 m), and below 40 m at four of seven stations. Apart from observations in Coronation Gulf and Queen Maud Gulf, the properties of low Ω waters are similar to PWW found in the Canada Basin domain (S ∼33, low T, high nutrient) and thus low Ω waters at depth in the CAA are due to the outflow of Pacific-origin water from the Canada Basin. Local regeneration of organic matter at depth, as evident in high nutrient and low oxygen concentrations [Carmack and McLaughlin, 2011], further lowers Ω. Both Coronation and Queen Maud gulfs receive a large input of freshwater from regional rivers [Carmack and McLaughlin, 2011] which freshens the entire water column to S < 30. Observations of δ18O and Ba in 2002 showed that main source of freshwater in Coronation Gulf is meteoric water and that sea ice meltwater is the secondary source of freshwater [Yamamoto-Kawai et al., 2010]. Thus, dilution by local river water and sea ice meltwater lowers Ω in surface waters in the western part of the CAA as suggested by Chierici and Fransson , on an underlying foundation of low Ω Pacific-origin waters.
5.1.5. Baffin Bay and Labrador Sea
 Waters exiting from the Arctic Ocean and the CAA from the north and west and Atlantic-origin water from the south entering via the West Greenland Current meet and mix in Baffin Bay and Labrador Sea [Lobb et al., 2003; Tan et al., 2004]. In the cyclonic gyre of Baffin Bay, Ω is relatively low and deep water is undersaturated with respect to both aragonite and calcite. Deep regeneration of surface-derived organic matter in the semienclosed Baffin Bay [Tremblay et al., 2002] increases nutrients, pCO2, and AOU, and decreases Ω at depth [Azetsu-Scott et al., 2010] (Figures 2c and 2d). The saturation horizon here lies at ∼600 m for aragonite (Figure 2c) and ∼1300 m for calcite. In the Labrador Sea, south of Davis Strait, the water column above ∼2300 m is oversaturated with respect to aragonite and the entire water column is oversaturated with respect to calcite (Figure 2c), owing to the presence of Atlantic-origin water with high Ω values that enter from the south via the West Greenland Current and retroflect westward above the Davis Strait sill [Azetsu-Scott et al., 2010]. This circulation feature creates a pronounced Ω front above Davis Strait, with low values to the north and high values to the south. Arctic outflow with high Pacific-origin water content and low Ω remains traceable on the western Labrador Sea continental shelf, although it exhibits modifications due to tidal mixing and freshwater input during transit [Azetsu-Scott et al., 2010].
6. Quantitative Analysis
 In surface and subsurface layers, low Ω waters were found along the pathway of Pacific-origin water (Figure 2) that enters the Bering Sea, upwells onto and crosses the Bering/Chukchi shelves, and flows into the Atlantic side of northern North America via the Canada Basin and CAA. Along this pathway, Ω is modified by input of anthropogenic CO2, biological activity, cooling/warming, mixing with river water, and the formation or melting of sea ice. The effects of each of these processes, described above in the calculation section, are estimated and shown in Figure 3. Except for ΩNoBio, calculations are limited to stations where CFC-12 data are available and for all stations the age is assumed to be zero at depths of 0–20 m. Values for ΩNoMW and ΩNoSIM were calculated only for the Canada Basin stations where δ18O data are available.
 The distribution of ΔΩ, shown in Figure 3, represents the cumulated effects of each process, and the difference between regions reflects the modification that occurs en route. Waters in the subsurface North Pacific are preconditioned by global-scale circulation history and biological activity has lowered Ω in subsurface water by up to 1.6 (Figure 3a). This signal is then carried into Bering Sea, Canada Basin, CAA, and Baffin Bay. The large effect of biological activity on the Bering Sea shelf and in Baffin Bay, compared to values in upstream regions, indicates a significant influence of in situ remineralization. Although ΔΩBio was set to 0 for waters shallower than 20 m and thus underestimates the effect of biological activity in this layer, a localized increase in Ω was found between 20 to 50 m due to photosynthesis (Figure 3a). High ΔΩBio values are derived from largely negative AOU (oxygen oversaturation) in these waters. The depth of this subsurface photosynthesis signal coincides with the top of the nutricline and subsurface Ω maximum in the Canada Basin [McLaughlin and Carmack, 2010].
 When compared to a reference temperature of 3.5°C, cooling of Pacific-origin waters in and along the way to the Arctic can lower Ω by 0.1–0.3 (Figure 3b). For example, the distribution of ΩNoTemp (not shown) shows that most of the waters in the Canada Basin would be Ω > 1 if there was no cooling, even in the presence of other processes. In the Arctic, cooling accompanies sea ice formation and brine rejection which removes freshwater from seawater. This process is reflected in the slight increase of Ω in subsurface layers (Figure 3c). On the other hand, melting of sea ice lowers Ω by ∼0.2 at the surface. This large influence is the result of recent extensive input of sea ice meltwater [Yamamoto- Kawai et al., 2008, 2009b, 2011]. Melting of sea ice further lowers surface Ω by another 0.2 through enhancing air-sea gas exchange (reduce ΔCDiseq close to 0) [Cai et al., 2010; Yamamoto-Kawai et al., 2011], a process which is not examined in the present study. Meteoric water also contributes to the lowering of Ω in the Canada Basin by ∼0.2 at the surface (Figure 3d).
 The uptake of anthropogenic CO2 has lowered Ω by >0.4 in surface waters of the North Pacific, Bering Sea, and Labrador Sea (Figure 3e). In the Canada Basin, CAA, and Baffin Bay, changes in Ω due only to the uptake of anthropogenic CO2 are smaller, owing to a higher Revelle factor in the colder surface waters in these regions [Sabine et al., 2004]. Deep winter convection in Labrador Sea delivers the anthropogenic CO2 signal to deeper layers (Figure 3e). Comparison of Figures 1c and 4 shows that the aragonite saturation horizon has shoaled by 50 m in the North Pacific due to the input of anthropogenic CO2, as pointed out by previous studies [Feely and Chen, 1982; Feely et al., 2008]. This shoaling results in the upwelling of “corrosive” subsurface water over the North American continental shelf and occurs more widely and frequently than before and will influence the ecosystem and commercial fisheries [Feely et al., 2008; Gruber et al., 2012; Barton et al., 2012]. In the absence of an anthropogenic CO2 input, surface and subsurface waters of Canada Basin and CAA are oversaturated with respect to aragonite except at a few localized undersaturation sites (Figure 4). The effect of anthropogenic CO2 on Ω is ∼0.3 in Canada Basin surface water. Yamamoto-Kawai et al.  estimated that Ω has decreased by 0.6 from preindustrial times to 2008 in the surface Canada Basin: of this decrease 0.3 was due to the increase in atmospheric CO2 concentration; 0.2 was due to dilution by sea ice meltwater; 0.2 was due to enhanced air-sea CO2 exchange; and a compensating increase of Ω by 0.1 was due to warming [Yamamoto-Kawai et al., 2011]. As ΩNoAnth in the present study was calculated with parameters S, T, and TA, and ΔCDiseq observed in 2008, Figure 3e shows the effect of the increase in atmospheric CO2 concentration when all the other conditions remains the same as found in preindustrial time. In other words, Figure 4 shows Ω without the input of anthropogenic CO2 but includes the effects of melting of sea ice and warming. The salinity of undersaturated surface waters shown in Figure 4 is <25, indicating dilution by sea ice melt. Anderson et al.  also estimated Ω in Canada Basin in preindustrial times using data observed in 2005, prior to the massive melting of sea ice and surface freshening that occurred in 2006–2007 [Yamamoto-Kawai et al., 2009a], and showed that surface waters were oversaturated with respect to aragonite in preindustrial period. They also estimated that PWW in most places in the Canada Basin will experience Ω undersaturation by the year 2050, even for calcite. Undersaturation of surface and subsurface waters in the Canada Basin impact the downstream regions of the CAA, Baffin Bay, and Labrador Sea, where local freshening and remineralization further modify Ω at the surface and at depth, respectively.
7. Summary and Conclusions
 We examined the distribution of Ω along a section following the global freshwater pathway from the North Pacific, through the Arctic and into the North Atlantic. Substantial modifications occur en route owing to regional physical and biogeochemical processes. Specifically, waters in the North Pacific are preconditioned by global-scale circulation history and remineralization to be high nutrient and low Ω source waters and these waters subsequently upwell onto the Bering Sea shelf. As these waters move northward and cross the highly productive Bering and Chukchi seas, they are further modified by photosynthesis and remineralization which increase Ω at the surface but reduce Ω at the bottom. These Pacific-origin waters then enter the Canada Basin, both as PSW and PWW varieties, where the former is modified locally by photosynthesis at the top of the nutricline and the latter is modified by cooling in and along the way to the Canada Basin. In the Canada Basin, there is a globally unique Ω profile. Surface waters are undersaturated due to inputs of fresh river water and the recent melting of sea ice and this could negatively impact planktonic calcifiers [Comeau et al., 2012]. The undersaturation of the underlying PWW is a result of cooling and remineralization on upstream shelves and the input of anthropogenic CO2. Undersaturated PWW abuts the upper slope of the Beaufort Sea shelf-break where upwelling favorable winds move it onto and across the shelf where it can affect benthic organisms as observed by Mathis et al. . Such exchange is likely to increase owing to the increased retreat of summer ice cover and exposure of the shelf-break to wind forcing over the past decade [Carmack and Chapman, 2003; Schulze and Pickart, 2012]. Surface and subsurface waters exit the Canada Basin and continue eastward through the CAA. Low Ω is found in surface waters in areas such as Coronation and Queen Maud gulfs where dilution due to local river inputs are high. Waters are also undersaturated at depth in the CAA, reflecting the outflow of PWW from the Canada Basin as well as local remineralization. Arctic waters exit the CAA into Baffin Bay where undersaturated low Ω values extend from the surface to depth due to outflow from the Arctic, upwelling within the cyclonic gyre and remineralization of organic matter at depth. A strong Ω front between Pacific/Arctic and Atlantic waters forms above Davis Strait.
 These data show that in order to predict future changes in pH and Ω, one must have a thorough, multiscale understanding of local physical and biogeochemical processes that act on waters which follow advective pathways. Physical processes include changes associated with large-scale advection, cooling, dilution from rivers and or ice melt, gyre circulation (i.e., upwelling or downwelling), and shelf-break upwelling. Biogeochemical processes include changes associated with production, rain out and remineralization of organic matter. Similarly, local impacts of reduced Ω will also be shaped by the biogeography of species which occupy a given region, because the response to low Ω varies largely between species [Ries et al., 2009].
 This research was supported by the Canada International Polar Year Office, by the National Science Foundation Office of Polar Programs (grant OPP-0424864) and by Fisheries and Oceans Canada. We acknowledge Sarah Zimmerman, other scientists and technicians, officers, and crews for their help in sampling and analysis during the cruises of the CCGS Louis S. St-Laurent and CCGS Sir Wilfrid Laurier. We appreciate Marty Davelaar for his careful analysis of DIC and TA samples. Figures were produced using Ocean Data View: R. Schlitzer at http://www.awi-bremerhaven.de/GEO/ODV, 2001.