Supercooling at Freeman Sound, Svalbard

Miles McPhee

3 Apr 2007

1. Background

During the recent cruise aboard the Norwegian Coast Guard icebreaker K/V Svalbard, we had the opportunity to deploy two turbulence instrument clusters (TICs) during three short on-ice stations at different locations in or near the Svalbard archipelago. The second station was situated near the edge of fast ice approximately 45 cm thick in the northern end of Stor Fjord (between Spitsbergen and Edge Islands), near the mouth of Freeman Sound. We measured there at two lev­els, 1 and 3 m below the ice/water interface, for about 20 hours, or slightly more than 1¹/₂ tidal cycles. This is an area of much interest because the combination of strong tidal action and cold northerly winds often forms a polynya capable of producing exceptionally dense water that is both highly saline and cold. In winter the water is nearly always near its freezing temperature, and may become supercooled. In the measurements we made during our brief station in Freeman Sound, we observed what appear to be supercooled conditions, but also noticed a response of the SBE 04 conductivity meters in those conditions that may bear on other discussions of supercool­ing observed in polar waters, based on measurements made with Sea-Bird and similar conductivi­ty measuring instrumentation.

2. Measurements

The two TICs differed in some respects. Both included a Sontek ADVOcean three-dimen­sional, backscatter acoustic current meter with its measuring volume in the same plane as fast re­sponse SBE 3F thermometers and standard SBE 4 conductivity meters. For the upper cluster (1 m below the ice), the T/C combination was pumped, in the standard SBE configuration. The upper cluster also included an open electrode SBE 07 microstructure conductivity sensor (with prewhit­ening circuitry removed). The lower cluster was not pumped but relied on the flushing of the con­ductivity duct by mean flow, which was maximized by orienting the cluster so that the duct aligned with the major tidal axis. Microstructure conductivity measurements were lacking from the lower cluster.

Data, averaged in one-minute bins, are shown in Fig. 1. The tidal flow (Fig. 1A) was ener­getic (more than an order of magnitude greater than tidal flow measured 2 days earlier in VanM­ijen Fjord), with peak speeds approaching 0.9 m s^-1. The positive velocity indicates NNE flow into Freeman Sound, i.e., advection of water from outside the fast ice zone toward the instru­ments. As the peak in the tidal velocity was approached before time 82.1 (time is indicated in dec­imal days of 2007, day 82 is 23 Mar 2007), salinity began decreasing and temperature rose slightly. During times when we are reasonably sure that the water column was well mixed, tem­peratures from the two SBE 03F thermometers match to within 1 mK. Conductivity, however, dif­fered by enough to cause a salinity difference of about 0.03 psu. The C sensor in the upper cluster (pumped) was calibrated just before the project, while that in the lower cluster was last calibrated in 2003. We assume that conductivity measured by the upper cluster is accurate.

Shortly before 82.1 (~02:15 UT), we observed a sudden drop in conductivity at the upper cluster indicating a change in salinity of about 0.5 psu (blue trace in Fig. 1C) which was not present at 3 m (green trace). This persisted for some time, then conductivity at 1 m returned to values close to those observed at 3 m. Then about 15 min after the anomaly at TIC 1 disappeared, the unpumped C sensor at 3 m indicated a similar, nearly instantaneous, drop in conductivity indi­cating a decrease in salinity exceeding 1 psu. These events are illustrated in greater detail in Fig. 2. At first, we interpreted the sudden drop in conductivity at 1 m as evidence of fouling by some foreign matter, perhaps biofouling or frazil crystals. But generally, our experience has been that when this occurs there is a rather drastic change in conductivity, leading to nonsensical values for salinity. This is clearly not the case here, and the fact that a similar event occurs in roughly the same time frame, 2 m lower in the water column, suggests a somewhat repeatable, rather than ran­dom, phenomenon.

Taken at face value, the sudden drop in conductivity also indicates a significant degree of supercooling, if the resulting salinity is used to determine the freezing temperature, as shown in Fig. 3. Freezing temperature at surface pressure is calculated from salinity using UNESCO formula as given by Gill (1982, Atmosphere and Ocean Dynamics, p. 602). We doubt, however, that the decrease in salinity is real, because this is an extremely turbulent environment: Reynolds stress at 1 m approached 1 Pa during this time (i.e., u_* ~ 0.03 m s^-1). It is hard to conceive how real salinity gradients as large as implied by the difference in conductivity between the two clusters could be maintained for any length of time. Hence the magnitude of supercooling is probably greatly overstated.

The red trace in Fig. 3 is based on salinity at TIC 2, as adjusted by a constant offset of 0.028 psu described above. Note the downward trend in DT during the first part of the sample at both levels (Fig. 3B), despite a rise in actual water temperature (Fig. 1). The thermometer read­ings differ by about 1 mK, which is probably smaller than either sensor’s calibration accuracy. If in fact the 1-m thermometer agreed with the lower sensor, water at 1 m would reach freezing co­incidentally with the sudden drop in conductivity. A linear fit to DT for the 3 m cluster during the time interval indicated by the arrow in Fig. 3B provides a slope of about -1.8 mK h^-1. The change in freezing temperature for a pressure difference of 2 db is approximately -1.5 mK. So if indeed the upper water column was thoroughly mixed (as we would expect with the high turbulence lev­els), given the downward trend in DT, water at 3 m would reach freezing at a time later than at 1 m, given by -1.5/-1.8 = 0.84 h. This is quite close to the time between the onset of the sudden drops in conductivity observed at the two levels. While perhaps fortuitous, the observed time lag is consistent with a well mixed water column that reaches freezing at different times for different pressures.

3. Discussion

Several possible reasons for the behavior of the conductivity sensors observed under the Freeman Sound fast ice exist. First, it could result from random contamination of the conductivity ducts in the SBE 04 instruments by biological or other fouling. This seems unlikely, because it oc­curs at both depths within about an hour, persists for about the same amount of time, and does not disturb the conductivity enough to indicate salinity radically different from the ambient surround­ings.

A second more plausible cause could be advection of frazil crystals in the tidal flow from the upstream open water area, where it is likely that if supercooled water existed, blowing snow would provide adequate near surface nucleation sites for substantial frazil production. The crys­tals would be advecting toward the instrument clusters and mixed downward by strong turbu­lence. Two factors argue against this interpretation. First, the change in conductivity is considerably smaller than one might expect if the duct in either the pumped or free-flowing in­strument was clogged by frazil crystals. Second, because of shear in the under-ice boundary layer, we would expect advection of frazil crystals to reach higher concentrations at the lower level first. Instead, the conductivity drop occurs at 1 m about an hour before it is seen at 3 m.

A third possibility is that water from outside the fast ice area is slightly lower in salinity (as suggested by the downward trend in salinity with time when velocity is positive, and vice ver­sa— Fig. 1), but remains above freezing until it mixes with slightly saltier water during the posi­tive inflow part of the tidal cycle. If there is any double diffusive tendency in the mixing (i.e., if heat mixes faster than salt, as it does at molecular scales), the fresher water (which is also slightly warmer) will lose heat faster than it gains salt, and may thus become temporarily supercooled. With regard to the SBE 04 instruments, this might induce a thin coating of ice in the instruments’ ducts that would account for the unrealistically low conductivity.

Presuming the last hypothesis to hold, the instruments would indicate the presence of su­percooling by a sudden drop in conductivity, but would not accurately describe its magnitude. The upper TIC during the Freeman Sound deployment included an open-electrode, microstructure conductivity probe (SBE 07). For these instruments, conductivity is almost linear with output fre­quency, and in situ calibration coefficients were estimated by linear regression of TIC 1 standard conductivity against mC frequency during the times shown in Fig. 2B after excluding the negative anomaly. One-minute averages of salinity thus calculated from the mC conductivity are compared with salinity from the pumped standard C cell in Fig. 4A. It is clear that the open-electrode instru­ment does not sense the same sudden drop in conductivity measured by the pumped standard cell. Nevertheless, there does appear to be a relatively rapid decrease in salinity starting at the time of the sudden drop, consistent with energetic mixing of fresher water from outside the fast ice zone. Despite the fact that absolute temperature rises during this time (Fig. 1A), the elevation of temper­ature above freezing inferred from the mC instrument (Fig. 4B) becomes negative (supercooled) and remains that way for at least half an hour, perhaps for as much as hour. The supercooling is about an order of magnitude less than implied by the standard pumped instrument.

The main purpose of this document is to investigate possible reasons for episodes of marked decrease in conductivity observed during the short deployment in energetic flow under fast ice in Freeman Sound, Stor Fjord, Svalbard. Obviously, the sample is too small to draw defi­nite conclusions. However, the scenario that I think most likely is as follows. Overall the water column is close (within a few mK) to its freezing temperature. There exist substantial horizontal gradients in salinity, at least near the fast ice edge, that are advected back and forth during the tid­al cycle. Differential mixing of heat and salt can lead to transient conditions of supercooling, be­cause the fresher water (which is higher in absolute temperature) loses heat to the more saline water faster than it gains salt. This leads to supercooling of a few (in our case possibly about 2) mK, which is gradually relieved either by nucleation near the ice or further mixing.

This supercooling affects the standard SBE 04 instrument by lowering the measured con­ductivity by amounts that are unrealistically large in physical terms of the turbulent system we measured, but nevertheless still within a reasonable (i.e., believable) range of inferred salinity. Al­though not shown, a battery operated Aanderaa RDCP instrument deployed about 75 m away showed similar transient excursions in salinity, at somewhat different times. Since the SBE instru­ments are de facto standards for oceanographic work, understanding this effect is important for gauging the magnitude of supercooling observed in polar waters.

Are these transient supercooling events just an interesting scientific curiosity? or do they represent a significant impact on ice and brine production in Stor Fjord? It seems to be a question worth pursuing.

Figures

Figure 1. A. Current velocity along the major tidal axis in Freeman Sound. Positive values are from the fast ice edge into the sound. The ice edge was approximately 400 m from the measure­ment site. B. Temperatures measured 1 and 3 m below the ice water interface. C. Salinity from standard SBE T/C measurements. The pair at 1 m is pumped, the lower pair is not. Box indicates period discussed in Fig. 2.

Figure 2. A. Detail of temperature during the “conductivity events” between 02:00 and 04:00, 23 Mar 2007. B. As in A, except salinity. The red trace is salinity at the lower cluster, adjusted by a constant offset of 0.028 psu chosen so that the end values bracketing the excursions agree.

Figure3. A. Difference between measured temperature and freezing temperature (at surface pres­sure) for salinities shown in Fig. 2B. B. Detail showing the decrease in DT with time prior to the sudden drop in conductivity at about 02:20 UT.

Figure 4. A. Comparison of salinities derived from the different conductivity sensors mounted in TIC 1. B. Elevation of temperature above freezing for salinity calculated from the mC instrument as described in the text. TIC 1 temperature has been adjusted downward by 1 mK.