First-year land-fast Antarctic sea ice as an archive of ice shelf meltwater fluxes

Cold Regions Science and Technology 113 (2015) 63–70

Contents lists available at ScienceDirect

Cold Regions Science and Technology journal homepage: www.elsevier.com/locate/coldregions

First-year land-fast Antarctic sea ice as an archive of ice shelf meltwater fluxes I.J. Smith a,⁎, A.J. Gough a, P.J. Langhorne a, A.R. Mahoney b, G.H. Leonard c, R. Van Hale d, S. Jendersie a,e, T.G. Haskell f a

Department of Physics, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand Geophysical Institute, University of Alaska Fairbanks, P.O. Box 757500, Fairbanks, AK 99775, USA National School of Surveying, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand d Department of Chemistry, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand e National Institute of Water & Atmospheric Research Limited (NIWA), 301 Evans Bay Parade, Hataitai, Wellington 6021, New Zealand f Callaghan Innovation, P.O. Box 31-310, Lower Hutt 5040, New Zealand b c

a r t i c l e

i n f o

Article history: Received 11 August 2014 Received in revised form 7 January 2015 Accepted 19 January 2015 Available online 26 January 2015 Keywords: Ice shelf basal meltwater Sea ice formation Oxygen isotope fractionation Surface water isotope changes Ice-ocean interaction Growth rates

a b s t r a c t Sampling beneath Antarctic ice shelves is sparse; therefore, tracking changes in ocean δ18O composition adjacent to ice shelves holds promise as an indicator of ice shelf basal melting. Sea ice archives of ice shelf–ocean interaction in particular could be important tools for future observational climate studies. Ocean δ18O values near the McMurdo Ice Shelf were reconstructed using observational data (sea ice δ18O, snow depth, and ice formation dates) from McMurdo Sound, Antarctica, by combining a recently revised version of an isotope fractionation model with an established thermodynamic sea ice model, resulting in improvements compared to previous approaches. Growth rates from the thermodynamic sea ice model were validated using direct growth rate measurements. That validation and supporting analysis indicated that a change was needed in ocean heat flux assumption from 0 W m−2 to around −13 W m−2 part way through the sea ice growth season. A well-constrained range (+ 1.84 ‰ to + 2.21 ‰) of effective fractionation coefficients for sea ice was derived, along with a mean of 1.99‰. For the first time, reconstructed ocean δ18O values were validated using winter-long measurements of Antarctic near-surface water δ18O. Taking uncertainties into account, the reconstructed ocean δ18O values generally agreed to within ±0.2‰ with the measured ocean δ18O mean values. Results indicated an overall decrease in measured ocean δ18O during the winter, but this was not statistically significant given the uncertainties in the measurements. Although the method works, it currently has limited utility for determining the presence and scale of any step-changes in ocean δ18O composition associated with present day ice shelf basal melting. This is because the uncertainty of the reconstructed values (±0.2‰) is of the same magnitude as the expected change. Also, the requirement to parameterise the ocean heat flux is a barrier to the method being an entirely retrospective method (i.e., one requiring only data from the end of the sea ice growth season). In a future Antarctic scenario of increased basal melting of the ice shelves, the method may become more valuable in an Antarctic context. The method developed in this paper will be useful currently in the Arctic, because Arctic waters exhibit much larger fresh water fluxes. © 2015 Elsevier B.V. All rights reserved.

1. Introduction 1.1. Ice shelves in a warming world Warming of Earth's climate system was stated by the Intergovernmental Panel on Climate Change (2013) to be “unequivocal,” with it “extremely likely” that anthropogenic effects dominated warming since the 1950s (Intergovernmental Panel on Climate Change, 2013). Under the SRES A1B scenario, climate and regional ⁎ Corresponding author. Tel.: +64 3 479 7755; fax: +64 3 479 0964. E-mail address: [email protected] (I.J. Smith).

http://dx.doi.org/10.1016/j.coldregions.2015.01.007 0165-232X/© 2015 Elsevier B.V. All rights reserved.

modelling indicate potential for increased melting of Antarctic ice shelves in the future due to ocean warming (Yin et al., 2011) and changes in ocean currents (Hellmer et al., 2012). Basal melt already occurs under Antarctic ice shelves, including deep beneath cold cavity ice shelves such as the Ross Ice Shelf and the Filchner-Ronne Ice Shelf, when High Salinity Shelf Water (HSSW) interacts with the ice to form Ice Shelf Water (ISW) (Jacobs et al., 1985). Detection and attribution of this basal meltwater from ice shelves (e.g., Hattermann et al., 2012) is rare but is important for monitoring ice shelves' response to warming conditions. There are complex and competing feedbacks that have implications for the stability of ice sheets if basal melting of ice shelves increases (Gagliardini et al., 2010).

64

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

1.2. Determining the response of ice shelves to climate change Identifying the timing and extent of meltwater fluxes from beneath ice shelves is particularly important for determining the response of ice shelves to climate change. However, sampling beneath Antarctic ice shelves is sparse due to logistical and financial challenges. Methods that use observations from surface waters adjacent to ice shelves have been used to obtain information on ice shelf basal melting but are often restricted to summer observations. Sea ice archives of ice shelf– ocean interaction are therefore potentially important tools in future observational climate studies, particularly in conjunction with other observations, or in places where winter observations are difficult logistically. Monitoring of changes in ISW is an important part of determining existing and future responses of ice shelves to climate change. ISW is seawater with its potential temperature below the surface freezing point temperature (Jacobs et al., 1985). Although the definition of ISW is based strictly on potential temperature, oxygen isotopes are sometimes used as an indicator of glacial input to the water mass. An increase in meltwater flux from beneath ice shelves has been suggested as the cause of lower salinity and more negative ocean δ18O values in the Ross Sea from the 1970s to 2000 (Jacobs et al., 2002). Despite observations indicating that ISW plumes only affect the sea ice at certain locations, those locations may be important archival sites for determining ice shelf–ocean processes. ISW is often thought of as a deep water mass, but there have been observations of ISW reaching the surface (e.g., Fer et al., 2012; Mahoney et al., 2011). ISW rising through the water column can become in situ supercooled due to pressure relief (Foldvik and Kvinge, 1974), and in situ supercooling has been observed as deep as 70 m (Leonard et al., 2011) . Sea ice that forms from ISW at the surface holds promise as part of the monitoring of ice shelves (Langhorne et al., revised version under review 2015). Incorporated platelet ice is a form of sea ice composed of dendritic crystals with columnar/congelation ice between those

crystals. Structural characteristics of incorporated platelet ice indicate that it is associated with the presence of ISW because dendritic crystals are indicative of growth in supercooled sea water. Incorporated platelet ice and related sub-ice platelet layers are observed in some regions in the later parts of the growth season (e.g., Paige, 1966) close to an ice shelf (e.g., Crocker and Wadhams, 1989; Gough et al., 2012a; Gow et al., 1998; Jeffries et al., 1993; Leonard et al., 2006). Although ISW is a necessary precursor for platelet ice formation, incorporated platelet ice does not always appear in cores simultaneously with the appearance of ISW (Gough et al., 2012a; Mahoney et al., 2011). Incorporated platelet ice cannot be distinguished from columnar ice just by the use of isotopes; for example δ18O values of incorporated platelet ice from a McMurdo Sound core were reported by Smith et al. (2012, supplementary material) to be between 1.55‰ to 1.77 ‰ (±0.02 ‰), which sits within the range of δ18O values for the columnar ice above it, which was 1.09‰ to 1.88‰ (±0.02‰). Also, as discussed by Smith et al. (2012), measuring the isotopic values of individual platelet crystals is problematic, so such measurements are not reported here. As a monitoring site for sea ice and ice shelf interaction processes, McMurdo Sound is an ideal location. McMurdo Sound (Fig. 1) is situated north of the McMurdo Ice Shelf. The ocean beneath the McMurdo Ice Shelf is connected to the Ross Ice Shelf cavity through Haskell Strait, with a flow-through of ISW likely originating from beneath the larger adjacent Ross Ice Shelf. Platelet ice observations are common in McMurdo Sound and in the south-western Ross Sea (e.g., Crocker and Wadhams, 1989; Gough et al., 2012a; Gow et al., 1998; Jeffries et al., 1993; Leonard et al., 2006; Smith et al., 1999; Smith et al., 2001). A map indicating platelet ice abundance in McMurdo Sound was produced by Dempsey et al. (2010), and a pan-Antarctic map showing all published observations of platelet ice was produced by Langhorne et al. (revised version under review 2015). On the eastern side of McMurdo Sound, platelet ice is observed in the late winter (Smith et al., 2001; Smith et al., 2012). Crocker and

Fig. 1. Locations of sea ice (EB-09) and ocean water samples (EB-09 and CA-09) taken in McMurdo Sound, Antarctica 2009. The insert shows the location of McMurdo Sound (arrow) in relation to Antarctica.

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

Wadhams (1989) suggested that platelet ice formed in 1986 in McMurdo Sound subsequent to the arrival of ISW. Mahoney et al. (2011) noted that although AASW (Antarctic Surface Water) and HSSW were present early in 2009, net currents were consistent with outflow from under the ice shelf from 30 April, and ISW was the only water mass present by the end of July. Smith et al. (2012) suggested that a step-change in the oxygen isotope composition of the surface sea water may have occurred at their 1999 McMurdo Sound site at the time incorporated platelet ice formed. Leonard et al. (2006) reported the first direct measurements of δ18O for surface sea water for McMurdo Sound (from the winter of 2003) but did not have water samples from periods prior to platelet ice appearing in cores. On the western side of McMurdo Sound, sea ice cores can be comprised entirely of incorporated platelet ice (Dempsey et al., 2010; Hughes et al., 2015), a consequence of a persistent, year-round outflow of ISW on the western side of the sound (Hughes et al., 2015; Robinson et al., 2014). Hughes et al. (2015) suggested an expansion of the lateral extent of ISW emerging from beneath the McMurdo Ice Shelf occurs as winter progresses. A significant outflow of ISW from the McMurdo-Ross Ice Shelf cavity, therefore, passes through McMurdo Sound, and this ISW signal is archived in the sea ice. The ISW-platelet dependence is confirmed by observations in the Weddell Sea. A “tentative total mean concentration” of 20–30% (in volume percent) platelet ice for coastal waters of the Weddell Sea was reported by Lange (1988) for the eastern and southern Weddell Sea. This is a region where ISW has been observed to reach the surface seasonally (Fer et al., 2012; Nicholls et al., 2001). Such a coincidence of ice shelf outflow and platelet ice observations are not as common in the central and eastern Ross Sea; for example, Kawamura et al. (2004) found that platelet ice contributed only 1% of the pack ice in the central and eastern Ross Sea, as a percentage of total cores from all sites. This lack of platelet ice observations may be due to the ISW plumes not reaching the ocean surface (e.g., see Jacobs et al., 1985), although the episodic nature of ISW plumes (Bergamasco et al., 2002) may also play a role. It is worth noting that reported glacial meltwater fractions in the Ross Sea are less than 1%; for example, a late summer (February–March 2000) mean glacial meltwater fraction of 0.22 ± 0.01 % at the Ross Ice Shelf front, with minimum values of 0– 0.2% (eastern Ross Sea: 100–200 m depth; western Ross Sea: all depths) and maximum values of 0.73 ± 0.07% (eastern Ross Sea: around 400 m depth), was reported by Loose et al. (2009).

1.3. Sea ice oxygen isotopes as a tracer of refrozen ice shelf meltwater This paper investigates whether a similar approach to Pfirman et al. (2004) can be used to determine the presence of wintertime ice shelf meltwater in regions where platelet ice is present. The key differences to the approach taken by Pfirman et al. (2004) are as follows: (1) use of the more detailed thermodynamic sea ice growth rate model of Bitz and Lipscomb (1999) rather than Thorndike (1992); (2) validation of modelled sea ice growth rates with measurements made at the site; (3) calculation of isotopic fractionation from sea ice growth rates using the modified Eicken (1998) equations proposed by Smith et al. (2012) and Toyota et al. (2013). Validation of the reconstructed sea water δ18O values against water samples taken at the same location and time as sea ice cores is then examined. Such a comparison has not been reported previously, probably because sea water samples taken during an Antarctic winter, such as those presented in this paper, are rare. The sea water δ18O measurements and reconstruction are then discussed in the context of detection of possible step-changes in δ18O, which Smith et al. (2012) suggested could range from summer values (e.g., δ18O = −0.31‰ for Antarctic surface water without ice shelf water input (Jacobs et al., 1985)) to winter values (e.g., δ18O = −0.50‰ near an ice shelf (Smith et al., 2012)).

65

2. Methods 2.1. Observational data needed for simulations All data presented in this paper are from analyses of sea water and first-year land-fast sea ice samples from McMurdo Sound, Antarctica (Fig. 1). McMurdo Sound lies at the south-western end of the Ross Sea and is bounded to the west by the Victoria Land Coast, to the east by Ross Island, and to the south by the McMurdo Ice Shelf, with inflow from beneath the ice shelf via Haskell Strait. Sea ice samples were cores of first-year sea ice that had formed and remained land-fast in Erebus Bay on 28 April 2009. Cores were taken on 12 May, 22 June, 10 August, 21 September, and 27 September 2009 (Gough et al., 2012a). The sea ice samples were comprised predominantly of columnar ice in upper levels of the cores and incorporated platelet ice in the lower levels. Sea water samples were collected at two sites, as described below. The presence of incorporated platelet ice observed in sea ice cores and the negative ocean heat flux measured in McMurdo Sound in 2009 are unarguably the result of ice shelf basal meltwater reaching the surface during the late winter (Gough et al., 2012a; Mahoney et al., 2011). 2.1.1. Oxygen isotopes Oxygen isotope ratios are reported using delta notation (δ18O). The delta value is a function of the difference in the 18O/16O ratio between the sample and an international standard, expressed in parts per thousand (or per mille, written‰):  δ¼

 Rsample −1  1000 Rreference

ð1Þ

where R is the 18O/16O ratio and Vienna Standard Mean Ocean Water (VSMOW) is normally used as the reference for ocean samples. By definition, δ18O = 0‰ for VSMOW. Surface ocean waters in the Arctic have δ18O values that range from approximately − 6 ‰ to + 1 ‰ (Pfirman et al., 2004, using data from G.A. Schmidt and others, 1999 Global seawater oxygen-18 database, http://www.giss.nasa.gov/data/o18data/), reflecting the significant contribution of river run-off in the Arctic. In Antarctica, where there are no large rivers flowing to the ocean, the range of δ18O values over all depths is much narrower, for example, − 0.71 ‰ to − 0.07 ‰ for the Ross Sea (Jacobs et al., 1985). Snow and ice shelves composed of compacted snow have highly negative δ18O values (e.g., δ18O = −42‰ near the base of the Ross Ice Shelf (Grootes and Stuiver, 1986)). Meltwater from the base of ice shelves therefore lowers the ocean values (e.g., δ18O = − 0.71 ± 0.06 ‰ for ocean water from beneath the Ross Ice Shelf (Jacobs et al., 1985)). Given the narrow range of Antarctic sea water δ18O values, measurements require a high level of instrumental accuracy and precision to determine seasonal and interannual changes. Traditional CO2 equilibration techniques on mass spectrometers (Epstein and Myeda, 1953) can achieve precisions around ± 0.02 ‰ to ± 0.03 ‰ (Jacobs et al., 1985), which is usually taken to be sufficient to resolve such changes. Using Arctic pack ice samples, Pfirman et al. (2004) inferred surface water δ18O values at the time the lower parts of the sea ice cores formed. Those samples were mainly slow growing columnar ice at the base of multi-year sea ice that had drifted through the Arctic Basin as it grew. Pfirman et al. (2004) used the Eicken (1998) stagnant boundary layer model, which gave equations for the derivation of effective fractionation coefficients, εeff,si, from sea ice growth velocities, where εeff,si is given by 18

18

εeff;si ¼ δ Oi −δ Ow

ð2Þ

where δ18Oi is the oxygen isotope value for sea ice and δ18Ow is the oxygen isotope value for sea water. To use this model, Pfirman et al. (2004) used ice drift back trajectories and a simple thermodynamic sea ice

66

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

growth rate model (Thorndike, 1992) to assign a date and geographic location to each sample depth in the sea ice core. Pfirman et al. (2004) provided compelling evidence that in the Arctic sea ice archives ocean δ18O changes. However, no work to date has explored whether such archives can be used in the Antarctic. 2.1.2. Data collection methods Sea water and sea ice cores samples in 2009 were collected for isotope analysis (see Toyota et al. (2013) for a full description of the isotope analysis methods used). Note that Toyota et al. (2013) only used columnar ice data, whereas incorporated platelet ice data (i.e., deeper sea ice core sub-samples) are also included here. Twenty sea water samples from Niskin casts were taken between February and September 2009, with eight δ18Ow samples from beneath the Cape Armitage site (CA-09) and twelve from the Erebus Bay site (EB-09) at depths of 8 m and 10 m (see Fig. 1 for site locations). For safety and logistical reasons, the CA-09 sea water samples were taken between February and June 2009 beneath stable multi-year sea ice, while EB-09 sea water samples were collected from beneath first-year sea ice. Mahoney et al. (2011) noted that “Overall, the data indicate that the water columns at the two mooring sites experienced the same seasonal evolution,” so we are confident that the sea water samples from these two sites can be taken to represent a continuous record of change at the EB-09 site. The ocean was accessed through an ice hole below a heated hut. Each sample was collected by opening the tap of the Niskin bottle and passing the sea water through a sieve to avoid the collection of ice or other debris before being decanted into 150 ml sample bottles, which were rinsed twice before being filled and capped. Samples were stored and transported at above freezing temperatures to New Zealand for isotope analysis. Seventy-seven sea ice sample δ18Oi measurements were made of columnar (n = 66), columnar/platelet transition (n = 3) and incorporated platelet ice (n = 8) from 8 cores collected from EB-09 site (see Fig. 1 and Gough et al., 2012a). In the analysis presented in this paper results from 59 sea ice sample δ18Oi measurements were included in the reconstruction of δ18Ow values (see Tables 1 and 2). The precision of the δ18Oi measurements was 0.04‰ for one core, and 0.05‰ for all the other cores. Sea ice growth rates, vi, derived from temperature probe measurements for 2009 were used to validate the application of the Bitz and Lipscomb (1999) model to McMurdo Sound. For the full details of the method used to obtain the growth rate data, vi, see Gough et al. (2012a). Table 1 Derived effective fractionation coefficients and measured δ 18 O i used to calculate reconstructed δ18Ow values shown in Fig. 3b. Ice structure

Derived effective fractionation coefficients (‰)

Measured δ18Oi (‰)

Depth (m)

Columnar (c) or incorporated platelet ice (p)

Minimum

Maximum

Mean

Number of cores contributing to mean at each depth

0.4–0.5 0.5–0.6 0.6–0.7 0.7–0.8 0.8–0.9 0.9–1.0 1.0–1.1 1.1–1.2 1.2–1.3 1.3–1.4 1.4–1.5 1.5–1.6 1.6–1.7 1.7–1.8 1.8–1.9 1.9–2.0

c c c c c c c c c c c c p p p p

1.85 1.84 1.87 1.86 1.95 1.89 1.85 1.95 2.12 2.06 2.08 2.16 2.02 2.02 1.99 1.99

1.86 1.88 1.88 1.90 1.97 1.98 1.87 2.02 2.16 2.11 2.15 2.21 2.08 2.06 2.03 2.03

1.21 1.37 1.43 1.39 1.58 1.50 1.46 1.50 1.58 1.51 1.60 1.39 1.40 1.50 1.46 1.39

6 4 5 5 5 5 5 3 3 3 1 3 3 3 3 2

Table 2 Mid-point values and half “height” of the reconstructed δ18Ow box (‰) with depth for Fig. 3b results. Depth (m)

Reconstructed δ18Ow (‰) (2 d.p.)

Half “height” of the reconstructed δ18Ow box (‰) (2 d.p.)

0.4–0.5 0.5–0.6 0.6–0.7 0.7–0.8 0.8–0.9 0.9–1.0 1.0–1.1 1.1–1.2 1.2–1.3 1.3–1.4 1.4–1.5 1.5–1.6 1.6–1.7 1.7–1.8 1.8–1.9 1.9–2.0

−0.65 −0.49 −0.45 −0.49 −0.38 −0.44 −0.39 −0.49 −0.56 −0.58 −0.52 −0.79 −0.65 −0.55 −0.55 −0.62

0.03 0.05 0.03 0.04 0.03 0.07 0.03 0.06 0.05 0.05 0.09 0.05 0.06 0.05 0.05 0.05

2.2. Simulation methods 2.2.1. Sea ice growth rate calculations from reanalysis data Sea ice growth rates were calculated using a one-dimensional thermodynamic model from Bitz and Lipscomb (1999) (denoted as BL99 from here on) forced with NASA MERRA 0.25° daily output (Rienecker et al., 2011) from the grid point nearest to the site. The BL99 model default (Arctic) parameters were modified to represent first-year Antarctic fast ice by setting the sea ice salinity to 5 parts per thousand, and the ocean heat flux, Fw, initially to be 0 W m−2 (i.e., no heat flux between the ocean and the sea ice) over the whole time period, based on the early season calculations of Gough et al. (2012a). Although Fw can be positive and large (and therefore reduce sea ice growth rates) in open ocean areas such East Antarctica (Heil et al., 1986) and the Weddell Sea (Allison, 1981), in McMurdo Sound, Fw values have been reported to oscillate about 0 W m−2 in the early sea ice growth period, with consistently negative values later on (Gough et al., 2012a; Purdie et al., 2006). Using data from multiple thermistor probes frozen into the sea ice, the changing energy balance near the sea ice-ocean interface throughout the 2009 winter was carefully calculated by Gough et al. (2012a), with Fw calculated at each time step as a residual heat flux (McPhee and Untersteiner, 1982; Purdie et al., 2006). In other words, Fw was calculated as the heat flux not attributable to heat conduction to the atmosphere, internal temperature or phase changes within the sea ice, nor to phase changes at the bottom of the sea ice. In order to neglect convection effects, Gough et al. (2012a) selected a depth to calculate the energy balance where the bulk salinity was stable, which was determined to be 0.15 m above the sea ice-ocean interface. Gough et al. (2012a) noted that when Fw is negative, this is likely to be because of both downward heat flow from the sea ice to the supercooled ocean and the accumulation of ice crystals that have been advected from elsewhere. The major uncertainty in correctly simulating land-fast sea ice growth using this approach arises from poor knowledge of (i) a site's snow accumulation history and (ii) the date that ice cover first formed and remained land-fast. To address (i), the measured total snow accumulation at the time the last sea ice sample was taken was used to determine a range of likely snow accumulation histories by scaling the reported MERRA precipitation field to reproduce 80% to 120% of the measured total (i.e., 0.12 to 0.18 m). To address (ii), a feasible range of ice formation dates (25–30 April 2009) was determined from visual observations of the site. Simulations were then performed using combinations of start date and scaling of the snow accumulation history, which resulted in a range of possible growth rates and formation times for each sub-sample (layer) of the sea ice.

εeff;si ¼ a1 þ a2 expða3 ðvi þ a4 ÞÞ þ a5 expða6 ð vi þ a4 ÞÞ

ð3Þ

where for vi b2.0 × 10−7 m s−1: a1 = −7.8328 × 10−3 ‰, a2 = 2.2416‰, a3 = −1.3692 × 106 s m−1, a4 = −4.6278 × 10−8 m s−1, a5 = 0.4918‰, a6 = −2.9447 × 107 s m−1; and for vi ≥2.0 × 10−7 m s−1: a1 = 1.2280‰, a2 = 0.7311‰, a3 = −1.2484 × 107 s m−1, a4 = 0 m s−1, a5 = 0.8441‰, a6 = −1.2821 s m−1. Previous papers, such as Eicken (1998), have not included an uncertainty associated with the equations for determining εeff,si from vi. It is beyond scope of this paper to quantify the uncertainties for such equations. Eq. (3) is therefore treated as if the only uncertainties arise from the values of vi used, and not from any inherent uncertainties in the equation itself. 2.3. Combining sea ice δ18Oi measurements (2.1) with simulations (2.2) to calculate sea water δ18Ow On any given sampling date at the EB-09 site, multiple sea ice cores (“samples”) were taken. Each core was then cut into 0.1 m sections (“sub-samples”). Results for sub-samples taken from the same depth interval at each site were combined to produce an estimate of the true δ18Oi of the sea ice. This gave 18 δ18Oi values, representing the mean of each set of sub-samples from the same depths. Sub-samples from 0 to 0.20 m depth were excluded from the analysis because the presence of granular snow ice/frazil in the first 0.10 m meant that the ice was not representative of early season congelation ice growth, and for 0.10–0.20 m, the BL99 model runs had significant uncertainty in the rates due to the length of time step in the code and the high growth rates. The effective fractionation coefficients for sea ice, εeff,si, calculated using Eq. (3) were then used in conjunction with the measurements of δ18Oi to calculate the δ18Ow values for surface sea water by rearranging Eq. (2). 3. Results Fig. 2a shows that for 2009, assuming Fw = 0 W m−2 (i.e., zero heat flux between the ocean and the ice), the sea ice growth rates, vi, calculated from BL99 (circles) closely match the sea ice growth rates measured by Gough et al. (2012a) for ice less than or equal to 1.50 m

Sea ice growth rate (10-7 m s-1)

3.47

3

2.89

2.5

2.31

2

1.74

1.5

1.16

1

0.58

0.5

0.00

0

0.5

1

1.5

2

0 2.5

Sea ice thickness (m)

b

3.47

3

2.89

2.5

2.31

2

1.74

1.5

1.16

1

0.58

0.5

0.00 0

0.5

1

1.5

2

Sea ice growth rate (cm day-1)

2.2.2. Sea ice effective fractionation coefficient calculations For each sub-sample, the sea ice growth rates, vi, calculated using the method described in Section 2.2.1 were used to calculate the subsample's expected range of effective fractionation coefficients for sea ice, εeff,si. For low growth rates, vi b2.0 × 10−7 m s−1, see Toyota et al. (2013) for the physical reasons for this choice, εeff,si was calculated using Eicken's (1998) Eq. (22) with a growth velocity offset (− 4.6278 × 10− 8 m s−1) as applied by Smith et al. (2012). For high growth rates (vi ≥ 2.0 × 10− 7 m s− 1), εeff,si was calculated using Eq. (9) of Toyota et al. (2013). These equations are expressed below as Eq. (3), for vi in units of m s−1:

a

Sea ice growth rate (10-7 m s-1)

The simulated growth rates were then validated through comparison with the measured growth rates. This indicated that the assumption of Fw = 0 W m−2 was reasonable in the early part of the season, but changes were needed to the ocean heat flux assumption for later in the season. Specifically, the values of the inputs to BL99 were modified to include a change to a negative ocean heat flux, as reported by Gough et al. (2012a). After incorporated platelet ice was observed in the core, Fw was taken to be the lower bound of Fw reported by Gough et al. (2012a) for September 2009 (Fw ≈ − 13 W m− 2, using lower limit of Fw = −10.6 ± 2 W m−2 and rounding to two significant figures). According to Gough et al. (2012a), platelet ice was first observed in the cores when the sea ice thickness was between 1.6 m and 1.7 m, and sea ice thickness reached 1.6 m between 10 and 15 August, so the change in Fw was applied after 12 August.

67

Sea ice growth rate (cm day-1)

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

0 2.5

Sea ice thickness (m) Fig. 2. Growth rate validation graphs. The output of the Bitz and Lipscomb (1999) model runs are shown as circles, with each circle representing different combinations of ice formation date and snow accumulation scaling. Green circles indicate a specimen model run for ice formation data of 28 April 2009 and 0.15 m final snow depth. Measured growth rates are shown as a red line with uncertainties as pink shading. Dash line indicates the designation of high and low growth rates, with Toyota et al. (2013) used for high growth rates and Smith et al. (2012) used for low growth rates. BL99 model runs performed with. (a) ocean heat flux Fw = 0 W m−2 for all runs; (b) ocean heat flux Fw = 0 W m−2 for all runs prior to 12 August 2009, then ocean heat flux Fw = −13 W m−2 after that.

thick. However, when the ice thickness exceeded 1.50 m, the calculated growth rates deviate substantially from the measured sea ice growth rates. Measured sea water δ18Ow results from 8 m or 10 m depth are plotted along with reconstructed sea water δ18Ow values (from Eqs. (2) and (3)) in Fig. 3a, assuming Fw = 0 W m−2 for the reconstruction. Error bars on measured sea water δ18Ow values indicate the uncertainties given by the precision for ocean measurements. The height of each box indicates the maximum and minimum limits on each reconstructed δ18Ow value by taking into account the precision on ice isotope values, and the uncertainty in the modelled sea ice growth rates. The width of each box indicates the range of times associated with each reconstructed δ18Ow value, which is also a function of ice growth rate uncertainty. The boxes widen as time progresses due to the increase in sea ice thickness uncertainty as the growth rates are progressively integrated over time. Since the first two calculated bands in Fig. 2a were not validated, reconstructions of sea water δ18Ow relating to those depths (0.20–0.30 m and 0.30–0.40 m) have been omitted from Fig. 3(a). Measurements of c-axis distributions (Gough et al., 2012a) have been used to divide the reconstructed δ18Ow values according to

68

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

a

-0.4

box values corresponds to the uncertainties in the reconstructed δ18Ow values and range from 0.03 to 0.09‰ for Fig. 3b (see Table 2 for tabulated values). As a result of the change in assumption of Fw, all but one of the reconstructed δ18Ow values bound the mean of the measured δ18Ow values by up to ±0.2‰ (Fig. 3b).

-0.6

4. Discussion

Sea water δ18O (‰)

-0.2

-0.8

-1

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Date (2009)

b

Sea water δ18O (‰)

-0.2

-0.4

-0.6

-0.8

-1

Mar

Apr

May

Jun

Jul

Aug Sep

Oct

Nov

Date (2009) Fig. 3. Ocean δ18Ow validation graphs. Measured data δ18Ow (sea water) values from 8 m, and 10 m depth are shown in blue: double casts (crosses), single casts (circles), and mean (line). Reconstructed ocean δ18Ow values are shown as boxes, with coloring from structural information: columnar ice (green boxes), incorporated platelet ice (red boxes). Reconstructed ocean δ18Ow values calculated with. (a) ocean heat flux Fw = 0 W m−2 for all runs; (b) ocean heat flux Fw = 0 W m−2 for all runs prior to 12 August 2009, then ocean heat flux Fw = −13 W m−2 after that.

sea ice core depths corresponding to columnar ice (green boxes) or to incorporated platelet ice (red boxes). Up to and including 23 July, the reconstructed δ18Ow values bound the mean of the measured δ18Ow values by up to ±0.2‰ (Fig. 3a). During that period, the reconstructed δ18Ow values also reflect trends in the measured δ18Ow. For dates after 23 July, the reconstructed δ18Ow values are consistently lower than the measured values (see Fig. 3a, last five boxes on right). The validation of the BL99 simulated growth rates, by comparison with the measured growth rates, indicated that the initial assumption of a constant ocean heat flux was not correct. Based on the observations of Gough et al. (2012a), the assumption was therefore then changed to a step-change in ocean heat flux from Fw = 0 W m−2 to Fw = − 13 W m−2 from 12 August onwards. The resulting growth rates are presented in Fig. 2b. These revised calculated growth rates were then used to obtain the reconstructed δ18Ow values presented in Fig. 3b. Overestimated growth rates for the early part of the growth season remain, however, and therefore the time period associated with each box in Fig. 3b is earlier than structural core data indicate. The values of the derived effective fractionation coefficients used to obtain the reconstructed δ18Ow values in Fig. 3b range from 1.84‰ to 2.21‰ (Table 1). Measured mean δ18Oi values, also shown in Table 1, range from 1.21‰ to 1.60‰. Also shown is the number of cores contributing to the mean at each depth and ice structure type (columnar or incorporated platelet ice). Half the “height” of the reconstructed δ18Ow

Reasonable agreement between modelled and measured sea ice growth rates was obtained using BL99 and Fw = 0 W m−2 for sea ice 0–1.5 m thick (Fig. 2a). Once sea ice thickness exceeded 1.5 m, the measured sea ice growth rates deviated from modelled thickness (Fig. 2a). For this same site and year, Gough et al. (2012a) reported that the sea ice structure changed from columnar to incorporated platelet ice at 1.57 m depth. The sharp increase in measured sea ice growth rates at 1.5 m (Fig. 2a) is coincident with the presence of incorporated platelet ice, consistent with earlier observations by Smith et al. (1999), who reported that sea ice in McMurdo Sound in 1997 showed an increase in ice thickness growth rate at 1.4 m depth, which corresponded to the start of incorporated platelet ice that year. Similarly, Trodahl et al. (2000) noted that negative oceanic heat flux in late winter-early spring 1997 in McMurdo Sound was associated with the presence of incorporated platelet ice. The increase in growth rates can therefore be attributed to the presence of supercooled water late in the winter that enhanced sea ice growth rates (e.g., Gough et al., 2012a). A comparison of Fig. 2a and b shows that the introduction of the step-change gives better agreement between the measured and modelled growth rates. The effect of applying a step-change to Fw on reconstructed δ18Ow is seen by comparing Fig. 3a with Fig. 3b. As noted earlier, Fig. 3a shows that holding Fw = 0 W m−2 for the entire run produces reconstructed δ18Ow that agreed well with measured δ18Ow before 23 July but fails to reproduce δ18Ow after this time. When a negative ocean heat flux of − 13 W m−2 was applied from 12 August onwards (i.e., times corresponding to platelet ice growth (Gough et al., 2012a)), reconstructed δ18Owvalues more closely matched measured δ18Ow in the latter part of the record, as evidenced by comparing the location of the red boxes relative to the mean of the δ18Ow observations in Fig. 3a and b. There is only one case, on 2 August (green box 5th from the right, Fig. 3b), where the reconstructed δ18Ow lies significantly below the mean measured δ18Ow trend. We deduce that this indicates a real change in water composition. Specifically, the reconstructed δ18Ow seems to have captured a temporary decrease in δ18Ow of 0.2‰ associated with the arrival of ISW prior to the onset of platelet ice formation. No water sample was taken on that date to directly confirm this. However, Mahoney et al. (2011) noted in 2009 that ISW arrived prior to the onset of platelet ice, so that a decrease in δ18Ow without a corresponding negative heat flux and without platelet ice formation is consistent with the Mahoney et al. (2011) observations. The very good agreement between BL99 model runs and measured growth rates for 2 August using Fw = 0 W m−2 (Fig. 2a) and the consistency of the three sub-samples used to obtain the δ18Oi value with the samples at the next depth below led us to discard the alternative interpretations that the reconstructed δ18Ow is not correct because a negative heat flux occurred earlier than modelled, or because an incorrect δ18Oi value was measured. As noted above, applying a negative ocean heat flux Fw = −13 W m−2 based on Gough et al. (2012a) in BL99 model runs after 12 August improved the model growth rate calculations (Fig. 2b) and reconstructed δ18Ow values (Fig. 3b). However, for a retrospective method, measurements of ocean heat flux over the growth season would not be available, which is a significant challenge to using this method. Further research is needed on parameterisation of the ocean heat flux and the uncertainties that would result from using such a parameterisation. Furthermore, incorporated platelet ice is comprised of a combination of individual dendritic ice crystals and ice formed by conduction of

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

heat to the air-ice interface (congelation ice). The dendritic crystals' growth rates are an order of magnitude faster than the congelation ice (Smith et al., 2012), and this may complicate the calculation of effective fractionation coefficients used in the reconstruction if large proportions of incorporated platelet ice make up the cores. The range of derived effective fractionation coefficients (1.84 ‰ to 2.21‰, see Table 1) and the mean (1.99‰) provides validation of the Pfirman et al. (2004) estimate from Arctic field measurements of approximately 2‰ fractionation during sea ice formation. This paper presents results from a system that was better constrained than previous studies such as Pfirman et al. (2004) in three key ways. First, direct sea ice growth rate measurements from temperature probes frozen into the sea ice were available for many of our sub-samples. Second, all of our samples were from first-year Antarctic land-fast ice, so the samples remained spatially fixed during ice growth. Third, and crucially, we have δ18O ocean measurements from the same location as sea ice sub-samples throughout most of the ice growth season. We are therefore able to carry out a more robust reconstruction of δ18O ocean values during sea ice growth periods by validating it with actual measurements. However, despite the well-constrained nature of this site, the challenge of applying this method in McMurdo Sound is the narrow range of δ18Ow values expected. The mean of the measured δ18Ow values varied by approximately 0.4 ‰ during the sea ice growth period, and exhibited an overall decreasing trend from March onwards. However, measurements on samples taken at the same time, location, and depth varied by as much as 0.3‰, so the trend is not statistically significant (R-squared value of approximately 0.3). There was greater scatter about the mean for the measured values than the uncertainty for the reconstructed δ18Ow values (Fig. 3b and Table 2). Reconstructed δ18Ow for 2009 varied by as much as 0.5 ‰ (Fig. 3b, taking uncertainties into account) over the sea ice growth period at observation sites on the eastern side of McMurdo Sound. Note that for the size of sub-samples (0.1 m long sections of sea ice core) analyzed in this paper, the sea ice δ18Oi, and therefore derived δ18Ow, was averaged over approximately a week, so this would smooth out any short term fluctuations in δ18Ow. This method should therefore be better than measurements of water samples taken at one time as a representative sampling of significant and persistent changes in water mass composition. Taking multiple cores and averaging at the same depth intervals has been done elsewhere for salinity (Gough et al., 2012b); it has been done here for δ18Oi to further reduce the uncertainties in the reconstruction of δ18Ow. Although the method described in this paper works, it currently has limited utility for determining the presence and scale of any stepchanges in ocean δ18O composition associated with present day ice shelf basal melting. This is because the uncertainty of the reconstructed ocean δ18O values (±0.2‰) is of the same magnitude as the expected change. Also, the requirement to parameterise (or measure) the ocean heat flux is a barrier to the method being an entirely retrospective method (i.e., one requiring only observational data from the end of the sea ice growth season on sea ice δ18O, snow depth, and ice formation dates). However, the method developed here will be useful in the present day Arctic Ocean, where Pfirman et al. (2004) originally used a similar approach, since Arctic waters exhibit much larger fresh water fluxes. In addition, in a future Antarctic scenario of increased basal melting of the ice shelves, the method may become more valuable in an Antarctic context. Mapping the expected future distribution, timing and extent of low δ18O composition as an indicator of ice shelf melting would therefore be a useful exercise. Acknowledgments Collection of the 2009 McMurdo Sound data was part of New Zealand's contribution to the International Polar Year and was funded by the then Foundation for Research Science and Technology (FRST). The University of Otago, Industrial Research Ltd and the National

69

Institute of Water and Atmospheric Research Ltd (NIWA) also provided financial and in-kind support. We are grateful to Antarctica New Zealand, the Scott Base 2009 winter team, Brian Staite, and Brett Grant for logistical and field support. Thanks to Myles Thayer, Richard Sparrow, and Peter Stroud for constructing instruments; Joe Trodahl, Daniel Pringle, Mike Williams, and Craig Stevens for advice; Russell Frew and Dianne Clark for isotope analysis; and Cecilia Bitz and Bill Lipscomb for making their code available via: http://www.atmos.washington. edu/~bitz/column/.

References Allison, I., 1981. Antarctic sea ice growth and oceanic heat flux. IAHS Publ. 131, 161–170. Bergamasco, A., Defendi, V., Meloni, R., 2002. Some dynamics of water outflow from beneath the Ross Ice Shelf during 1995 and 1996. Antarct. Sci. 14 (1), 74–82. Bitz, C.M., Lipscomb, W.H., 1999. An energy-conserving thermodynamic model of sea ice. J. Geophys. Res. 104 (C7), 15669–15677. Crocker, G.B., Wadhams, P., 1989. Modelling Antarctic fast-ice growth. J. Glaciol. 35 (119), 3–8. Dempsey, D.E., Langhorne, P.J., Robinson, N.J., Williams, M.J.M., Haskell, T.G., Frew, R.D., 2010. Observation and modeling of platelet ice fabric in McMurdo Sound, Antarctica. J. Geophys. Res. Oceans 115, C01007. http://dx.doi.org/10.1029/ 2008JC005264. Eicken, H., 1998. Deriving modes and rates of ice growth in the Weddell Sea from microstructural, salinity, and stable-isotope data. Antarct. Res. Ser. 74, 89–122. Epstein, S., Mayeda, T., 1953. Variation of O18 content of waters from natural sources. Geochim. Cosmochim. Acta 4 (5), 213–224. Fer, I., Makinson, K., Nicholls, K.W., 2012. Observations of thermohaline convection adjacent to Brunt Ice Shelf. J. Phys. Oceanogr. 42, 502–508. http://dx.doi.org/10. 1175/JPO-D-11-0211.1. Foldvik, A., Kvinge, T., 1974. Conditional instability of sea water at the freezing point. Deep-Sea Res. 21, 169–174. Gagliardini, O., Durand, G., Zwinger, T., Hindmarsh, R.C.A., Le Meur, E., 2010. Coupling of ice-shelf melting and buttressing is a key process in ice-sheets dynamics. Geophys. Res. Lett. 37, L14501. http://dx.doi.org/10.1029/2010GL043334. Gough, A.J., Mahoney, A., Langhorne, P.J., Williams, M.J.M., Robinson, N.J., Haskell, T.G., 2012a. Signatures of supercooling: McMurdo Sound platelet ice. J. Glaciol. 58 (207), 38–50. http://dx.doi.org/10.3189/2012JoG10J218. Gough, A.J., Mahoney, A.R., Langhorne, P.J., Williams, M.J.M., Haskell, T.G., 2012b. Sea ice salinity and structure: a winter time series of salinity and its distribution. J. Geophys. Res. 117, C03008. http://dx.doi.org/10.1029/2011JC007527. Gow, A.J., Ackley, S.F., Govoni, J.W., Weeks, W.F., 1998. Physical and structural properties of land-fast sea ice in McMurdo Sound, Antarctica. Antarct. Res. Ser. 74, 355–374. Grootes, P.M., Stuiver, M., 1986. Ross Ice Shelf oxygen isotopes and West Antarctic climate history. Quat. Res. 26, 49–67. Hattermann, T., Nøst, O.A., Lilly, J.M., Smedsrud, L.H., 2012. Two years of oceanic observations below the Fimbul Ice Shelf, Antarctica. Geophys. Res. Lett. 39, L12605. http://dx. doi.org/10.1029/2012GL051012. Heil, P., Allison, I., Lytle, V.I., 1996. Seasonal and interannual variations of the oceanic heat flux under a landfast Antarctic sea ice cover. J. Geophys. Res. 101 (C11), 25,741–25,752. Hellmer, H.H., Kauker, F., Timmermann, R., Determann, J., Rae, J., 2012. Twenty-firstcentury warming of a large Antarctic ice-shelf cavity by a redirected coastal current. Nature 484, 225–228. http://dx.doi.org/10.1038/nature11064. Hughes, K.G., Langhorne, P.J., Leonard, G.H., Stevens, C.L., 2015. Extension of an Ice Shelf Water plume model beneath sea ice with application in McMurdo Sound, Antarctica. J. Geophys. Res. Oceans 119. http://dx.doi.org/10.1002/2013JC009411. Intergovernmental Panel on Climate Change, 2013. Summary for Policymakers. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Jacobs, S.S., Fairbanks, R.G., Horibe, Y., 1985. Origin and evolution of water masses near 16 the Antarctic continental margin: evidence from H18 2 O/H2 O ratios in seawater. Antarct. Res. Ser. 43, 59–85. Jacobs, S.S., Giulivi, C.F., Mele, P.A., 2002. Freshening of the Ross Sea during the late 20th century. Science 297 (5580), 386–389. http://dx.doi.org/10.1126/science.1069574. Jeffries, M.O., Weeks, W.F., Shaw, R., Morris, K., 1993. Structural characteristics of congelation and platelet ice and their role in the development of Antarctic land-fast ice. J. Glaciol. 39, 223–238. Kawamura, T., Jeffries, M.O., Tison, J.-L., Krouse, H.R., 2004. Superimposed-ice formation in summer on Ross Sea pack-ice floes. Ann. Glaciol. 39, 563–568. Lange, M.A., 1988. Basic properties of Antarctic sea ice as revealed by textural analysis of ice cores. Ann. Glaciol. 10, 95–101. Langhorne, P.J., Hughes, K.G., Gough, A.J., Smith, I.J., Williams, M.J.M., Robinson, N.J., Stevens, C.L., Rack, W., Price, D., Leonard, G.H., Mahoney, A.R., Haas, C., Haskell, T.G., 2015. Observed platelet ice distributions in Antarctic sea ice: an index for ocean-ice shelf heat flux. Geophys. Res. Lett. http://dx.doi.org/10.1002/2014GL062978 (revised version under review). Leonard, G.H., Purdie, C.R., Langhorne, P.J., Haskell, T.G., Williams, M.J.M., Frew, R.D., 2006. Observations of platelet ice growth and oceanographic conditions during the winter

70

I.J. Smith et al. / Cold Regions Science and Technology 113 (2015) 63–70

of 2003 in McMurdo Sound, Antarctica. J. Geophys. Res. 111 (C4), C04012. http://dx. doi.org/10.1029/2005JC002952. Leonard, G.H., Langhorne, P.J., Williams, M.J.M., Vennell, R., Purdie, C.R., Dempsey, D.E., Haskell, T.G., Frew, R.D., 2011. Evolution of supercooling under coastal Antarctic sea ice during winter. Antarct. Sci. 23 (4), 399–409. http://dx.doi.org/10.1017/ S0954102011000265. Loose, B., Schlosser, P., Smethie, W.M., Jacobs, S., 2009. An optimized estimate of glacial melt from the Ross Ice Shelf using noble gases, stable isotopes, and CFC transient tracers. J. Geophys. Res. Oceans 114, C08007. http://dx.doi.org/10.1029/ 2008JC005048. Mahoney, A.R., Gough, A.J., Langhorne, P.J., Robinson, N., Stevens, C.L., Williams, M.J.M., Haskell, T.G., 2011. The seasonal appearance of ice shelf water in coastal Antarctica and its effect on sea ice growth. J. Geophys. Res. 116, C11032. http://dx.doi.org/10. 1029/2011JC007060. McPhee, M.G., Untersteiner, N., 1982. Using sea ice to measure vertical flux in the ocean. J. Geophys. Res. 87 (3), 2071–2074. Nicholls, K.W., Østerhus, S., Makinson, K., Johnson, M.R., 2001. Oceanographic conditions south of Berkner Island, beneath Filchner-Ronne Ice Shelf, Antarctica. J. Geophys. Res. Oceans C6 (106), 11481–11492. http://dx.doi.org/10.1029/2003JC002149. Paige, R.A., 1966. Crystallographic studies of sea ice in McMurdo Sound, Antarctica. Technical Report R494. US Naval Civil Engineering Laboratory, Port Hueneme, California, USA (31 page report). Pfirman, S., Haxby, W., Eicken, H., Jeffries, M., Bauch, D., 2004. Drifting Arctic sea ice archives changes in ocean surface conditions. Geophys. Res. Lett. 31, L19401. http:// dx.doi.org/10.1029/2004GL020666. Purdie, C.R., Langhorne, P.J., Leonard, G.H., Haskell, T.G., 2006. Growth of first-year landfast Antarctic sea ice determined from winter temperature measurements. Ann. Glaciol. 44, 170–176. Rienecker, M.M., Suarez, M.J., Gelaro, R., Todling, R., Bacmeister, J., Liu, E., Bosilovich, M.G., Schubert, S.D., Takacs, L., Kim, G.-K., Bloom, S., Chen, J., Collins, D., Conaty, A., da Silva, A., Gu, W., Joiner, J., Koster, R.D., Lucchesi, R., Molod, A., Owens, T., Pawson, S., Pegion,

P., Redder, C.R., Reichle, R., Robertson, F.R., Ruddick, A.G., Sienkiewicz, M., Woollen, J., 2011. MERRA: NASA's Modern-Era Retrospective Analysis for Research and Applications. J. Clim. 24 (14), 3624–3648. http://dx.doi.org/10.1175/JCLI-D-11-00015.1. Robinson, N.J., Williams, M.J.M., Stevens, C.L., Langhorne, P.J., Haskell, T.G., 2014. Evolution of a supercooled Ice Shelf Water plume with an actively-growing sub-ice platelet matrix. J. Geophys. Res. Oceans 119 (6), 3425–3446. http://dx.doi.org/10.1002/ 2013JC009399. Smith, I.J., Langhorne, P.J., Trodahl, H.J., Haskell, T.G., Cole, D.M., 1999. Platelet ice—the McMurdo Sound debate. Ice in Surface Waters, Proceedings of International Association for Hydraulic Research 14th International Symposium on Ice, Potsdam, New York. A. A. Balkema, Rotterdam, pp. 829–834. Smith, I.J., Langhorne, P.J., Trodahl, H.J., Haskell, T.G., Frew, R., Vennell, R., 2001. Platelet ice and the land-fast sea ice of McMurdo Sound, Antarctica. Ann. Glaciol. 33, 21–27. http://dx.doi.org/10.3189/172756401781818365. Smith, I.J., Langhorne, P.J., Frew, R.D., Vennell, R., Haskell, T.G., 2012. Sea ice growth rates near ice shelves. Cold Reg. Sci. Technol. 83–84, 57–70. http://dx.doi.org/10.1016/j. coldregions.2012.06.005. Thorndike, A.S., 1992. A toy model linking atmospheric thermal radiation and sea ice growth. J. Geophys. Res. 97 (C6), 9401–9410. http://dx.doi.org/10.1029/92JC00695. Toyota, T., Smith, I.J., Gough, A.J., Langhorne, P.J., Leonard, G.H., Van Hale, R.J., Mahoney, A.R., Haskell, T.G., 2013. Oxygen isotope fractionation during the freezing of sea water. J. Glaciol. 59 (216), 697–710. http://dx.doi.org/10.3189/2013JoG12J163. Trodahl, H.J., McGuinness, M.J., Langhorne, P.J., Collins, K., Pantoja, A.E., Smith, I.J., Haskell, T.G., 2000. Heat transport in McMurdo Sound first-year fast ice. J. Geophys. Res. 105 (C5), 11347–11358. Yin, J., Overpeck, J., Griffies, S., Hu, A., Russell, J., Stouffer, R., 2011. Different magnitudes of projected subsurface ocean warming around Greenland and Antarctica. Nat. Geosci. 4, 524–528. http://dx.doi.org/10.1038/NGEO1189.