Introduction

In Antarctica, brine-soaked firn has been observed in the Lazarev Ice Shelf (Reference DubrovinDubrovin 1962), in a small ice shelf near Wilkes station (Reference Stuart and BullStuart and Bull 1963), in the Wordie, Brunt, and Larsen ice shelves (Reference Smith and EvansSmith and Evans 1972), and in the Brunt Ice Shelf (Reference ThomasThomas 1973). However, the phenomenon is best known in the McMurdo Ice Shelf, as reported, for example, by Reference Stuart and BullStuart and Bull (1963), Reference Risk and HochsteinRisk and Hochstein (1967), Reference HeineHeine (1968), Reference CloughClough (1973), and Reference Kovacs and GowKovacs and Gow (1975, 1977). Several mechanisms have been suggested for brine soaking, e.g. vertical percolation of sea-water from the bottom, lateral infiltration from the ice front, and upward diffusion along ice-crystal boundaries. This paper presents new findings on brine migration in the McMurdo Ice Shelf.

Results and Discussion

The depth characteristics, lateral continuity, and inland boundary of sea-water infiltration into the McMurdo Ice Shelf were determined in January 1977 with a dual-antenna radio echo profile system. The methodology for use of the dual-antenna sounding system is described in Reference Kovacs and MoreyKovacs and Morey (1979).

The inland boundary of brine infiltration was profiled in detail in January 1977, between Ross and White islands (Fig.l). Stations along the boundary were fixed by triangulation survey. The northern limit of this brine layer was originally delineated in 1967 (Reference Clough and BentleyClough and Bentley 1967, Reference CloughClough 1973). In November 1973, new bamboo markers were placed at intervals along the boundary (J Clough personal communication). Resurvey of the existing markers, 2S, IS, IN, 2N, 3N, and 4N (Fig.l), on 18 January 1977 indicated that the brine layer had penetrated the ice shelf 197, 129, 73, 55, 10, and 9 m, respectively, or at an apparent rate of 0.168, 0.110, 0.062, 0.047, 0.008, and 0.008 m d⁻¹.

The brine-layer depth and the electromagnetic wave velocity velocities and dielectric constants determined from the radio echo profile data versus distance along the profile line shown in Figure 1 are listed in Table I. Also listed are surface elevations determined by topographic survey, and ice-shelf depths determined by direct drill-hole measurement. These data allowed construction of a representative cross-section of the McMurdo Ice Shelf along our profile line (Fig.2). The location on 10 January 1977 of a large (4.4 m) step in the brine layer is shown, as is the inland termination of the brine infiltration, which is characterized by a series of descending steps. These features, as displayed on the graphic recorder, are shown in Figures 3 and 4.

The position of the 4.4 m-high step in the brine layer was reprofiled on 24 November 1978 and again on 20 January 1981. Between 10 January 1977 and 24 November 1978 this step or brine front had moved from about 2 905 m to about 3 610 m in from the ice front, an average advance rate of 1.037 m d⁻¹.

As with the 4.4 m-high brine wave currently overriding an older brine-soaked firn horizon, the brine steps at the inland boundary represent past intrusions of brine triggered by periodic break-outs of the ice shelf. Through this process, firn at the ice front between sea-level and the top of the old brine horizon becomes exposed to the sea, and a new wave of sea-water begins to permeate the ice shelf. From the slope of the brine-soaked firn layer it was estimated that approximately 3 km of the ice shelf had to have broken off to allow for the 4.4 m-high brine step. Reference PaigePaige (1971) reports that “the most extensive observed periods of ice-shelf break-out occurred in February of 1964 and 1965”. Approximately 1 500 m of ice shelf calved. Further calving occurred in 1966, 1967 and 1968, and apparently in 1970. No brine-wave record of these events was observed in the radio echo profiles because it was apparently removed by subsequent break-outs. If the last ice-shelf breakout occurred in February 1970 and the brine wave was 2 850 m in from the ice front in January 1977, then the average infiltration velocity of the brine wave would have been 1.21 m d⁻¹.

Between 6 February and 22 March 1980, large sections of the McMurdo Ice Shelf broke off (C Roper personal communication). During one period, the ice shelf was observed moving vertically 30 to 50 mm at an inland crack. A resurvey of the brine-profile line in January 1981 revealed that 1 640 m of the ice shelf was missing. Based upon old records (Reference HeineHeine 1963, Reference PaigePaige 1971) this break-out is one of the largest recorded. A new brine-layer horizon about 1.4 m above above the old horizon at the ice front extended inland 375 m (Fig.2). This infiltration took 302 d, an average rate of 1.24 m d⁻¹.

Reference Risk and HochsteinRisk and Hochstein (1967) attributed the brine-soaked firn in the McMurdo Ice Shelf to lateral percolation from the ice front. They also agree with Reference Stuart and BullStuart and Bull (1963) that vertical brine infiltration occurs through low-density bottom firn. However, Risk and Hochstein limited the area where the latter mechanism occurs to a zone extending about 0.5 km in from the edge, where permeable firn was considered to be exposed due to bottom melting. If vertical infiltration is occurring near the ice front, the top of the brine-soaked firn in this area should be near hydrostatic equilibrium with the sea-water. Our radio echo profile surveys showed that the brine layer, beginning at the very edge of the ice shelf, has a negative gradient. It therefore appears that permeable firn does not exist on the bottom of the ice shelf. The freezing of fresh water from the brine in the infiltrated firn has effectively blocked off interconnected pores, preventing upward migration of sea-water.

Reference ThomasThomas (1975) estimated a brine infiltration rate of 1.10 m d⁻¹ for the Brunt Ice Shelf, where the firn density (0.570 Mg m⁻³), temperature (-10°C) ami pressure gradient (1 in 525) were more favorable to sea-water infiltration. We can calculate a brjne-flow velocity (u) at station 10 following the procedure of Thomas, where

and B_o is the specific permeability of the firn, η is the dynamic viscosity of the.brine fluid and ∂p/∂x is the pressure gradient of the brine layer. Thomas has suggested that B_o = 9 × 10⁻⁴ d_m ² where d_m is the grain diameter, which we estimate to be 2 mm for 0.7 Mg m⁻³ firn. The firn temperature was ∼-14°C and n for the brine at this temperature is 4.5 × 10⁻³ Pa s. The hydraulic head, over a distance of 2 580 m from station 1 to station 9, is 3.1 m. From Equation 1 we calculate a flow velocity of 9.68 × 10⁻⁶ m s⁻¹ or 0.84 m d⁻¹. This value fits nicely between the average brine-wave velocity of 1.04 m d⁻¹ which existed before the wave reached station 10 (10 January 1977 to 24 November 1978) and the average velocity of 0.67 m d⁻¹ which occurred after the wave passed station 10 (24 November 1978 to 20 January 1981).

Core data obtained in November 1978 about 120 m inside and 60 m beyond the maximum inland position of the brine-layer intrusion, which occurred in January 1977, revealed that “impermeable” firn of density 0.82 Mg m⁻³ (ice by definition) was encountered at a depth of about 43 m, or 7 m above the brine layer. The inland limit of brine-movement infiltration in permeable material of the McMurdo Ice Shelf is therefore controlled mainly by the depth at which brine encounters the firn/ice transition. At this boundary, firn containing liquid brine is presumably carried downward and densified by continued snow accumulation on the ice-shelf surface.

The slope of the brine layer was 1 in 832 from the ice front to station 9, 1 in 910 from station 11 to station 20, and 1 in 1 500 from station 22 to 24. The surface of the ice shelf beyond station 11 has a slope of 1 in 606, while the bottom is 1 in 137.

If the brine layer along the profile line could maintain a slope of 1 in 830 beyond station 9 and could move instantly inland to intersect the “impermeable” firn/ice transition horizon, it would do so at a maximum about 14 km inland from the 1977 ice front. However, this cannot be expected to occur under existing infiltration conditions; as the brine moves inland, it will encounter denser and therefore less permeable firn as well as lower temperatures. These effects will slow the infiltration rate and probably limit inland penetration in the area of our profile to less than 11 km.

A rate of bottom melting of the order of 0.9 to 1.2 m of ice a⁻¹ has been estimated for stations 202 and 207 (see Fig.l) by Reference Risk and HochsteinRisk and Hochstein (1967). By assuming an ice bottom ablation rate of 0.9 m a⁻¹, it would take 69 a before the brine-infiltrated firn at station 24 became exposed to the sea. Similarly, if bottom melting was occurring at 1.2 m a⁻¹, it would take only 52 a before the brine-infiltrated firn reached the bottom of the ice shelf.

The typical ice-shelf movement rates shown in Figure 1 come from Reference HeineHeine (1967), except that of station 307 which came from the 1971-73 to 1978-79 summer-survey data on file at the New Zealand Scott Base, Ross Island, Antarctica. In 1978-79, station 307 was moving at 84 m a⁻¹. The survey data also indicate that this station is slowing. It is assumed that the ice shelf in t he area of station 307 will move westward at an average rate of 80 m a⁻¹ during the next 50 to 70 a and that the ice-shelf cross-section depicted in Figure 2 will remain unchanged. In so doing, it is possible to approximate the area of the ice shelf composed of brine-soaked firn and saline ice. Thus, at a forward movement rate of 80 m a⁻¹ and a bottom ablation rate of 0.9 m a⁻¹, we can expect that, when the saline ice near station 307 reaches the ice-shelf bottom, it will be some 5 520 m ahead of its present position and will have moved downward along path C (Fig. 2). Similarly, if the bottom ablation rate was 1.2 m a⁻¹, the brine ice would move forward some 4 160 m along path D. Path D intersects the firn/ice transition horizon about 10 700 m inland on the cross-section. If path C was similarly extended it would intersect the firn/ice transition about 15 km inland. From earlier considerations this distance appears excessive and suggests that the ice-shelf bottom is melting at a rate nearer to 1.2 m a⁻¹ than 0.9 m a⁻¹ in the area of the cross-section. This is in keeping with measurements by Reference PaigePaige (1969), who determined, from ice-movement records and drill-hole measurements, an ice-shelf thinning rate of 1.06 m a⁻¹ and a snow accumulation rate of 0.27 m a⁻¹ or a total bottom melting rate of 1.33 m a⁻¹ for an area about 0.5 km south of the southern end of our profile line. This ablation rate was confirmed by Reference HoffmanHoffman (1974).

Fig.1. Map of the McMurdo Ice Shelf area. The ice/firn transition controls the western limit of brine infiltration. The ice-shelf surface west of this boundary is ice.

Fig.2. Cross-section of the McMurdo Ice Shelf along the radio echo profile line. Reference Stuart and BullStuart and Bull (1963) first showed the ice shelf to be wedge-shaped. The height of the ice front varies in time and space and is a function of ice-shelf movement rates, break-out dates and magnitude, and the rate of bottom melting.

The above analysis does not consider vertical brine migration, which will occur first by diffusion of brine along crystal boundaries and later by inter-granular vein flow as the saline ice moves downward into an increasingly warmer environment. A tacit assumption 1s also Implied that the current brine terminus, being 1n “Impermeable ice”, is no longer moving Inland. This 1s not the case. Between 18 January 1977 and 24 November 1978, the deepest brine layer moved 27.5 m closer to station 307, a rate of 0.041 m d⁻¹. From 24 November 1978 to 20 January 1981, the brine layer moved from 12.5 m before to 18 m beyond station 307, an average of 0.039 m d⁻¹. Core data from this site revealed that the brine was migrating into ice with density of 0.85 Mg m⁻³ and temperature of about -16°C. The calculated brine salinity and viscosity at this temperature are 230°/_°° and 5.3 × 10⁻³ Pa s.

Fig.3. Graphic record of radio echo data showing the brine step of 4.4 m.

Fig.4. Fig.4. Graphic record of radio echo data showing several brine layers at the inland boundary of brine penetration. The fact that the ice-shelf bottom could be observed under the lowest brine layer indicates that the layer is “thin” and/or low loss with respect to electromagnetic energy transmission as the radio echo would not have been observed otherwise. Time between horizontal scan lines is about 325 ns.

The brine velocity is at least three orders of magnitude higher than simple diffusion theory through the ice lattice would predict (Reference Hoekstra, Osterkamp and WeeksHoekstra and others 1965, Reference Seidensticker, Hoekstra, Osterkamp and WeeksSeidensticker 1966). Also, calculation of fluid migration based upon Darcy's law of flow through a permeable material indicates a need for a material having a permeability significantly higher than that in the 0.85 Mg m⁻³ density ice through which the brine is migrating. “Ice with a density of 0.85 Mg m⁻³ could be permeable to brine movement because the veins at triple grain junctions would allow liquid movement under a potential gradient. However, at a temperature of -16°C, the veins would have to be enlarged by the moving brine in order for the permeability to reach that of temperate glacier ice (Reference Nye and FrankNye and Frank 1973). Even then, the permeability due to the veins alone would be too low to account for the observed rate of movement. Accordingly, some other mechanism must account for the brine's mobility” (S Colbeck personal communication).

The calculated depth of the brine layer, as determined from the radio echo data, was verified by drill-hole measurements made in December 1978. For example, from the 1977 radio echo data, we calculated the depth to the brine layer at station E, 9 560 m in from the ice front, to be 49 m. Coring in December 1978 revealed a depth of 50.4 m. During the preceding two winters, about 0.9 m of new snow accumulated. Adding this to the depth of 49 m gives a difference between the calculated and measured depths of 0.5 m or 1%. In 1974 to 1977, cores were obtained from several drill holes to a maximum depth of about 22.5 m and from four holes extending to the bottom of the ice shelf from which ice-shelf thicknesses were determined. In 1978, six holes were cored and one pit was dug along the survey line. Five of the core holes, stations B, C, D, E, and F were located 770, 3 590, 7 370, 9 560, and 9 700 m, respectively, in from the ice front. The first four holes met the brine layer at depths of 8.85, 9.4, 33.9, and 50.4 m, respectively. The pit, station A, located »5 m from the ice-shelf front, was excavated 2.5 m to sea-level and a hole was cored 6 m to the ice-shelf bottom. The core consisted of brine-Infiltrated firn which was now saline ice. The last core had a sea-ice Skeleton growth layer of 10 to 20 mm thickness on the bottom, indicating minor sea-water freezing due to winter cooling. The deepest hole at station F, located 37 m beyond station 307, reached a depth of nearly 60 m. The depth versus temperature profiles for three of the stations are shown in Figure 5. Extrapolating the temperature profiles in Figure 5 to a temperature of -1.8°C (the freezing point of the sea-water), we find an apparent ice-shelf thickness of 25.5 m at station B, 49.5 m at station 10b, and 80 m at station D. The thickness values for stations B and 10b agree well with the thickness that can be obtained from Figure 2 The value for station D is about 9 m less than the indicated depth in Figure 2 This suggests that the temperature gradient becomes steeper with depth at this location.

Fig.5. Depth-temperature profiles for stations 10b, and D.

Least squares fit curves passing through the depth versus density core data obtained at stations B, C, D, E, and F are shown in Figure 6. Each curve represents the firn depth-density trend down to the top of the brine layer except at station F, where no brine was encountered. This figure, along with the temperature profiles in Figure 5, reveals that, as one moves inland on the ice shelf, the thermal and density regime of the firn changes. In short, at a depth of 10 m at station B, the snow is much warmer and denser than at stations further inland. This is due to factors related to heat transfer. Where the ice shelf is thin, more heat escapes to the atmosphere from the sea-water and larger seasonal temperature variations occur in the firn. These effects accelerate the densification process.

Fig.6. Depth-density curves for the firn at five stations along the ice-shelf profile line.

From the data listed in Table I we can calculate the effective bulk density of the ice shelf (ρ_a) at station 24 from

where F is freeboard, ρ_w is density of sea-water (∼1.027 5 Mg m⁻³), and D_T is total ice-shelf thickness. Given that the surface elevation is 25 m and the keel depth K is 88 m then D_T = 113 m and, from Equation 2, ρ_a is 0.80.

From the values in Table I, the K to F ratio is 3.53:1 and the ice-shelf thickness to freeboard ratio is 4.5:1. The latter value is higher than the 4.15:1 ratio that would be derived from Reference GowGow's (1963) relationship based on density data from Little America V. The higher value for station 24 is believed to be due to brine loading and higher firn densities deflecting the ice shelf downward. A deflection due to brine loading was evident in our 1977 elevation survey, which showed a marked change in slope near station 10 (see insert A, Fig.2). This change occurred over such a short distance that it was first detected by eye and later verified by the elevation survey.

Another interesting finding from the January 1981 resurvey was that brine had risen 8.7 m inside the bore hole of diameter 0.15 m at station E and 0.5 m in the bore hole at station D. Lead-line sounding indicated a hard bottom at station E. At station D, the hole bottom had 0.7 m of “hoar-frost”, below which was a firn surface slightly moist with brine. At station E, the radio echo data revealed at least two horizons where the brine had migrated laterally away from the hole. At the up-well level of 8.7 m, the brine had moved 2 to 3 m laterally, while at the mid-up-well level, it had moved outward approximately 5m. It is unknown whether the up-welling of 8.7 m is the maximum hydraulic head in the brine layer at station E as up-welling may have been arrested by liquid freezing under the lower temperatures at the higher elevation.

Chemical analysis of brine-infiltrated cores from drill holes along the profile shows that freeze-fractionation of the sea-water, as it penetrates the ice shelf, preferentially precipitates virtually all sodium sulfate before the inland boundary of brine penetration is reached. For example, near the nose of the brine wave of height 4.4 m, the sulfate to sodium ratio is essentially the same as for sea-water, or slightly higher where sodium sulfate has accumulated locally. However, this ratio decreases four-fold immediately in front of the brine wave in samples from an earlier infiltrated zone across which the new brine wave is migrating. Near the terminal boundary of brine infiltration, the sulfate to sodium ratio is reduced to less than one-tenth the sea-water value, and concomitant removal of water by freezing within the pore spaces of the firn grains has produced brines approximately seven times more concentrated than the original sea-water.

Acknowledgements

The assistance of Thomas Fenwick and Rexford M Morey during various field surveys is acknowledged. John D Palmer, of the New Zealand Department of Lands and Survey, took our triangulation station data and arranged for its computer analysis and plotting on the McMurdo area map grid. The personal consultation provided by Dr Samuel Colbeck on aspects of water migration in ice is most appreciated.

Funding for this study was provided by the US National Science Foundation, Division of Polar Programs.