2.1. Instrumentation

Figure 1 shows the deployment locations of the IMBs in the Bellingshausen and Amundsen Seas in October–December 2007 and February–March 2009, respectively. All IMBs used in the study were built by the US Army Cold Regions Research and Engineering Laboratory (CRREL) (Reference Perovich, Light, Eicken, Jones and RuncimanPerovich and others, 2007). The three IMBs, named ‘Brussels-1’, ‘Brussels-2’ and ‘Liege’, were deployed on a pack ice floe in the Bellingshausen Sea during the Sea Ice Mass Balance in the Antarctic (SIMBA) expedition (Reference LewisLewis and others, 2011; Reference XieXie and others, 2011). The general snow and ice thicknesses were 0.08–0.14m and 0.47–0.69m at the Brussels sites, and 0.25–0.70 m and 0.90–1.20 m at the Liege site, respectively (Reference LewisLewis and others, 2011). Before the icebreaker N.B. Palmer (NBP) left the station, the Liege IMB was retrieved and was redeployed on the NBP 2009 cruise at a fast-ice site in the Amundsen Sea. Concurrently with the IMB data collection at the Amundsen Sea site, an ice-tethered profiler (ITP) was installed on the ice mass on which the IMB was positioned (within 50 m of each other) and collected water-column temperature and salinity profiles on a sampling schedule of one round-trip profile between 7 and 760 m depth each day. The ITP data were relayed by Iridium satellite and are available from the Ice-Tethered Profiler Program based at the Woods Hole Oceanographic Institution, MA, USA (http://www.whoi.edu/itp, ITP31; Reference Toole, Krishfield and ProshutinskyToole and others, 2011). This site was stationary (‘fast ice’) from 6 February to 10 March 2009, detached from the land and drifted from 10 to 14 March 2009 (total distance of 60 km), then stopped until drift resumed on 18 March 2009 and continued briefly through the end of transmission on 19 March 2009. Table 1 shows the parameters and times when data were collected for all stations. Each IMB measured geographic position (Garmin GPS16-HVS), surface-elevation changes (Campbell SR50A-L24 acoustic sensor mounted on a surface mast ∼2 m high), changes of the bottom ice surface (Benthos PSA-916 sonar altimeter mounted on an underwater mast 1-2 m below the ice bottom, within a 14° beamwidth and at 0.01 m resolution), and the vertical temperature profile that extended from the air, through the snow, ice and surface water to 1-2 m below the ice at 0.05-0.10 m intervals (YSI-44033-BP thermistors, at 0.01°C resolution). The IMB at Brussels-1 also measured irradiance at the ice under-surface (not reported on here) (Satlantic OCR-504 4 channel underwater radiometer on an L-arm looking upward at ∼1.2m depth), while the IMBs at Brussels-2 and the Amundsen Sea site were equipped to measure conductivity, temperature and depth (CTD) below the ice (Sea-Bird MicroCAT SBE 37-SI with pressure sensor mounted on an underwater mast ∼1.2 m below the ice bottom, at 0.002°C temperature resolution and 0.001 psu salinity resolution). All IMB sensors were autonomously sampled at 30min intervals. Data were saved in a data logger (Campbell CR1000-ST-SW-NC), transmitted and downloaded from the Advanced Research and Global Observation Satellite (ARGOS) system. The sensors and IMBs experienced life spans varying from 18 days (minimum) to 75 days (maximum).

Fig. 1.

IMB tracks (a) from October to December 2007, deployed during NBP 2007 (SIMBA) on a floe in the pack ice of the Bellingshausen Sea and (b) from February to March 2009, deployed during NBP 2009 on fast ice in the Amundsen Sea.

Table 1.

Duration of IMB data records by data type and site

2.2. Parameterization of bulk OHF from ocean measurements

Water temperature elevation (or temperature elevation hereafter), defined as the deviation of the sea-water temperature above the freezing point at the associated seawater salinity, has been identified as the measure of the available heat content in the mixed layer beneath the ice (e.g. Reference JosbergerJosberger, 1987; Reference Morison and McPheeMorison and others, 1987; Reference Wettlaufer, Untersteiner and ColonyWettlaufer and others, 1990; Reference WettlauferWettlaufer, 1991; Reference OmstedtOmstedt and Wettlaufer, 1992). Reference Morison and McPheeMorison and others (1987) parameterized the OHF, using the temperature elevation, the friction velocity and a constant equal to the bulk heat transfer coefficient (C _H) divided by the density (p) and specific heat (c _w) of sea water. Reference JosbergerJosberger (1987) showed the bulk heat transfer coefficient varied little over a wide range of ice-melting conditions. After Reference MorisonMorison (1995) and Reference McPhee, Ackley and KottmeierMcPhee and others (1999), C _Hρc _w ∼700 W m^–2 ° C^–1 m^–1 s) (C _H = 1.7 x 10^–4ρ= 1024kgm ^–3, c _w = 3980J kg^–1 °C^–1), derived also from various experiments and modeling for melting sea ice in the Arctic (e.g. Reference JosbergerJosberger 1987; Reference Morison and McPheeMorison and others, 1987; Reference OmstedtOmstedt and Wettlaufer, 1992), is used here. By direct comparison with turbulence measurements, Reference McPhee, Ackley and KottmeierMcPhee and others (1999) found that this parameterization of heat flux from sea-water temperature deviation is in good agreement with measurements derived from turbulence measurements and also showed that friction velocity could be parameterized by ice-drift speed. Equation (1) (called OHF₁ hereafter) with this value (700) is used to determine the OHF (Wm^–2):

(1)

where u _i (m s–¹) is the relative speed of the ice mass, T (°C) is the temperature of the surface water and T _f ^(°C) is the

(2)

where S is the surface water salinity and p (dbar) is pressure (Reference FujinoFujino and others, 1974)

Water salinities, computed from measurements of conductivity, temperature and pressure by a CTD measuring device suspended at 1.2 m below the ice (sea-water equation of state), were used to calculate the freezing points (Eqn (2)). The freezing point is subtracted from the measured temperature to determine the water temperature relative to its freezing point (i.e. the water temperature elevation), as a measure of the available heat content at the ice/water interface (Reference McPheeMcPhee and others, 1996; Reference Sirevaag, McPhee and MorisonSirevaag and others, 2010). However, the Bellingshausen Sea CTD data were confined to Brussels-2, and also ceased transmission on 1 November 2007, well before the end of the drift record on 6 December 2007 (Table 1 ). In order to extend the CTD sea-water temperature record, we instead use the thermistor records as a proxy. All of the IMBs were equipped with thermistors measuring surface water temperature under ice. The precision of the thermistors was, however, only 0.01 °C (due to the loss of the last place by transmission of only an 8-bit number) vs 0.002°C for the CTD temperature. Any single thermistor record was therefore too coarse in resolution to estimate variations in OHF. However, since there were 14-21 thermistors positioned in the uppermost 1.5– 2 m of the surface water under ice, we are able to increase precision by averaging. The standard deviation is reduced to only ±0.005°C for these combined records and therefore provides a more precise measure of temporal water temperature variations. Since the IMBs on the moving Belgica floe were in sufficiently close proximity (within a few hundred meters of each other) to reasonably assume that they were exposed to the same water masses, the average of the underwater thermistor measurements of water temperature from each of the IMBs is calibrated to the more accurate CTD measurements from Brussels-2 when both measurements were taken, and compared with that record. When corrected for a constant bias for each thermistor record, computation of the heat flux, shown later, gives a ˂5% deviation from that computed from the CTD temperature.

Although salinity and pressure measurements are necessary to calculate the surface water freezing point (needed in Eqn (1)), the effect of the observed small salinity and pressure fluctuations on the freezing point as used in the OHF calculated from the CTD measurements is small compared to the effect of water temperature variations. Therefore, the average salinity and pressure measured by the CTD, up until 1 November 2007 when it failed, is used to estimate the surface water freezing point during the periods lacking a CTD record after 1 November 2007. The temperature above its freezing point using the average thermistor string records where CTD data are not available is then computed for the full length of the record in the Bellingshausen Sea and is used in estimating the OHF for the full duration (2 October to 6 December 2007) for all three sites in the Bellingshausen Sea. Errors in the OHF introduced by using the average freezing point, without accounting for the small fluctuations in freezing-point temperature with measured salinities, are within the 5% error introduced by using the thermistor temperatures instead of the CTD precision water temperature. Error propagation is further discussed in Section 4.1.

2.3. OHF computed from ice freezing or melting

When air temperatures are sufficiently low and the ice/water interface is not in thermal balance, heat may be conducted through the ice and snow, lost to the atmosphere, thereby establishing the potential for ice growth (Reference Weeks and UntersteinerWeeks and Ackley, 1986; Reference LytleLytle and Ackley, 1996). During such conditions, from the steady-state heat balance, the OHF (Reference Maykut and UntersteinerMaykut, 1986) can be written as

(3)

where ki is the thermal conductivity of ice (Eqn (4)) (Reference OnoOno, 1968), ∂T/∂z (°Cm^–1) is the within-ice temperature gradient at the ice bottom surface, p _i is the density of the sea ice (assumed 920 kgm^–3; Reference Weeks and UntersteinerWeeks and Ackley, 1986), L is the latent heat of fusion (Eqn (5); Reference OnoOno, 1968), ∆t (ms–¹) is the measured rate of ice thickness change and A is the area (A = 1 m² in this context).

(4)

(5)

where k0 is the temperature-dependent pure ice thermal conductivity (however, a constant value of 2.1 Wm–¹ (Reference YenYen, 1981) is assumed in our calculations over the narrow temperature range used here), Si is the salinity of the ice (used 5 according to ice-core records from Reference LewisLewis and others, 2011) and Ti is the temperature of the ice at the measurement level.

Except for early March 2009 for the Amundsen Sea buoy, however, the temperature gradients near the ice bottom approached zero (Reference LewisLewis and others, 2011), since surface air temperatures alternated between warm and cold episodes and the snow cover was consistently deep enough to insulate the bottom layers from the lowest air temperatures and therefore prevented heat conduction between ocean and atmosphere through the ice. Ice at the bottom was instead usually melting, under the influence of the OHF alone. Equation (3) can therefore be simplified, without the conductive term (called OHF₂ hereafter):

(6)

The average OHF is used to melt the ice at a particular melt rate ∆t (m s^–1). The melt rate can be determined as the slope of the plot of ice thickness (net loss) with time.

The average OHF values from method OHF₁ are compared with the average flux (OHF₂) derived from the ice melt rate in this study.

3.1. Bellingshausen Sea

3.1.1 OHF derived from ocean measurements (OHF₁)

Figures 2-4 show the results from the Bellingshausen Sea IMBs in 2007 for Liege, Brussels-1 and Brussels-2, respectively. The average OHF₁ measurements are in close agreement at 6, 7 and 8Wm^–2, respectively, with the small differences accounted for by different record lengths from 2 October to 6 December 2007 for the three buoys (Table 1 ). Brussels-2 provides the longest record of OHF, which included a spike of ˃55Wm^–2 in the final days (December 2007). Ice-floe drift speeds as great as 0.6-0.7 ms–¹ were measured. Water temperature elevations spiked to ∼0.13°C intermittently during October and November, and increased to ˃0.24°C at the end of the Brussels-2 record in early December 2007. These higher surface water temperatures and highest OHF₁ in early December 2007 also coincided with ˃0°C air temperatures (not shown).

Fig. 2.

Liege ocean heat flux from Eqn (1) (OHF₁ method; mean of 6 W m^–2) (a), ice-floe drift speed (b) and water temperature elevation based on calibrated thermistor data (c), from 5 to 23 October 2007.

Fig. 3.

Brussels-1 OHF from Eqn (1) (OHF1 method; mean of 7Wm–2) (a), ice-floe drift speed (b) and water temperature elevation based on calibrated thermistor data (c), from 2 October to 1 November 2007.

Fig. 4.

Brussels-2 OHF from Eqn (1) (OHF₁) and ice thickness change (a), ice-floe drift speed (b) and water temperature elevation based on CTD (up to 1 November 2007) and calibrated thermistor data (after 1 November 2007) (c), from 12 October to 6 December 2007. The mean OHF of 8 W m^–2 calculated from Eqn (6) (OHF₂) based on ice thickness time series (melt rate), also shown in (a), matched well with the mean OHF of 8Wm ^–2 using the OHF₁ method for the same period, 12–29 October 2007.

3.1.2. OHF2 derived from ice melt rate

The ice thickness time series (Fig. 5) shows ice thickness much larger at the Liege site than at the two Brussels sites. Under-ice photographs (http://www.utsa.edu/lrsg/Antarctica/SIMBA/Pictures/album/Under%20ice%20photos/index.html) indicate that the bottom surface of ice at Brussels-1 was relatively level, but scalloped (0.01–0.02m variation) due to ice melt, while the under surface at Liege(a) had a complex topography resulting from deformation and melt events, with ice thickness spiking 0.5 m, suggesting the upward-looking sonar periodically had returns from an irregular feature instead of the leveler ice. This phenomenon was also seen at an Arctic fast-ice site (Reference WangWang and others, 2013). The sensor was then relocated to Liege(b) on 12 October 2007, which largely reduced this noise in the ice thickness record. Therefore, only Liege(b) data are used to calculate the OHF using the OHF₂ method. The top and bottom surfaces at the Brussels-1 and Brussels-2 sites were uniform and similar (not shown). The average OHF derived from the ice melt rate (OHF₂) at Brussels-2 in the Bellingshausen Sea during 12-19 October 2007 was ∼8Wm^–2 (Fig. 4a), which matches well with the mean OHF₁ result from ocean temperature and salinity for the same time. The total derived OHF₂ for Brussels-1 is 12 Wm–² (2 October to 6 November 2007), larger than the 7Wm–² from OHF₁.

This difference was probably due to the difference in time, since the latter was only from 2 October to 1 November 2007 (Fig. 4). The derived OHF2 for the Liege(b) site is 19Wm^–2 (12–22 October 2007), much larger than the 6Wm^–2 (OHF₁) for 5–23 October 2007 (Fig. 4). This is a large difference and might be due to (1) difference in time and short overlap between the records, and (2) the rough bottom surface that resulted in larger error in ice thickness measurements as seen in Liege(b) (Fig. 5). Although the thickness variation in Liege(b) was much less than in Liege (a), the change of 0.2 m in 10 days (12–22 October 2007) in Liege(b) was still much larger than that seen from the Brussels sites. Therefore, the 19 Wm–2 appears to be a much overestimated OHF value for the area, with rough bottom surface that caused large measurement variations in thickness change probably the primary cause.

Fig. 5.

Ice thickness time series from upward-looking sonar at the three IMB sites of the Belgica Floe, September to December 2007. The sensor at Liege(a) was relocated to Liege(b) on 12 October 2007.

3.2. Amundsen Sea

Since the ice was stationary most of the time at the Amundsen Sea site, the ice-drift speed (zero) cannot be used to estimate the friction velocity using the OHF₁ method. Therefore, the OHF₂ ice thickness change method is used. Figure 6a shows the full ice thickness time series, with a mean ice melt rate of 0.0054 md^–1corresponding to an OHF of 14Wm^–2. During short-lived events, the melt rate varies from 0.0049 to 0.0063 m d^–1corresponding to OHF of 12–16Wm^–2with a short exception where the OHF is almost 0 due to a melt rate close to 0 (not shown). Figure 6c shows the sea-water temperature elevation above its freezing point, which is overall much larger than that in the Bellingshausen Sea (Fig. 2–4). For the drift from 10 to 14 March 2009, the OHF is also calculated using the OHF₁ method (Fig. 6a). The maximum OHF was up to 55 Wm– 2, with a mean ∼17Wm–² . This mean value therefore generally agreed with the OHF₂ (12-16Wm^–2) computed when the ice was stationary.

Fig. 6.

Fast-ice site at Amundsen Sea (February to March 2009): (a) ocean heat flux (OHF₂) for the entire period derived from ice thickness melt rate, and instantaneous OHF (OHF₁) for the drift period (10–14 March 2009); (b) drift speed for the drift period; and (c) under-ice water temperature elevation based on calibrated thermistor data for the entire period.

4.1. Error propagation on OHF calculations

Close agreement is found between the average OHF₂ derived from the smooth ice-thickness profile at Brussels-2 (∼8Wm^{– 2} ) and the average of OHF₁ derived from the salinity and temperature measured on the same IMB over the same time interval (also 8Wm^{– 2} ) in the pack ice of the Bellingshausen Sea. The latter method using ocean temperature and conductivity values (OHF1) can, however, show short-term temporal variability that is not detectable by the ice melt rate method (OHF₂) at the given data resolution.

Error in a single heat-flux value can arise from three sources: calculated velocity, measured temperature and estimated freezing-point temperature. The error in measured temperature of ±0.01°C leads these, as the freezing point, determined by the salinity, does not vary much since the full range of salinity for these waters is ˂1 psu. Errors in velocity arise from differencing GPS positions that are closely spaced in time, so the error in an individual heat flux calculation can be up to an estimated 2 Wm^{– 2}, or 25% of the average heat flux. However, we are still using a bulk parameterization of the heat flux, so it is only an approximation to the ‘real’ (turbulent) heat flux in any case, so in our view it is not so important to quantify the errors in an individual measurement. The time-averaged bulk parameterization of the heat flux, derived from ice velocity, ocean temperature and salinity, agrees well with the heat flux given by the change in ice thickness. The agreement suggests the individual measurement (bulk heat flux or ice thickness change) does not propagate but is instead smoothed out when time-averaged.

4.2. Temporal variations of OHF correlate with wind

The validated parameterization demonstrates the effect of ice-drift speed on calculating OHF in this location. Ice-drift speed can be highly variable, with rapid acceleration caused by intermittent changes in wind speed and direction (Reference GeigerGeiger and others, 1998; Reference Worby, Massom, Allison and HeilWorby and others, 1998). Storm events and sharp wind fronts may therefore increase OHF under pack ice by means of added turbulent stress at the ice/ water interface (Reference GordonGordon and Huber, 1984; Reference McPhee and UntersteinerMcPhee, 1986; Reference McPheeMcPhee and others, 1987; Reference MorisonMorison, 1995). The net drift of pack ice in the Bellingshausen Sea is largely wind-driven. However, studies (Reference GeigerGeiger and others, 1998; Reference Worby, Massom, Allison and HeilWorby and others, 1998; Reference HeilHeil and others, 2008, Reference Ackley, Lewis and Xie2009) also measured peaks in drift speed at ∼12-13hour intervals, which is approximately the semi-diurnal tide period and near the inertial period at the given latitudes. Therefore, while wind is a significant force on ice drift, the effects of inertial/tidal oscillations can dominate the drift pattern at semi-diurnal timescales (Reference GeigerGeiger and others, 1998). Although the semidiurnal drift pattern is not as strong as seen in the Arctic summer sea ice, the drift speed itself in the Bellingshausen Sea is larger than in the Arctic (0.17±0.08 ms^–1 ) (Reference XieXie and others, 2013). Consequently, where these variations noticeably affect drift speeds, particularly from mid-November to early December (Fig. 4b), they should also be expected to affect turbulent properties in the surface water beneath the ice and vertical shear across the pycnocline (Reference McPhee and UntersteinerMcPhee, 1986; Reference McPheeMcPhee and others, 1987; Reference MuenchMuench and others, 2001).

The roughly sinusoidal OHF shown for the Bellingshausen Sea is not seen at the fast-ice site in the Amundsen Sea due to zero drift speed during the stationary status. However, after the fast ice broke away from land, its drift pattern showed semi-diurnal peaks (Fig. 6b) (over a short record) similar to those seen in the Bellingshausen Sea. The drift speeds were well above the expected magnitude of tidal currents in the study area. Tidal speeds are only ∼0.01 – 0.02 ms^–1 during neap tides and 0.05 ms^-1 during spring tides, based on the Circum-Antarctic Tidal Simulation version 2008b (CATS2008b) barotropic tide model, an update to the model described by Reference Padman, Fricker, Coleman, Howard and ErofeevaPadman and others (2002). Therefore, wind-driven currents and inertial oscillations could have been the primary suppliers of energy to the turbulent mixing processes under the sea ice in the Amundsen Sea, similar to the Bellingshausen Sea.

However, seasonal difference is expected (Allison, 1979). Mean OHF of 6-8 Wm^{– 2} in the Bellingshausen Sea is much less than the 12-16Wm^{– 2} in the Amundsen Sea. Indeed, even the presence of inertial oscillations in drift patterns is expected to vary seasonally, being especially evident during the melt season mostly because ice becomes too weak to dampen and resist the transmitted stress (Reference McPhee and UntersteinerMcPhee, 1986).

Under the pack ice in the Bellingshausen Sea, OHF measurements peaked (up to 55 Wm^{– 2} ) in December after the ice floe had drifted into the marginal ice zone. It is possible that shortwave solar radiative heating of the surface water through opening leads contributed to the elevated surface water temperatures that contributed to the increased OHF at that time (Reference Perovich, Light, Eicken, Jones and RuncimanPerovich and others, 2007), although the drift carrying the ice floe over a new water mass with different characteristics cannot be discounted (Reference GordonGordon and Huber, 1995).

4.3. Amundsen Sea: water stratification under fast ice

The water stratification under fast ice is clearly visible (Fig. 7a), with the colder water found on the top of the ∼300 m column. The stratification of salinity is not so clear, although less saline water was mainly in the top 200 m (Fig. 7b). Comparisons between the CTD measurements at 1-2 m depth and the uppermost ITP CTD measurements at 8 m below the fast ice suggest, however, the very top surface water was colder and fresher (Fig. 7c and d) than that found below. As well, the CTD salinity measurements did not mirror the ITP salinity measurements at 8 m (Fig. 7d). In fact, in most cases, salinity spikes at 8 m are generally followed by salinity drops at 1-2 m. We suspect that warm, saline water ascending from the deep warmer water below caused melt events which lowered the salinity in the uppermost 1 or 2 m of the surface water just under the ice. This may also account for the slightly lower temperatures measured directly below the ice. In particular, the large and wide salinity spike from 7 to 17 March 2009 seen in the ITP 8 m profile matches well with the increase in temperature at both 1-2 m and 8m, and the large decrease in salinity at 1-2 m, presumably due to ice melt. Persistent stratification may have been possible under the fast ice because drift was not contributing to turbulent mixing. Furthermore, a smooth ice-thickness time series measured by the sonar pinger (Fig. 6a) indicates that the under-ice surface was relatively level, perhaps also enhancing the stratification (Reference McPhee and UntersteinerMcPhee, 1986). When the ice mass broke away and began to drift (18-19 March 2009), both the temperature and salinity values in the upper 10m from the two sensors began to converge (Fig. 7c and d), suggesting the water was mixing rather than remaining stratified.

Fig. 7.

ITP31 temperature profile (a) and salinity profile (b) by Julian day in 2009 at the fast-ice site in the Amundsen Sea. (c, d) Temperature (c) and salinity (d) measurements at the uppermost ITP (8m depth) with the corresponding IMB’s CTD measurements positioned at 1–2m depth.

Another factor that may have led to changing surface water stratification in March was possible ice growth, rather than melt, during the final 2 days of the record, as indicated by small positive ice thickness changes (Fig. 6a). Moreover, at the end of the record the required conditions for ice growth were measured by a bottom temperature gradient within the ice, conducting heat away from the previously isothermal bottom layer (Reference AllisonAllison, 1981; Reference Maykut and UntersteinerMaykut, 1986). Figure 8 clearly shows that the bottom four depths (-0.25, -0.35, -0.40 and -0.50 m) of the ice were mostly isothermal, with temperature around -1.7°C, and were gradually melted away, with the first bottom (-50 m) on 15 February (melted away), the second bottom (-40 m) on 9 March and the third bottom (-0.35 m) on 15 March. After 15 March, the fourth bottom was at-0.25 m and was once at the isothermal status with the -0.20 m for ∼2 days. But this isothermal status was rapidly broken on 17 March. A clear temperature gradient between the fourth bottom (-0.25 m, ∼–2°C, at freezing point) and above (-0.2 m, ∼–3°C or smaller, below freezing point) then clearly appeared, favorable for ice growth, continuing for the next 2 days, 18 and 19 March.

Fig. 8.

Temperature gradient within the ice measured by a fixed thermistor string at the fast-ice site in the Amundsen Sea, February to March 2009. The vertical positions of the thermistors are referenced relative to the snow/ice interface at 0.00 m (the lowest temperature time series on the top, air/snow interface).

Therefore, the end of under-ice melt and onset of ice growth marked the seasonal shift to autumn conditions in the under-ice water regime, lagging cold air temperature changes in the region by only a few days since the ice was ∼0.25 m thick at this time. Moreover, as sea ice freezes, the winter pycnocline is destabilized by brine-initiated convection so that deep water is then more susceptible to entrainment (Reference GordonGordon and Huber, 1984). For example, brine rejection from rapidly growing ice was observed in March (Reference AllisonAllison, 1981), so brine rejection during freezing may have led to the onset of high convective heat flux that was measured in the Amundsen Sea near the end of the record (Fig. 7c). The convergence of temperature and salinity values measured at 1-2 m depth from the IMB and 8m depth from the ITP (Fig. 7c and d) is also consistent with mixing initiated by brine rejection during ice growth in the last few days of the record.

Unlike the Amundsen Sea summer, temperature and salinity time series at depth below the pack ice in the Bellingshausen Sea during spring, also collected from ship-based CTD casts of the SIMBA experiment, did not indicate uppermost surface water stratification under the drifting pack ice and generally showed a ‘well-mixed’ winter mixed layer with fairly uniform temperature and salinity, down to depths ˃50m (Reference LewisLewis and others, 2011). One possibility is that the turbulent stress caused by drifting pack ice over a large area may have eroded the potential for uppermost surface water stratification caused by limited melt processes under the pack ice. Brine rejection below the ice was also observed by underwater photos (http://www.utsa.edu/lrsg/Antarctica/SIMBA/Pictures/album/Under%20ice%20photos/) at the IMB sites, caused by freezing in the upper layers (only) of the ice during cold episodes, and new ice growth was also observed in lead areas nearby during the ship drift. Therefore, freshening by melt may have been balanced by salinization from this brine rejection and coincident growth of thin ice during spring 2007 in the Bellingshausen Sea.

4.4. Comparison with previous findings

Several previous studies have reported a large range of OHF in the Weddell Sea, 1.7-41 Wm^{– 2} on average, with a maximum value of ˃150Wm^–2 (Table 2 ). In the western Weddell Sea, similar values (7-8Wm^–2) to those shown here for the Bellingshausen Sea are found. In the Bellingshausen Sea, we measured average water temperature elevations of only 0.06°C. A maximum water temperature elevation of 0.24°C was measured only in early December 2007 when it is likely that solar heating in the lesser ice concentration of the marginal ice zone contributed to the heat content of the surface water. This rapid increase in OHF (up to 20Wm^–2) was also reported at the fast-ice site of Prydz Bay, East Antarctica, in mid-December 2006 and was also attributed to possible warm deep water intrusion and increased solar heating (Reference Lei, Li, Cheng, Zhang and HeilLei and others, 2010).

Table 2.

Under-ice ocean heat flux (OHF) estimates determined by various studies in the Southern Ocean

The most extreme OHF (˃150Wm^–2) measured at the end of the record reported by Reference MorisonMorison (1995) was also associated with drift into open water. In the Amundsen Sea we measured average temperature elevation of 0.14°C (over twice the average value (0.06°C) in the Bellingshausen Sea) and a maximum of 0.35°C. The temperature elevations that we measured at the fast-ice site in the Amundsen Sea were of similar magnitude to that measured by Reference MorisonMorison (1995), perhaps an indication of the existence of upwelling effects on the continental shelf in the Amundsen Sea. Similarly, a typical temperature elevation of ∼0.1°C was measured during the 1986 austral winter in the eastern Weddell Sea (Reference GordonGordon and Huber, 1990), which was greater than what we measured under the pack ice in the Bellingshausen Sea and closer to what we measured under the fast ice in the Amundsen Sea.

A consistent feature between previous results in the Weddell Sea and the results of our study was the detection of short-term average surface water current speeds that were presumably due to semi-diurnal inertial/tidal oscillations (Reference MuenchMuench and others, 2001; Reference Sirevaag, McPhee and MorisonSirevaag and others, 2010).

4.5. The driving forces of OHF

Ocean heat flux may be constrained by both the availability of heat and the intensity of mechanical mixing at the ice/ water interface (Reference MuenchMuench and others, 2001). In the Bellingshausen Sea, we find that water temperature elevations are not correlated with measured drift speeds. This is consistent with what was found in the Weddell Sea (Reference MuenchMuench and others, 2001). Reference MartinsonMartinson and Iannuzzi (1998) found that drift speed and surface water temperature elevation vary independently, although the average of their combined effect on OHF may be less variable. During summer, when there is no sea-ice freezing process to upset the density balance in the surface water, a greater proportion of OHF may be attributed to diffusion across the pycnocline (Reference GordonGordon and Huber, 1990; Reference Hohmann, Schlosser, Jacobs, Ludin and WeppernigHohmann and others, 2002). Autumn freezing in the Amundsen Sea initiated greater convection between the deep and shallow waters (Reference GordonGordon and Huber, 1990). Hence, both shallower bottom topography and seasonal freezing may have initiated greater entrainment of deep water under the fast ice in the summer/autumn transition in the Amundsen Sea as compared to beneath the spring pack ice in the Bellingshausen Sea. Consequently, turbulent stress as a means to transport heat across the pycnocline may have been more important in the Bellingshausen Sea. This is consistent with the greater drift speeds measured in the Bellingshausen Sea and the greater water temperature elevations measured in the Amundsen Sea.

The OHF in the Amundsen Sea is ∼17Wm^–2with the primary source of heat flux attributed to the deep-water upwelling. The deep water is also the source of heat at several hundred meters depth to the base of icebergs and the ice shelves. The surface heat flux found here would be commensurate with a melt loss rate of 2-3 micea–¹ under nearby icebergs and ice shelves. This melting rate is compatible with tracer estimates of the significant freshwater flux from glacial melt into the southeast Pacific sector bordering the Amundsen and Bellingshausen Seas (Reference Jacobs and JenkinsJacobs and others, 1996; Reference Jenkins, Vaughan and JacobsJenkins and others, 1997; Reference Hellmer and JenkinsHellmer and others, 1998; Reference Hohmann, Schlosser, Jacobs, Ludin and WeppernigHohmann and others, 2002).