Carbonate saturation state of surface waters in the Ross Sea and Southern Ocean

Introduction Conclusions References

where K sp is the solubility product constant for the specific CaCO 3 mineral and depends on salinity, temperature, and pressure (Mucci, 1983). Aragonite is ∼ 1.6 times more soluble than calcite at 0 • C whereas the solubility of magnesium calcite varies depending on the mole fraction of magnesium ions (Dickson, 2010). Ω Ar represents the saturation state of aragonite and Ω Ca represents the saturation state of calcite. Ω < 1 rep-5 resents undersaturation where dissolution is thermodynamically favorable and Ω > 1 represents supersaturation where precipitation is favorable. Most surface waters of the global oceans are currently supersaturated with respect to CaCO 3 (Feely et al., 2009). However, decreasing CO 2− 3 concentrations can decrease calcification rates even in supersaturated conditions (e.g. Moy et al., 2009;Andersson et al., 2011). 10 The Southern Ocean is especially vulnerable to Ocean Acidification (OA) due to its relatively low total alkalinity (TA) and because of increased CO 2 solubility in cold water. In addition, Antarctic continental shelves have insignificant sedimentary CaCO 3 to buffer against OA (Hauck et al., 2013). Surface waters in the Southern Ocean may start to become undersaturated with respect to aragonite by 2050 and be fully undersatu- 15 rated by 2100 (Orr et al., 2005;Feely et al., 2009). McNeil and Matear (2008) have suggested that wintertime aragonite undersaturation in the Southern Ocean may begin as early as 2030.
OA induced decreases in Ω have potentially serious consequences for Antarctic food webs. In the Ross Sea the aragonitic shelled pteropod Limacina helicina is a dominant stars, and brittle stars (McClintock et al., 2011). Conversely non-calcareous phytoplankton may benefit in the Ross Sea in a high pCO 2 world, especially the larger diatom Chaetoceros lineola (Tortell et al., 2008;Feng et al., 2009).
There are only a few surface carbon system data sets from the Ross Sea (Bates et al., 1998;Sweeney et al., 2000b;Mattsdotter Björk et al., 2014) that can be used to establish baselines in order to understand the relative importance of the physical, chemical, and biological processes that drive the large spatial and seasonal variability of Ω. With no winter Ω measurements, it is challenging to predict when the Ross Sea will become undersaturated with respect to aragonite and calcite. A model by McNeil et al. (2010) suggests that winter surface waters in the Ross Sea will become under- 15 saturated with respect to aragonite by the year 2045 since sea ice, upwelling of deep water, and short residence times prevent these surface waters from reaching equilibrium with the atmosphere. However, McNeil et al. (2010) indirectly estimated surface winter Ω Ar values by using limited carbon system data from the spring (Sweeney et al., 2000b).
Antarctica, accounting for up to half of all primary production over the Antarctic continental shelf (Arrigo and McClain, 1994;Smith and Gordon, 1997;Arrigo and van Dijken, 2003). Photosynthesis reduces the concentration of nutrients and dissolved inorganic carbon (DIC) in the mixed layer, causing Ω to increase in surface waters (McNeil et al., 2010). Once the sea ice reforms during autumn and winter, remineralization of 10 organic matter and deep convective mixing produces a relatively homogeneous water column, causing surface DIC concentrations to increase and Ω to decrease (Gordon et al., 2000;Sweeney et al., 2000b;Petty et al., 2013).
The Southern Ocean is composed of multiple fronts that separate distinct water masses (Rintoul et al., 2001). The most prominent are the Antarctic Polar Front (PF) 15 and the Sub-Antarctic Front (SAF). Circumpolar Deep Water (CDW) upwells south of the PF where it becomes modified into Antarctic surface water (AASW). We define the fronts based on the sea surface temperature (SST) gradient after averaging the underway SST data into 0.25 • bins. Following Dong et al. (2006), the PF is defined as the southernmost location at which the SST gradient exceeds 1.5 × 10 −2 • C km −1 . Follow-Introduction  Fig. 1. Underway TA measurements were conducted using the shipboard uncontaminated continuous flow system with an intake located at ∼ 5 m depth. Seawater from the ship's 10 underway system was redirected to the bottom of a 250 mL free surface interface cup flowing at 2 L min −1 and was drawn from the bottom of the cup for TA analysis without filtration. The entire system was automated and relatively unattended. The sampling cycle was every 24 min on a custom-configured Metrohm 905 Titrando equipped with three Metrohm 800 Dosino syringe pumps (two 50 mL units for sample handling and 15 rinsing and one 5 mL unit for acid titration). Temperature was measured at the cup and in the titration cell. We used certified 0.1 N HCl provided by A. Dickson (Scripps Institution of Oceanography) for the potentiometric titrations and TA calculations follow Dickson et al. (2003). Since we consumed the certified HCl after ∼ 1000 measurements in the Ross Sea, we mixed our own 0. (2) the difference between the measurement and mean was greater than 6 µmol kg −1 .
A total of 65 measurements (out of 1716) were rejected. Surface pCO 2 measurements were made every 3 min using the LDEO air-sea equilibrator permanently installed on the NBP (data available at http://www.ldeo.columbia. edu/res/pi/CO2). The estimated precision is ±1.5 µatm. 5 In order to evaluate the controls of seasonal surface Ω Ar variability in the Ross Sea, we made discrete water column TA and DIC measurements at 85 stations using Niskin bottles attached to a 24 bottle rosette from 13 February through 18 March 2013 (Fig. 1a). We collected samples for TA and DIC following the protocols of Dickson et al. (2007) and immediately added saturated mercuric chloride (< 0.1 % by volume).
For TA, we ran each sample within 12 h of collection using a second potentiometric titrator, a Metrohm 855 Robotic Titrosampler equipped with two 800 Metrohm Dosino syringe pumps (one 50 mL unit for rinsing and sample handling and one 5 mL unit for acid titration). The samples were prefiltered through 0.45 µm polyvinylidene fluoride filters and the estimated precision based on the CRMs (n = 108) is ±1.5 µmol kg −1 . 15 We measured DIC on hydrocast samples within ∼ 4 h of collection without filtration. We acidified 1.25 mL of the sample using a custom built injection system coupled to an infrared gas analyzer (LI-COR LI7000). As described by Long et al. (2011), the infrared absorption signal vs. time is integrated for each stripped gas sample to yield a total mass of CO 2 . Samples were run in triplicate or greater and were calibrated 20 using CRMs between every 3-4 unknowns. Micro-bubbles regularly appeared within injected samples due to sample warming between acquisition and DIC analysis. Each integration curve was visually inspected and integration curves that exhibited evidence for bubbles were rejected. The estimated precision based upon unknowns (> 3500 runs) and CRM replicates (n = 855) for cruise NPB-1302 is ±3 µmol kg −1 . Introduction

Carbon system calculations and crosschecks
We calculate Ω and DIC (hereafter called DIC calc ) with CO2SYS for MATLAB (Lewis and Wallace, 1998;van Heuven et al., 2009) with TA, pCO 2 , SST, and salinity as input variables. Calculations are only conducted for pCO 2 measured within 3 min of the TA measurement (n = 1034), the average cycle time for the automated pCO 2 measure-5 ments. We use the equilibrium constants of Mehrback et al. (1973) as refit by Dickson and Millero (1987) since previous studies have found that they are the optimal choice, including for Antarctic waters (e.g. McNeil et al., 2007;Lee et al., 2000;Millero et al., 2002). For the hydrocast data, we calculate Ω using TA, DIC, temperature, and salinity as input variables.
As a means of internal quality control, we use the initial pH reading from the TA titration as a third carbon system parameter to crosscheck the accuracy of our Ω Ar estimates. In terms of consistency, Ω Ar calculated using TA and pCO 2 is 0.02 ± 0.07 greater than Ω Ar calculated using TA and pH. In addition, DIC calc using TA and pCO 2 is 2 ± 7 µmol kg −1 lower than DIC calc using TA and pH. Finally, measured pCO 2 is 15 4 ± 14 µatm lower than pCO 2 calculated from TA and pH. These strong consistencies suggest that our pCO 2 and TA measurements are accurate. Our surface TA and DIC calc measurements vs. latitude for the Southern Ocean are within the ranges of other studies (Mattsdotter Björk et al., 2014;McNeil et al., 2007;Metzl et al., 2006). We compare the TA measurements from the surface hydrocasts (< 5 m deep) to the 20 underway TA measurements made while the ship was still on station within ∼ 15 min of when the surface samples were collected. The underway values are 3 ± 5 µmol kg −1 higher than the hydrocast TA values.

Ross Sea and Southern Ocean calculations
The Ω Ar of surface waters in the Ross Sea increases during the austral summer months 25 (McNeil et al., 2010). We use DIC, TA, SST, and salinity to determine the controls on the seasonal cycle of surface water Ω Ar . We normalize DIC and TA to a salinity summer and early autumn, observations show that sDIC and sTA concentrations are relatively uniform below 200 m across space and a given season (Table 1).
Following Hauri et al. (2013), the change in Ω Ar of surface hydrocast samples (upper 10 m) from winter conditions can be expressed as: ∆ sDIC and ∆ sTA are the difference in sDIC and sTA for each sample from the winter value. The term ∆T is calculated using a winter SST of −1.89 • C (per Sweeney, 2003).
∆S Ω represents the total contribution of salinity changes to ∆Ω Ar . 15 Since salinity between 200 to 400 m is variable across the Ross Sea (Orsi and Wiederwohl, 2009), ∆S is calculated as the difference between the salinity of a surface sample and the average salinity for samples from that station that are between 200-400 m.
∆ DIC s and ∆ TA s represent changes to DIC and TA due to dilution/concentration 20 from freshwater input and sea-ice processes: The partial derivatives quantify the change in Ω Ar per unit change in DIC, TA, temperature, and salinity respectively. To determine the partial derivatives, we calculate Ω Ar for all hydrocast samples within the upper 10 m using DIC, TA, temperature, and salinity as input parameters. We recalculate Ω Ar after independently increasing DIC, TA, temperature, and salinity by one unit. The partial derivatives are the average difference 5 between the initial Ω Ar and the recalculated Ω Ar .
We use the same equations to evaluate the relative importance of DIC, TA, temperature, and salinity on the variability of Ω Ar from 75 to 55 • S. For the ∆ terms, we calculate the change in sDIC, sTA, temperature, and salinity from the mean of the first 6 underway measurements at 75 • S. For Eqs. (4) and (5), instead of using DIC, TA, and salinity 10 values from 200-400 m, we use the mean of the first 6 underway measurements at 75 • S.

Ω in the Ross Sea
We define the area west of 171 • E as the western region and the area between 171 15 and 180 • E as the central region. This demarcation is similar to regions defined by Sweeney et al. (2000a) and roughly traces the western boundary between sea ice and open water when the Ross Sea polynya is opening during austral spring (Fig. 2a). Surface water salinity is lower in the western region (33.79 ± 0.27) than the central region (34.11 ± 0.10) (Fig. 2b). While sea ice advects northwards in the central region 20 as the Ross Sea polynya forms, in the western region much of the sea ice melts in place lowering the surface salinity and increasing stratification. Monthly Aqua MODIS chlorophyll concentration data shows that the highest chlorophyll concentrations during February 2013 are in the western region (Fig. 2c). Arrigo et al. (1999) also observed that diatom blooms in the highly stratified western region peak during the early autumn 25 ∼ 6 weeks after the main Phaeocystis antarctica bloom in the central Ross Sea. Surface TA values range from 2268 to 2346 µmol kg −1 (mean = 2314±16 µmol kg −1 ).
Since TA strongly covaries with salinity (R 2 = 0.86, residual ±6 µmol kg −1 ), the lowest TA values are located in the western region. Values of sTA range from 2336 to 2386 µmol kg −1 (mean = 2360 ± 7 µmol kg −1 ) and are influenced by calcification/dissolution as well as phytoplankton photosynthesis since one unit of nitrate draw-5 down increases TA by one unit (Brewer and Goldman, 1978) (Fig. 2d).
Surface pCO 2 values range from 162 to 354 µatm (Fig. 2e). Surface pCO 2 values are lower in the western region (238 ± 34 µatm) compared to the central region (319 ± 16 µatm) due to greater phytoplankton photosynthesis in the west. Surface Ω Ar values range from 1.40 to 2.42 and Ω Ca ranges from 2.24 to 3.89 (Fig. 2f). The highest Ω Ar values are in the western region (1.94±0.18, Ω Ca = 3.09±0.30) compared to the central region (1.58 ± 0.09, Ω Ca = 2.52 ± 0.14). Greater phytoplankton photosynthesis in the western region increases Ω by both decreasing DIC and increasing TA.
Spatial and temporal variations in surface water Ω Ar are mainly controlled by sDIC in the Ross Sea (Fig. 3). The concentration of sDIC decreased by 58 ± 20 µmol kg −1 15 from a winter value, causing Ω Ar to increase by 0.5 ± 0.2. In addition, sTA increased by 11 ± 7 µmol kg −1 during the preceding summer months, causing Ω Ar to increase by 0.1 ± 0.1. Although there was a significant reduction in salinity compared to winter values (0.7±0.3), Ω Ar only decreased by ∼ 0.01 due to this freshening since both DIC and TA concentrations were reduced. Lastly, the effect of temperature on Ω Ar was negligi-20 ble since the Ross Sea only experiences a 2 • C seasonal change in SSTs (Sweeney, 2003). Two processes can reduce sDIC: calcification and phytoplankton photosynthesis. To evaluate the importance of calcification, we use time dependent changes in potential alkalinity (pTA, defined as sNitrate + sTA) from a winter value (2367±3 µmol kg −1 , defined 25 as average value for all samples between 200-400 m). While TA will increase during photosynthesis due to nitrate drawdown, pTA will be conserved. Therefore, changes in pTA can be attributed to calcification and dissolution. The average ∆pTA from a winter concentration is negligible (0 ± 5 µmol kg −1 ); therefore, calcification appears to be insignificant and the increase in sTA from winter conditions is largely driven by nitrate drawdown during photosynthesis. Earlier studies found that calcification contributed to only ∼ 5 % of the total seasonal DIC drawdown (Sweeney et al., 2000a;Bates et al., 1998). Therefore, we argue that photosynthesis exerts the dominant control on sDIC, sTA, and Ω Ar . While the highest Ω Ar value that we observed was 2.4, values up to ∼ 4 5 have been observed during December-January (McNeil et al., 2010). By the time we arrived in the Ross Sea, surface sDIC concentrations would have already increased relative to the summer due to enhanced air-sea CO 2 fluxes (Arrigo and Van Dijken, 2007), deepening of the mixed layer (Sweeney, 2003), and remineralization of organic carbon (Sweeney et al., 2000b).
Mattsdotter Björk et al. (2014) also argue that phytoplankton photosynthesis is the major control on surface water Ω Ar variability between the Ross Sea and the Antarctic Peninsula based upon the covariance of Ω Ar and chlorophyll a. The largest contributor to seasonal Ω Ar change in the Chukchi Sea in the Arctic is also phytoplankton photosynthesis (Bates et al., 2013). However, unlike the Ross Sea, in parts of the Arctic sea 15 ice melt and river runoff leads to significant reductions in Ω Ar (Chierici and Fransson, 2009;Yamamoto et al., 2009;Robbins et al., 2013).

Ω in the Southern Ocean
The spatial changes in Ω Ar , SST, pCO 2 , and particulate organic matter (POC) between 75 and 55 • S are shown in Fig. 4 crease in Ω Ar is likely due to phytoplankton photosynthesis. This region may be along the Antarctic Slope Front that is known for higher biological activity (Jacobs, 1991). There is another step in Ω Ar from ∼ 1.4 to ∼ 1.5 between 67.5 and 67 • S. This step also corresponds with a decrease in pCO 2 from ∼ 370 to ∼ 340 µatm, likely due to phytoplankton photosynthesis. Elevated POC concentrations and lower pCO 2 values around the PF indicate enhanced phytoplankton photosynthesis. Ruben (2003) also found that pCO 2 is reduced south of the PF (170 • W) due to primary production.
We quantify the contribution of changing sDIC (calculated from TA and pCO 2 ), sTA, 5 SST, and salinity to changing Ω Ar along the transect (Fig. 5a). The dominant control is declining sDIC calc from ∼ 2240 to ∼ 2140 µmol kg −1 between 75 and 55 • S, which causes Ω Ar to increase by 0.87 if sTA, SST, and salinity are held constant. Declining sTA from ∼ 2340 to ∼ 2310 µmol kg −1 partially counters the influence of sDIC calc and reduces Ω Ar by 0.28. The influences of SST and salinity on Ω Ar are minimal.
10 Ω Ar variability is driven almost entirely by changes in sDIC calc from 75 • S to the PF.
Between the PF and SAF, variability in Ω Ar is influenced by the opposing effects of sDIC calc and sTA. The TA : DIC calc ratio and Ω Ar are constant between the PF and 60 • S since both sDIC calc and sTA decrease at the same rate. Between 60 • S and the SAF, Ω Ar increases since sDIC calc declines faster than sTA (Fig. 5b). North of the SAF, 15 Ω Ar variability is again driven by sDIC calc : Ω Ar increases due to a decrease in sDIC calc while sTA remains constant. The concentration of sDIC calc is highest south of the PF due to upwelling of CDW (Fig. 6a). To evaluate the properties of CDW, we use data from the 2011 Repeat Hydrography Cruise SO4P, which is part of the U.S. Climate Variability and Predictability 20 (CLIVAR) program (Swift and Orsi, 2012) (available at http://www.clivar.org/resources/ data/hydrographic). We only use data from hydrocasts located between 168 where the bottom depth is > 1000 m (Fig. 1b). We reject the data from hydrocast 46(B) where the deep DIC data below 200 m is ∼ 30 µmol kg −1 higher than the rest of the stations. Following Sweeney (2003) show that annual primary productivity increases from south to north in the Southern Ocean. In addition, surface waters north of the PF advect northwards and accumulate a sDIC deficit. Finally, warmer water holds less DIC while in equilibrium with the atmosphere. There is little net air-sea CO 2 flux between 75 and 55 • S (except for net efflux at 60 • S) since warming and increased biological production compensate each 15 other (Takahashi et al., 2012). The concentration of sTA is also highest south of the PF due to upwelling of CDW. Based off the CLIVAR dataset, the sTA of CDW is 2334 ± 3 µmol kg −1 . Surface water between 74 • S and the PF has a relatively constant sTA concentration of ∼ 2340 µmol kg −1 , slightly higher than its CDW source (Fig. 6b) Millero et al. (1998) that a negative linear relationship between sTA and SST is due to colder water being indicative of greater upwelling of TA rich water. This dataset supports the argument that increased upwelling of CDW from strengthening westerly winds will increase OA in the Southern Ocean (Lenton et al., 2009). While the TA : DIC ratio for CDW is 1.040±0.002, the TA : DIC calc ratio for surface waters 5 between 75 • S and the PF ranges from 1.046 to 1.064 (Fig. 5b). Therefore increased upwelling will lower the TA : DIC ratio and cause Ω Ar to decrease.

Estimate of wintertime surface Ω Ar values in the Ross Sea
Efforts to predict winter Ω Ar undersaturation in the Ross Sea are complicated by the complete lack of carbon system measurements from the winter months in the Ross 10 Sea.
McNeil et al. (2010) estimated winter surface water Ω Ar by using the lowest observed Ω Ar value from early spring when the Ross Sea is still covered by sea ice. They used mid October and early November carbon system measurements from the Joint Global Ocean Flux Study (JGOFS) (Sweeney et al., 2000b). Although sea ice algae produc-15 tivity peaks in November, their impact on water column DIC concentrations is likely to be negligible (Saenz and Arrigo, 2014). McNeil et al. (2010) found that early spring surface water Ω Ar was ∼ 1.2. There was a single Ω Ar value < 1.1 that they used as an initial condition along with the IPCC US92a scenario to predict that surface waters of the Ross Sea could begin to experience seasonally undersaturated conditions with 20 respect to aragonite as early as 2015 if full equilibrium with rising atmospheric CO 2 is achieved. Based on a three-dimensional Coupled Ice, Atmosphere, and Ocean model Arrigo, 2005), McNeil et al. (2010) argued that only 35 % of the atmospheric CO 2 signal equilibrates with Ross Sea surface waters due to sea ice, upwelling of CDW, and short residence times, thereby delaying the onset of 25 aragonite undersaturation until 2045. Decadal wintertime surface carbon system measurements do not exist to directly validate this disequilibrium assumption. In addition, McNeil et al. (2010)  dersaturated with respect to aragonite if the minimum wintertime surface Ω Ar value used was low due to measurement error. To independently calculate Ω Ar from early spring surface waters, we use the LDEO pCO 2 measurements from November 1994November , 1997November , 2005November , and 2006 that are from the Ross Shelf (defined by the 1000 m isopleth) and are south of 74 • S (Fig. 7a). The earli-5 est pCO 2 measurements are from 16 November 1994, 17 November 1997, 6 November 2005 November 2006 when much of the Ross Sea is still covered in sea ice. The earliest measurements from 2005/06 are more likely to represent winter conditions since they are from 74 • S as the NBP entered the Ross Sea. Conversely, the earliest measurements from 1994/97 are from the 76.5 • S line, close to where the Ross 10 Sea polynya opens up from. We calculate wintertime TA in the Ross Sea by establishing a salinity-TA relationship using data from Bates et al. (1998), Sweeney et al. (2000b, and our own hydrocast TA measurements from the upper 10 m (Fig. 7b). Since one unit of nitrate drawdown increases TA by one unit, the TA measurements are adjusted to winter nitrate concen- 15 trations of 29 µmol kg −1 (the mean nitrate concentration between 200-400 m from our cruise). The relationship between TA and salinity is consistent among these independent datasets and the standard deviation of the residuals for TA is ±5 µmol kg −1 .
We calculate historical Ω Ar using historical pCO 2 measurements, TA calculated from salinity, SST, and salinity. Phosphate and silicate are set to the winter values of 2.1 and 20 79 µmol kg −1 respectively. The thermosalinograph (TSG) salinity data from the historical pCO 2 measurements appears reasonable and is uncalibrated. While the largest offset in TSG salinity compared with Autosal measurements is 0.3, such error is not typical. For instance, on our cruise the difference between TSG and Autosal measurements is less than 0.02. To test the possible impact of a poor salinity calibration, we recalculate Ω Ar for all pCO 2 measurements after increasing salinity by 0.3. TA calculated from the observed TA-salinity relationship increases by ∼ 21 µmol kg −1 and Ω Ar increases by 0.024 ± 0.003. The lowest Ω Ar measurements are 1. 24 in 1994, 1.25 in 1997, 1.22 in 2005, and 1.20 in 2006 (Fig. 7c). Although Ω Ar declines from 1994 to 2006, we have low confidence in any trend due to spatial-temporal sampling biases. The lowest Ω Ar values are consistently between 1.2 and 1.3 as the ship crossed sea ice covered regions and open water that had experienced DIC drawdown. With the exception of a single mea- A simple calculation also suggests that wintertime Ω Ar values may be closer to 1.2 than 1.1. If salinity is 34.5, approximately the mean salinity of the water column, TA would be 2339 µmol kg −1 based on the observed TA-salinity linear relationship. Sweeney (2003) estimates winter pCO 2 values of ∼ 425 µatm based on deep pCO 2 measurements made during early spring. Setting salinity to 34.5, TA to 2339 µmol kg −1 , pCO 2 to 425 µatm, temperature to −1.89, silicate to 79 µmol kg −1 , and phosphate to 2.1 µmol kg −1 yields a Ω Ar value of 1.22.

15
Although pCO 2 measurements of surface waters colder than −1.75 • C south of 60 • S typically reach ∼ 410 µatm by September, Takahashi et al. (2009) present a few measurements of ∼ 450 µatm. Even if pCO 2 reaches 450 µatm during winter in the Ross Sea, Ω Ar would be 1.16 (with salinity at 34.5 and TA at 2339 µmol kg −1 ). In order to obtain Ω Ar of 1.1, pCO 2 would need to be ∼ 480 µatm, a value that appears unreasonably 20 high given the available datasets from the Ross Sea. McNeil et al. (2010) calculated the Ω Ar of water arriving onto the Ross Shelf following the recipes of Jacobs and Fairbanks (1985): 50 % CDW, 25 % Tmin water (minimum temperature in upper 100 m), and 25 % AASW. To calculate the Ω Ar of these three source water masses, they used hydrocast temperature, salinity, and DIC data col-Introduction We independently calculate Ω Ar of incoming water using the 2011 CLIVAR hydrocast data from north of the Ross Shelf between 168 • E-73 • W as described earlier (Fig. 1b).
The Ω Ar of water in the upper 100 m (AASW and Tmin) from the CLIVAR dataset is 1.36 ± 0.13 and the Ω Ar of CDW (maximum temperature below 150 m) is 1.18 ± 0.03 (Fig. 7d). Even if 100 % of the incoming water onto the Ross Shelf is CDW, the Ω Ar of 5 this incoming water would be significantly greater than 1.08. While most properties of CDW are similar between the 2011 CLIVAR data and the 1994 data used by McNeil et al. (2010), the TA of CDW from the CLIVAR dataset is 18 µmol kg −1 higher (Table 2).
Another approach to estimate the Ω Ar of winter surface waters is to use the properties of water below 200 m. For the TRACERS data, sTA below 200 m is 2338 ± 2 µmol kg −1 .
For the JGOFS autumn cruise (NBP 97-3) sTA below 200 m is 2339±2 µmol kg −1 . Using the CLIVAR dataset, sTA of CDW from off the Ross Shelf is 2334 ± 3 µmol kg −1 . This consistency between independent datasets suggests that we can accurately estimate winter TA in the Ross Sea. The range in sDIC below 200 m is much greater than that for sTA ( Table 2). The 15 lowest value is 2220±5 µmol kg −1 from our cruise and the highest is 2237±3 µmol kg −1 from the summer JGOFS cruise (NBP 97-01). This range in sDIC concentrations below 200 m is not surprising given that sDIC concentrations vary across the input water masses. In addition, sDIC concentrations below 200 m will be influenced by carbon export particularly in summer and early autumn and over multiple seasons' air to sea 20 flux of CO 2 . Assuming that deep water concentrations of TA and DIC are relatively unmodified following wintertime deep convective mixing, we estimate the Ω Ar of winter surface water by setting TA to 2338 µmol kg −1 , salinity to 34.5, temperature to −1.89 • C, phosphate to 2.1 µmol kg −1 , and silicate to 79 µmol kg −1 . If DIC concentrations are 2220 µmol kg −1 , 25 Ω Ar would be 1.37. If sDIC concentrations are 2237 µmol kg −1 , Ω Ar would be 1.24 and pCO 2 would be 417 µatm. These results are consistent with a study by Matson et al. (2014) where early spring Ω Ar at 20 m depth calculated using pH and salinity derived TA was 1.  1993-1995in Prydz Bay (McNeil et al., 2011). Hofmann et al. (2015 report Ω Ar at 18 m depth (bottom depth < 30 m) at two coastal sites in McMurdo Sound, the Jetty and Cape Evans, for December-May and November-June respectively using pH and salinity derived TA as input variables. The 5 lowest Ω Ar observations were from May at both sites and were 1.22 and 0.96 at the Jetty and Cape Evans. The maximum calculated pCO 2 was 559 at Cape Evans. The low Ω Ar and high calculated pCO 2 values measured by Hofmann et al. (2015) may represent differences between coastal and open ocean systems -there may be a coastal amplification signal when sinking organic matter hits a shallow bed. Another possibility is that their carbon system time series, particularly at Cape Evans, is inaccurate. After conditioning and calibrating their pH measurements using discrete water samples, for logistical reasons Hofmann et al. (2015) could not collect additional validation samples during deployment or measure multiple carbon system parameters for crosscheck. Although the SeaFET pH sensors that they used are generally stable, they can   1996, Deep.-Sea Res. Pt. II, 47, 3491-3520, doi:10.1016 (00)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Reuer, M. K., Barnett, B. A., Bender, M. L., Falkowski, P. G., and Hendricks, M. Bates et al., 1998), December-January 1995/96 (red, Bates et al., 1998), and April 1997(magenta, Sweeney et al., 2000. TA has been corrected to a nitrate concentration of 29 µmol kg −1 to account for the effects of nitrate drawdown on TA (Brewer and Goldman, 1976). (c) Aragonite saturation state (Ω Ar ) of surface waters from November calculated from pCO 2 , salinity derived TA, temperature, and salinity (d) profiles of aragonite saturation state (Ω Ar ) from off the Ross Shelf (see Fig. 1) from NBP 11-02 calculated from TA, DIC, temperature, and salinity at surface pressures.