Water samples were collected during an expedition in August and September of 2014 onboard the Canadian Coast Guard Ship (CCGS) Amundsen. Samples were taken at the stations shown in Fig. 1 in the Beaufort Sea over the entire water column using a rosette system (24 12-L PVC Niskin bottles) equipped with a conductivity-temperature-depth sensor (CTD, Seabird^® SBE 911plus). Water samples for the analysis of DIC and total alkalinity (TA) were drawn directly from the Niskin bottles into 300 mL borosilicate glass bottles into which 20 µL of a supersaturated mercuric chloride (HgCl2) solution was injected to halt biological activity before being sealed with ground-glass stoppers and Apiezon^® Type-M high-vacuum grease. Samples were stored at 4 ∘C in the dark until analysis using a VINDTA 3C (versatile instrument for the determination of titration alkalinity, by Marianda) at Dalhousie University following the methods described in Dickson et al. (2007). The instrument was calibrated against certified reference materials provided by A. G. Dickson (Scripps Institute of Oceanography) and the reproducibility of the DIC and TA measurements was, respectively, better than 2 and 3 µmolkg-1. Water samples for measurements of the carbon isotope ratio of DIC (δ13C-DIC) and the oxygen isotope ratio of water (δ18O-H2O) were collected in tandem with DIC and TA samples. δ13C-DIC samples were analyzed at the Yale Analytical and Stable Isotope Center (Yale University) using a GasBench II system connected to a Thermo Delta Plus XP isotope ratio mass spectrometer (IRMS). δ18O-H2O samples were analyzed on a triple collector IRMS in dual inlet mode at the GEOTOP Stable Isotope Laboratory (Université du Québec à Montréal). The pCO2, pH (on the total scale), and aragonite saturation state (ΩAr) were computed using DIC and TA as input parameters with the standard set of carbonate system equations, excluding nutrients, using the CO2SYS program of Lewis and Wallace (1998) and the carbonic acid dissociation constants of Mehrbach et al. (1973) refit by Dickson and Millero (1987).

The distribution and vertical structure of water masses in the Arctic Ocean are well known and documented (e.g., Jones and Anderson, 1986; Aagaard, 1989; Swift et al., 1997; Jones et al., 1998). The variable inputs of freshwater from rivers and sea-ice melt have a strong impact on the local stratification and circulation (e.g., Aagaard and Carmack, 1989; Rabe et al., 2011; McClelland et al., 2012). In this study, the vertical water column is divided into three major water masses. At the surface is the polar mixed layer (PML), a mixture of meteoric/river water (MW), sea-ice melt (SIM), and upper halocline Pacific-origin water. The UHL originates in the Pacific Ocean and lies below the PML. It covers a depth range of approximately 25 to 200 m, a salinity range of 31.6 to 34.6 (with a core salinity of 33.1), and is characterized by a temperature minimum. Below the UHL is the Atlantic layer (ATL), covering the depths below 150 m to approximately 1000 m, with a salinity range of 34.6 to 34.9. Waters with a mixture of UHL and ATL properties, or without at least 80 % of the total fraction of one of these water masses, are defined as UHL-ATL waters. This described layering of water masses is applicable over the deep basin in the study area and disappears somewhat over the shelf as mixing and upwelling and/or downwelling dynamics occur.

Using a three endmember mixing scheme allowing for negative values, assuming conservative mixing, and knowing the TA and δ18O-H2O of each water mass, the relative fraction of each water mass in a sample of known TA and δ18O can be calculated. This technique has been applied successfully in multiple independent studies in the same geographic area (e.g., Yamamoto-Kawai et al., 2009; Shadwick et al., 2011b). The upper 150 m of the water column is assumed to be a mixture of MW, SIM, and UHL, and the following system of linear equations is solved to estimate the relative fraction of each water mass: fMW+fSIM+fUHL=1,TAMWfMW+TASIMfSIM+TAUHLfUHL=TA,δ18OMWfMW+δ18OSIMfSIM+δ18OUHLfUHL=δ18O, where fx represents the relative fraction of each water mass. Below 150 m, the water is assumed to be a mixture of net SIM, UHL, and ATL, and the following system of equations is used: fSIM+fUHL+fATL=1,TASIMfSIM+TAUHLfUHL+TAATLfATL=TA,δ18OSIMfSIM+δ18OUHLfUHL+δ18OATLfATL=δ18O. The endmember values used for each water mass are shown in Table 1. The TA value for MW is a flow-weighted value of the Mackenzie River water taken from Cooper et al. (2008). The δ18O value for MW is a flux-weighted value for the Mackenzie River taken from Yi et al. (2012). These endmembers give a balanced value for the river water endmember throughout the year. Although Alaskan and Eurasian rivers may have some influence on the upper water column in this area and contribute water with different properties, the dominant input of MW to the study area is the Mackenzie River and so these values are used. The DIC value for MW is taken from Shadwick et al. (2011b). The δ18O value for SIM is taken from Yamamoto-Kawai et al. (2009) and the TA, DIC, and salinity values for SIM are taken from Lansard et al. (2012). The δ18O, TA, and DIC values for the UHL are the average values of all water samples at a salinity of 33.1, considered to be the core of the UHL water mass. Likewise, ATL values are averages of all values at the deep temperature maximum and a salinity of 34.8.

The velocity fields used in this study are taken from the 0.25∘ Arctic and Northern Hemisphere Atlantic (ANHA4) configuration. This is a regional configuration of a coupled ocean–sea ice model, based on the Nucleus for European Modelling of the Ocean (NEMO, version 3.4) framework (Madec, 2008). The sea-ice model used is the Louvain-La-Neuve Sea-Ice Model version 2 (LIM2) with an elastic-viscous-plastic rheology, including both thermodynamic and dynamic modules (Fichefet and Maqueda, 1997). The horizontal mesh is a subdomain of the global ORCA025 tripolar grid with two open boundaries, one close to the Bering Strait in the Pacific Ocean and the other at 20∘ S in the Atlantic Ocean. In the vertical, there are 50 unequally spaced geopotential levels with higher resolution (∼1 m) in the top 10 m.

This simulation is integrated from January 2002 to December 2014. The initial conditions (3-D temperature, salinity, horizontal velocities, sea surface height, and sea ice) are extracted from the Global Ocean Reanalysis and Simulations (GLORYS2v3) produced by Mercator Ocean (Masina et al., 2015). Hourly 33 km horizontal resolution atmospheric forcing data (10 m wind, 2 m air temperature and humidity, downwelling and longwave radiation fluxes, and total precipitation) from the Canadian Meteorological Centre (CMC) global deterministic prediction system (GDPS) reforecasts (CGRF) dataset (Smith et al., 2014) are used to drive the model. The GLORYS2 dataset is also used to provide the open boundaries (temperature, salinity and, ocean velocities). Monthly interannual runoff from Dai et al. (2009) as well as Greenland meltwater provided by Bamber et al. (2012) are also carefully remapped onto the model grid to give a more realistic freshwater input from the land to the ocean.

Model output is 5-day averages of the velocity fields in the study area for August and September 2014. To look at along-shore and cross-shelf transports on the Mackenzie Shelf, the velocity was gridded along the shelf-to-basin sampled transect (Fig. 1) every 5 km, starting from 20 km closer to shore than the first sampled station and ending ∼ 20 km into the basin beyond the last sampled station. The along-shore direction runs parallel to the shelf-break, at a bearing of 52∘ to the east of true north (52 or 232∘ T, where T indicates relative to true north). The cross-shelf direction is then 38∘ to the west of true north (142 or 322∘ T) (directions indicated in Fig. 1). The ocean current vectors (u and v components) from the ANHA4 simulation were tilted to correspond to these along-shore and cross-shelf bearings, resulting in the u component of the velocity representing along-shore flow and the v component representing cross-shelf flow. In all cases, positive (negative) values indicate along-shore flow to the east (west) and cross-shelf flow in the off-shelf (on-shelf) direction.

Uncertainty was estimated using Monte Carlo simulations for the calculation of cross-shelf mass transports. The inputs for the simulations were randomly generated from normal distributions, and 1000 points were chosen randomly for each of the variables. Standard deviation for salinity (±0.0003) and temperature (±0.001 ∘C) were taken from error calculations with the Seabird 911plus CTD sensor, and the standard deviation for DIC measurements (±2 µmolkg-1) was taken from repeat measurements of sample and standard seawater using the VINDTA 3C. The error of the modelled velocity was estimated by taking the standard deviation of the velocity at each station and depth layer from the ANHA4 model over the 2-month study period.

High variability in the carbonate system properties of surface waters is observed throughout the study region with marked differences between the basin and distinct shelf regions (Fig. 2). Temperature varies from less than -1 ∘C in the central basin and beyond the shelf-break to greater than 6 ∘C on the Mackenzie Shelf and in the the Amundsen Gulf. Salinity ranges from less than 26 to greater than 30, with the lowest values in areas close to the Mackenzie River and in the central basin. Higher salinity values in the Amundsen Gulf and on the Mackenzie Shelf suggest upwelling of high-salinity deep water. DIC is lowest in the low-salinity and low-temperature surface waters of the central basin and higher in shelf seas, possibly a result of upwelling. TA follows a pattern similar to that of salinity with low values beyond the shelf-break and in the central basin, and higher values over the shelf. High pCO2 values are found in waters with higher temperature and DIC content, primarily over the shelf and in the Amundsen Gulf. Lower δ13C-DIC values are observed over the shelf in areas near the Mackenzie River outflow, as river-water values are presumably depleted due to accumulation of metabolic CO2 from the microbial respiration of terrigenous or riverine organic matter, while high δ13C-DIC values are found in the northern Amundsen Gulf.

These surface plots (Fig. 2) are constructed from surface samples taken over more than a month-long period, from 17 August to 21 September. The stations along each of the three transects (the Amundsen Gulf, Mackenzie Shelf, and Mackenzie Trough) were all sampled within a 3-day span, except for the furthest off-shore station on the Mackenzie Trough transect. Temporal variability between the three transects and the stations in the deep basin must be considered when looking at the spatial variability between these areas. Changes in wind and circulation that occurred during the sampling period may alter the surface conditions due to upwelling or downwelling (where the Mackenzie River discharge is directed) and concentrations of sea ice in different areas.

Relationships between DIC, TA, salinity, temperature, pCO2, and ΩAr for all samples on the Mackenzie Shelf, in the Amundsen Gulf, and in the Canada Basin are shown in Fig. 3. DIC and TA are both quite variable in the low-salinity PML with freshwater from sea-ice melt and river outflow contributing waters of different chemical properties (Fig. 3a and b). DIC ranges widely from 1822 to 2156 µmolkg-1 in the PML and TA ranges from 1892 to 2269 µmolkg-1 over this same wide salinity range of 25.8 to 32.0. The DIC reaches a maximum of ∼ 2225 µmolkg-1 in samples characterized by the highest relative contributions of UHL, a common feature in this study area. For example, Shadwick et al. (2011b) found a DIC maximum of 2240 µmolkg-1 at the UHL maximum, Brown et al. (2016) found a DIC maximum of 2225±8 µmolkg-1, and Lansard et al. (2012) found a DIC maximum of 2228±8 µmolkg-1 in this same water mass. TA shows a maximum of ∼ 2360 µmolkg-1 in waters with the greatest ATL contribution and a salinity maximum of ∼ 34.9. These parameters follow the trends seen in previous years in the same sampling areas (e.g., Shadwick et al., 2011b). A linear regression was performed to extrapolate to the values of DIC and TA at a salinity of zero, representative of the Mackenzie River water endmember (regression lines shown in Fig. 3a and b). Only samples with greater than 10 % fMW and less than 5 % fSIM were used for the regression. The resulting endmember values of DIC and TA for the Mackenzie River are 1441 and 1564 µmolkg-1, respectively. These values are close to the Mackenzie River properties published in Shadwick et al. (2011b) and Cooper et al. (2008) and used for the water mass deconvolution in this study (Table 1).

The relationship between DIC and TA is shown in Fig. 3c. For samples in the PML, the slope is < 1, with TA increasing at a faster rate than DIC. TA is nearly conservative while DIC is more responsive to biological processes, including photosynthesis and respiration in the PML. The slope of this relationship increases in the UHL to > 1, as DIC builds up in the mid-depth layer to its maximum value due to the accumulation of remineralization products. A greater DIC : TA ratio indicates that there is a greater amount of DIC in the water relative to TA, altering the chemistry and leading to an increase in pCO2. Values of the DIC : TA ratio for the core or average values of each of the water masses defined here are shown in Table 1. The value of pCO2 is variable in the PML, ranging from ∼ 200 to 500 µatm, but mostly staying at values below 400 µatm (Fig. 3d). Temperatures range from -2 to 7 ∘C in the PML, with resulting variability in pCO2 and ΩAr (Fig. 3e and f). The UHL is where pCO2 reaches its maximum, with values greater than 700 µatm in water associated with the DIC maximum. The UHL is also the layer where the temperature minimum is found (Fig. 3e and f). The high values of pCO2 and low ΩAr found in the UHL are restricted to a small temperature range. Waters return to the 200 to 500 µatm range in the ATL layer, with a few samples showing higher values.

Five stations were sampled along an onshore-offshore transect across the Mackenzie Shelf and beyond the shelf-break from 22 to 24 August (identified in Fig. 1). Water properties and carbonate system parameters in the top 300 m are shown in Fig. 4. Water temperature is warmest at the surface over the shelf, reaching values of > 6.5 ∘C. Surface waters over the deep basin are cooler, with a maximum temperature of < 1 ∘C. Salinity, DIC, and TA all show similar patterns in the top 300 m over the shelf and beyond the shelf-break. The lowest values (< 26.5, 1830 µmolkg-1, and 1920 µmolkg-1 for salinity, DIC, and TA, respectively) are seen in the surface water beyond the shelf-break. Salinity and TA increase down to 300 m, reaching maximum values of ∼ 34.8 and ∼ 2300 µmolkg-1, respectively. DIC has a mid-depth maximum at around 125 m depth of ∼ 2225 µmolkg-1, then decreases slightly with depth down to 300 m. All three of these parameters show isoclines sloping up onto the shelf, with the same values at shallower depths over the shelf as well as elevated surface values over the shelf.

Low pCO2 values are observed in the surface layer along the transect, ranging from 278 to 339 µatm. Beyond the shelf-break, pCO2 increases with depth to a maximum value of 749 µatm at around 125 m depth in the UHL (Fig. 4e). Below this depth, the pCO2 decreases gradually. Maximum values on the shelf of up to 665 µatm are found in the bottom layer (53 m depth). Measurements of δ13C-DIC show the expected pattern, with highest values of up to 1.50 ‰ in the surface or subsurface, due to photosynthesis and the preferential uptake of the lighter carbon isotope, and a decrease to minimum values of -0.5 ‰ in the DIC maximum layer in the deep basin and on the shelf bottom due to respiration of 13C-depleted organic matter. Maximum values of pCO2 and minimum values of δ13C-DIC are coincident on the shelf bottom and at depths of around 125 m beyond the shelf-break. The temperature effect on pCO2 is not overly important at these depths, as large variations in temperatures are most prominent in the top 30 m (Fig. 4a). Variations in pCO2 at these greater depths can be attributed to changes in DIC content (metabolic CO2 addition), as indicated by the depleted δ13C-DIC values at the same locations. The saturation state of the waters with respect to aragonite (ΩAr) and pH are also closely related to the DIC concentration and pCO2 levels, with the lowest values being observed along the shelf bottom and in the DIC maximum beyond the shelf-break (Fig. 4g and h). These parameters and the consequences for ocean acidification are discussed further in Sect. 4.4.

Surface water composition varies greatly over the study region with large freshwater input from the Mackenzie River as well as significant sea-ice melt in the central Canada Basin (Fig. 5a and b). The strong impact of the Mackenzie River is seen easily, with fractions of meteoric water greater than 20 % observed over the shelf. The river plume is detectable (> 10 %) in the surface waters to the east into the Amundsen Gulf and far into the central Canada Basin. Sea-ice melt has the strongest influence in the central basin with fractions of ∼ 15 %, and substantial fractions of up to 10 % also seen near the shelf-break.

The fraction of water-mass types along the sampled transect on the Mackenzie Shelf are shown in Fig. 5c–f. High meteoric water content is seen over the shallow shelf region, as well as in the surface layer out into the deep basin. There is a notable intrusion of sea-ice melt water from the deep basin onto the shelf-break on the Mackenzie Shelf. This sea ice may be transported south on the eastern arm of the Beaufort Gyre and pushed onto the shelf. The UHL dominates at mid-depth, around 100 m, off the shelf, with large fractions (> 95 %) pushed up to shallower depths along the bottom of the shelf. Atlantic water is restricted to the deep layer beyond the shelf-break and did not intrude up onto the shallow shelf during our sampling period.

Wind data were taken from the gridded reforecast data of the Canadian Meteorological Centre (Smith et al., 2014). The location selected is on the Mackenzie Shelf, near the middle of the transect located at 70.3125∘ N, 133.5937∘ W (Fig. 1). Figure 6 shows the wind direction and magnitude averaged every 6 hours during August and September of 2014. Oscillations between strong northerly and southerly winds are a prevalent feature. From 6 to 16 August, the wind field was dominated by northeasterly winds with magnitudes of > 20 m s-1. Between 26 August and 10 September, periods of strong southerly winds were interrupted by short periods of strong northerly winds with maximum velocities of > 50 m s-1 in both directions. The 14 to 16 and the 26 to 30 September were periods of sustained southeasterly winds reaching speeds of up to 40 m s-1. Sampling of the Mackenzie Shelf transect took place from 22 to 24 August, indicated by the red lines in Fig. 6. During that time, the wind direction was variable and wind speeds were comparably low. This sampling period followed both northeasterly and southeasterly winds throughout the beginning of the month.

The modelled velocity, averaged in the top 10 m of the water column, on the Mackenzie Shelf is shown in Fig. 7a and b for two time periods displaying the two opposite modes of circulation observed during the study period in the fall of 2014. During the period from 16 to 20 August, surface water flows to the west along the Mackenzie Shelf. From 5 to 9 September, mean surface flow is towards the east. In the middle of the shelf, cross-shelf transport of surface water is evident at both times, with off-shelf flow from 16 to 20 August and on-shelf flow from 5 to 9 September. In both cases, the Cape Bathurst topography induces intensified surface currents in the mean direction of flow. Surface flow is also altered at the Mackenzie Trough, as topography becomes complex and the influence of the Mackenzie River is greatest. The modelled velocity at the 56 m depth horizon shows the strong influence of the shelf-break jet from both 16 to 20 August (Fig. 7c) going to the west and from 5 to 9 September (Fig. 7d) to the east. The water flow closely follows the topography of the shelf in both cases, shown clearly at the Mackenzie Trough, but also evident with cross-shelf movement through the middle of the sampled transect, most likely due to bathymetric changes at the Kugmallit Valley, a previously documented location of enhanced cross-shelf transport (Williams et al., 2008). Cross-shelf movement is also seen at several other locations along the shelf-break. From 16 to 20 August, the modelled velocity at the 92 m depth horizon shows an accelerated shelf-break jet (Fig. 7e). The velocity from 5 to 9 September at the 92 m depth horizon (Fig. 7f) shows the eastward flowing shelf-break jet of similar velocity as at 56 m depth.

Two opposite modes of circulation are shown to occur on the Mackenzie Shelf during the study period. From 16 to 20 August, the shelf-break jet flows to the west, whereas from 5 to 9 September it flows to the east. In the former case, it generates conditions conducive to upwelling along the shelf and produces an estuarine circulation, with relatively fresher water in the surface layer pushed off-shelf towards the basin and water at depth being brought onto the shelf. In the latter case, downwelling could take place, producing an anti-estuarine circulation, with water at the surface moving shoreward and denser water at depth moving out past the shelf-break. These two periods are investigated further to look at the upwelling and downwelling circulation in the fall of 2014.

The 2-month average, modelled velocities of the along-shore and cross-shelf flow for the Mackenzie Shelf transect are shown in Fig. 8a and b. The surface flow over the shelf is dominated by eastward along-shore flow. The westward flow of the Beaufort Gyre beyond the shelf-break is evident. The cross-shelf flow is primarily in the on-shelf direction over the shallow shelf and off-shelf beyond the shelf-break. Due to Ekman transport, westward flow towards the Alaskan Shelf is associated with upwelling onto the shelf. This situation is seen from 16 to 20 August with strong westward along-shore transport and velocities on the shelf and in the core of the shelf-break of > 0.15 m s-1 (Fig. 8c). This westward transport throughout the water column coincides with water velocities in the cross-shelf direction indicative of upwelling (Fig. 8d). Surface water over the shelf reaches velocities of > 0.08 m s-1 in the off-shelf direction in the top 5 m and flow remains in an off-shelf direction for much of the top 15 m of the water column. Below this surface layer, water flow over the shelf is in an on-shelf direction, reaching velocities of > 0.10 m s-1 at mid-depth (∼ 50 m). This upwelling flow is readily noted, with water at depth being moved onto the shelf and water at the surface moving out toward the shelf-break. Flow from the west towards the Amundsen Gulf is associated with downwelling transport off the shelf, a situation seen from 5 to 9 September (Fig. 8e and f). Along-shore velocity over the shelf and shelf-break is high, dominated by flow in the eastward direction over the entire shelf area, reaching values of > 0.16 m s-1 at the surface and decreasing with depth. The westward flow in the deep basin decreases and is pushed further beyond the shelf-break relative to the average along-shore pattern. The strong eastward flow over the shelf and shelf-break is associated with a small on-shelf transport at the surface (> 0.09 m s-1) and an off-shelf transport along the shelf bottom reaching velocities of > 0.03 m s-1. This cross-shelf velocity pattern shows a downwelling loop, with water at the surface pushed shoreward and water at depth moving out towards the shelf-break.

These modelled velocities are representative of 5-day averages for the region and thus do not show the peak velocities reached during the study period. The shelf-break jet or boundary current has been measured throughout the year at various locations along the Alaskan and Canadian Beaufort shelves. For example, Pickart et al. (2009) reported a bottom-intensified shelf-break jet flowing at ∼ 0.15 m s-1 to the east over a year-long mean across the shelf at 152∘ W, as similarly reported by Nikolopoulos et al. (2009). This value is the same as the maximum modelled velocities found in the core of the shelf-break jet through the sampled transect. Reversals of the shelf-break jet due to variations in wind forcing have been reported in numerous studies along the western and eastern Beaufort Shelf (e.g., Nikolopoulos et al., 2009; Pickart et al., 2013b; Dmitrenko et al., 2016). Measurements of velocity from moorings anchored in the same region across the Mackenzie Shelf slope, such as those analyzed in Forest et al. (2015) from September 2009 to August 2012, show current surges of 0.20 to 0.80 m s-1 with frequent oscillations in current direction. The mean velocities during this period at these moorings were 0.08 to 0.14 m s-1. These mean velocities are comparable in magnitude to velocities produced by the model for 5-day averages. It is likely that similarly strong surges in current velocity took place throughout the study period of August and September 2014 on the Mackenzie Shelf.

Along-shore water transport across the Mackenzie Shelf transect throughout the study period was estimated in the top 200 m of the water column through the 120 km (in the cross-shelf direction) section identified in Fig. 1 (Fig. 9a). Transport was calculated by taking the product of the depth-averaged modelled velocity and the depth and cross-shelf area between stations. Positive values indicate eastward flow, whereas negative values indicate westward flow. Maximum westward transport of 0.74 Sv through the Mackenzie Shelf transect took place from 16 to 20 August; maximum eastward transport of 0.89 Sv through the transect occurred from 5 to 9 September. In comparison, the transport of northward flowing water through the Bering Strait is approximately 1.1 Sv (Woodgate et al., 2012). The along-shelf transport through these two transects and the changes in transport direction are substantial.

In August and September of 2014, the circulation along-shore changed from a strong flow to the west to a strong flow to the east and back again (Fig. 9). Correlation between along-shore wind in the northeasterly (southwesterly) direction and upwelling (downwelling) on the Mackenzie Shelf has been shown in numerous studies and models (e.g., Carmack and Kulikov, 1998; Yang et al., 2006; Mathis et al., 2012). In our study, the two largest upwelling episodes on the Mackenzie Shelf, 11 to 20 August and 15 to 24 September, were preceded by periods of strong (> 20 m s-1) and sustained northeasterly winds (Fig. 6). The largest downwelling episodes, from 31 August to 9 September, appear to be triggered by strong southwesterly winds (> 40 m s-1) and strong wind reversals throughout.

The oscillating along-shore water transport is dominated by changes over the slope, as evidenced by the transport calculated at the specific sampled station locations. Along-shore transport at the stations on the transect are calculated through the water column, to a maximum depth of 200 m, by taking the sum of the products of the along-shore modelled velocity and a cross-shelf distance of 1 km (Fig. 9b). Bottom depth was taken from the shipboard depth sounder when sampling took place. All three stations on the shelf (434, 432, and 428) as well as station 435 on the slope show the same oscillating change in along-shore transport. Station 421, located beyond the slope in the deep Canada Basin shows a negative total along-shore transport throughout the study period. This station is under the influence of the Beaufort Gyre, which is consistently flowing in a westward direction at this location. During the 2-month study period there is an oscillation between two dominant flow patterns represented by positive and negative total transport in the along-shore direction.

Cross-shelf flux calculations were carried out for the Mackenzie Shelf transect to investigate the exchange of DIC and TA between the shelf and the deep basin. Fluxes were calculated by taking the product of linearly gridded DIC or TA measurements and cross-shelf velocity values from the ANHA4 simulation through the water column with an along-shore distance of 10 km. The DIC flux (mol C d-1 m-2) at the shelf stations for the period of the largest on-shelf flux are shown in Fig. 10. From 16 to 20 August, during the period of greatest westward along-shore transport, the cross-shelf DIC flux is off-shore at all stations in the top 10 to 25 m of the water column, reaching values of 16.7×103 mol C d-1 m-2 (TA flux: 18.1×103 mol TA d-1 m-2) at station 432. The DIC flux is on-shelf below the surface layer, with stations 434 and 432 having maximum on-shelf fluxes at mid-depth. At station 428, the on-shelf flux reaches values of 16.9×103 mol C d-1 m-2 (TA flux: 17.4×103 mol TA d-1 m-2) near the bottom of the water column. This on-shelf flux near the ocean floor transports water sourced from the UHL, bringing metabolite-rich water and DIC onto the shelf.

The excess DIC brought onto the shelf and its impact on the pCO2 and pH depends on the amount and composition of the source-water. The DIC : TA ratio in the water being moved along the bottom of the shelf is a parameter that provides insight into the carbonate system properties of the water mass. At stations 434, 432, and 428 on the shelf, the DIC : TA ratio changes significantly with depth. Water being upwelled (Fig. 10; negative DIC flux) has DIC : TA ratios greater than 0.975 at all three stations. In other words, excess DIC is being upwelled, contributing to increased pCO2 and decreased pH and ΩAr levels of the shelf bottom waters. The DIC : TA ratio in the latter is higher than the ratio in the surface waters (0.92 to 0.93) being moved out towards the Canada Basin. During the time of measurement, surface pCO2 levels along the Mackenzie Shelf transect remained low, ranging from 278 to 339 µatm, allowing for the uptake of CO2 by the surface ocean.

Cross-shelf mass transport of carbon was calculated for the three shelf stations (434, 432, and 428) from 16 to 20 August (Table 2). These transports are integrated through a 10 km along-shore distance to include the area of the shelf influenced by similar water movement. From 16 to 20 August, when the along-shore flow was westward, the cross-shelf transport should have led to an upwelling flow over the shelf. As expected by the cross-shelf velocity in Fig. 8d, transport in the top 10 m at all three stations is off-shelf, reaching a maximum of 14±3×10-3 Tg C d-1. Below the surface layer, the cross-shelf transport is on-shelf, with maximum transport at station 428 of 65±15×10-3 Tg C d-1.

From 6 to 16 August, strong northeasterly winds were present over the shelf region, but had shifted to an easterly direction by the time sampling took place and through 22 August. Although no results from a coupled physical-biogeochemical model were available, it seems probable from the wind and physical sampling data that upwelling on the Mackenzie Shelf began around 6 August and continued through to the 22 August. From the analysis of the hydrodynamic model, there are two periods, lasting longer than 10 days in August and September of 2014 (approximately 10 to 28 August and 14 to 29 September), during which the cross-shelf transport brings water from beyond the shelf-break onto the Mackenzie Shelf. During these episodes, water from the DIC-rich UHL is brought onto the Mackenzie Shelf. The period of off-shelf transport lasts for ∼ 15 days in the region, offsetting the on-shelf flux of these waters and carrying remineralization products (metabolites) back to the deep basin. Nevertheless, the average transport through the 2 months carries water onto the shelf. This long-term upwelling activity acidifies the shelf bottom waters and may impact the air–sea exchange of CO2 due to the large air–sea pCO2 difference, including altering the sink or source status of the shelf. Surface waters with high meteoric water content are delivered beyond the shelf-break into the Beaufort Gyre as upwelling takes place.

The result of upwelling on the shelf is reflected by gradients of DIC, TA, and δ13C-DIC across the shelf (Fig. 11a–d). The on-shore transport of deeper water is evident when comparing the values at the 50 m depth horizon across the three identified stations. The onshore gradients for DIC and TA are -1.98 and -0.87 µmolkg-1 m-1, respectively. The DIC : TA ratio of these gradients is 2.28, meaning that more DIC, relative to TA, is brought onto the shelf. This is important because a higher DIC : TA ratio translates into an increased pCO2 of the water as well as a decrease in the pH and the saturation state of the water with respect to carbonate minerals. If these waters are brought to the surface by upwelling or vertical mixing, the net flux of CO2 over the shelf region may be altered. Outgassing may occur, releasing CO2 to the atmosphere, or the uptake of CO2 by the surface ocean may be suppressed. Although surface pCO2 was undersaturated with respect to the atmosphere when sampling took place, outgassing or a bloom may have occurred before sampling at the height of this upwelling period. The gradient of δ13C-DIC across the shelf is positive (Fig. 11c and d), opposite to that of DIC and TA. The δ13C-DIC values reach minima along the shelf bottom, reflecting the signature of UHL source water and remineralization products on the shelf.

To investigate the delivery of DIC to the shallow shelf, the effects on the source or sink status of the shelf, the pH conditions, the excess DIC (DICex), and excess protons (Hex+) at 50 m depth were calculated relative to those of a parcel of water of identical TA, temperature, and salinity in equilibrium with the atmosphere at a pCO2 of 400 µatm (following the methods of Burt et al., 2013): DICex=DICobs-DIC(pCO2=400),H+ex=Hobs+-H+(pCO2=400). The gradient in DICex is -1.19 µmolkg-1 km-1 and the gradient in Hex+ is -0.09 nmol kg-1 km-1 (Fig. 11e). Both quantities display negative gradients with positive values closer to shore along the shelf bottom at station 432, where DICex reaches a value of 51.4 µmolkg-1 and Hex+ reaches a value of 4.57 nmol kg-1. If this shelf water is upwelled, the surface water pCO2 and pH conditions will be altered, lowering the pH and raising the pCO2 at the surface leading to outgassing. At station 432 there is excess DIC in the water column and an on-shore water movement from 35 to 55 m depth. The total on-shore cross-shelf mass transport of DICex (relative to a pCO2 of 400 µatm) during the upwelling event from 16 to 20 August at the three shelf stations is shown in the last column of Table 2.

Distinguishing the ocean acidification signal in the Arctic is complicated because of the interplay of several changing environmental conditions including warming, sea-ice loss, surface freshening, and changes in primary production (Carmack et al., 2016). During sampling, a layer of high pCO2 and low pH water resided at depth on the shelf (Fig. 4e and h), having been upwelled from beyond the shelf-break onto the shelf. This bottom layer is sourced from the UHL where a DIC maximum consistently resides beyond the shelf-break (Anderson et al., 2010; Yamamoto-Kawai et al., 2013). The UHL source-water is of Pacific origin, and is preconditioned before entering the Arctic to have low ΩAr values (Yamamoto-Kawai et al., 2013). This water is seen in Fig. 3e and f, centred at a salinity of 33.1 and a temperature minimum, as having pCO2 values over 600 µatm and ΩAr as low as 0.83. This water can be corrosive to the mineral skeletons and shells of CaCO3-secreting organisms such as bivalves, mollusks, and echinoderms, with varying impacts on different organisms (Ries et al., 2011). These conditions can also have negative impacts on community structure and recruitment that may carry through to higher trophic levels (Fabry et al., 2008). In our study area, the undersaturated water mass is not at the surface over the shelf, but sustained periods of upwelling could bring this high pCO2 water to the surface and outgas CO2 to the atmosphere (e.g., Mucci et al., 2010; Mathis et al., 2012). An outgassing event following wind-induced upwelling on the Alaskan Beaufort Shelf was documented by Mathis et al. (2012). This was a 10-day event that brought water of ΩAr<1.0 to the surface over the Beaufort Shelf. Likewise, upwelling of elevated pCO2 (522±37 µatm) UHL water at Cape Bathurst was captured by Mucci et al. (2010). In the present study, sampling took place just after a suspected strong upwelling event. Aragonite undersaturated waters with high pCO2 were restricted to depths of greater than 20 m over the shelf at the time of sampling. A strong upwelling wind event could easily have brought this water to the surface and created events like those described by Mucci et al. (2010) and Mathis et al. (2012).