Other formats

    TEI XML file   ePub eBook file  


    mail icontwitter iconBlogspot iconrss icon

Tuatara: Volume 15, Issue 3, December 1967

The Lakes of the McMurdo Dry Valleys

page 152

The Lakes of the McMurdo Dry Valleys

ALTHOUGH much of the Antarctic continent is covered with ice and snow there are some small areas which are ice free. One of the largest of these is the McMurdo Oasis area into which Victoria University of Wellington has sent many expeditions.

The first question that might be asked is why are these areas not ice covered like the rest of the continent. To answer this question it is necessary to consider the precipitation/evaporation balance. In this cold and arid region (mean annual temperatures of McMurdo Oasis region is approximately -20°C) only a very small fraction of the snow that falls ever melts and almost all of it is lost directly by sublimation. The best way of understanding the precipitation/evaporation balance is to consider it in terms of the single parameter—nett precipitation. This is defined as being equal to the total precipitation less the total evaporation. If this value is positive for any area in this region, the land surface will be covered with ice and snow. If this value is negative (i.e. sublimation is greater than precipitation) the area will be ice free (a so called ’dry area’), that is unless ice can flow into the area from a region of positive nett precipitation. The imaginary line which divides these two regions is called the ‘snow line’ and is the point at which precipitation equals sublimation1. For a given region the nett precipitation increases as altitude is increased. As one moves from the sea inland the snow line rises; presumably because total precipitation decreases. As one descends below the snow line the environment becomes drier and drier. Parts of this area are so dry that CaCl2. 6H2O crystallizes from ice free saline ponds (e.g. Lake Don Juan). This implies that the mean relative humidity of this area is 45% R.H. Thus one could map the relative aridity of this region by drawing lines parallel to the snow line corresponding to say 80%, 70%, 60%, 50%, 40% relative humidities. These lines not only control such physical phenomena as the presence or absence of permafrost, depth to permafrost, length of time the ice in an ice cored moraine will survive, but also have important ecological consequences in this region so inhospitable to living things. An example is the growth of lichen. It is well known that fungi can only grow at relative humidities above 80%, thus lichens can only inhabit a strip between the 80% relative humidity line and the snow line.

page 153

Thus the dry areas are those areas which lie below the snow line and into which ice, from above the snow line, cannot flow.

The McMurdo Oasis Dry Valley Area consists of a number of enclosed drainage basins. The lowest part of each is occupied by a saline lake. If we consider the evaporation/precipitation balance of the entire drainage system it can be seen that the nett excess precipitation of a snow field will flow below the snow line as a glacier. If the surface area of the glacier is insufficient to balance the sublimation/precipitation budget the glacier will advance further and further below the snow line toward a situation where the total positive nett precipitation above the snow line is balanced by the total negative nett precipitation below the snow line. Usually the glacier has pushed sufficiently far below the snow line so that some summer melting takes place. In such cases for a few days during the hottest part of the summer a stream flows away from the glacier snout and feeds a lake which occupies the lowest point of that particular enclosed drainage basin, or into that rather special saline lake called the sea. The size of the lake is determined by that area needed to balance the evaporation/precipitation equation for that particular drainage area. If there is a nett precipitation increase to the area the lake levels will rise and if there is a decrease in precipitation the lake levels will fall.

The susceptibility of the lake levels to change depends on what fraction of the total evaporation of the system is accounted for by the lake surface. If this is small a very small increase in nett precipitation to the area will lead to an increase in lake area. Thus some lakes are much more susceptible to fluctuation of surface area than others. What is usually measured is lake level. This is not only a function of lake area but also depends on the shape of the depression occupied by the lake. For a flat basin like that occupied by Lake Fryxell a large increase in area can be achieved with little increase in depth, whereas in a steep U-shaped trough such as occupied by Lake Bonney, the reverse is true.

Thus these Antarctic lakes are very sensitive indicators of changes in nett precipitation and hence of glacial advances and retreats.

The above treatment is an over simplification of the situation and deals with climatic changes only in terms of nett precipitation. Let us consider the effect of temperature changes. If the nett precipitation were to remain constant but the temperature were to decrease there would be less summer melting of the glacier so that it would advance and produce a larger area for evaporation below the snow line. This means that a smaller lake area than before would be needed to balance the system and the lake levels would drop. Conversely if the climate were to warm, more of the glacier would melt and the glacier would retreat leaving less area below the snow line so that a larger lake area would be needed to page 154 balance the system and the lakes would rise. In such systems it is always difficult to separate the temperature and nett precipitation effects, and the lake areas (levels) really only tell us the amount needed to balance the system. However the effect of temperature is limited for a number of reasons. The whole temperature range between glacial and interglacial periods is only about 6°C. Also to consider the effects of temperature alone we should hold the precipitation (rather than nett precipitation) constant. If then the temperature is raised the rate of evaporation would rise since the vapour pressure of ice is very temperature dependent. This would lead to a rise in the snow line which would tend to negate the effect of the ice retreat since it would increase the evaporation per unit area per unit time from the stream and lake surface, which would lead to less lake area being needed to balance the system.

A further effect is that the retreating glacier is replaced by a stream which can contribute almost as much evaporation as the glacier it replaces. Chemical analysis of the water of the Onyx River flowing into Lake Vanda shows that the 18 mile length of the Onyx River provides as much exaporation as the whole surface of Lake Vanda (4 miles by 1 mile).

It is therefore concluded that in this area lake level fluctuations are controlled largely by fluctuations in the nett precipation.

There is abundant evidence of past changes in lake levels. This evidence takes the form of lake shore lines higher than at present and the presence of chemical concentration gradients in the lakes.

These changes in lake levels must be a record of past climatic change. The question that immediately arises is when was this climate change and what was its cause. Let us consider some examples: Lake Vanda receded from its upper levels (shown very well in Fig. 1) some 3000 years before present. This was determined by the C-14 dating of algae found in these upper levels by Professor Wellman. While Lake Vanda stood at its upper level its surface area would have been approximately twice that at present. This suggests, as will be discussed below, that the nett precipitation to the snow fields which feed Lake Vanda (i.e. the eastern end of the Wright Valley) would have been 20% to 40% greater than at present.

In most areas of the world precipitation to a given area is controlled by its position relative to the open sea. A very attractive hypothesis is that the Ross Ice Shelf was much further to the south than its present position prior to 3000 years ago. This would mean that there would have been more open sea closer to the snow fields supplying Lake Vanda. The hypothesis that the local alpine glaciers, as distinct from those fed from the polar plateau, are controlled by mean distance to the sea (i.e. the position of the Ross Ice Shelf) is further supported by evidence that the coastal regions page 155 to the south of the McMurdo Oasis area appear to have been more heavily glaciated up until a few thousand years ago.

All the lakes that occupy the lowest part of the various enclosed drainage basins are chemically stratified and these chemical concentration gradients contain paleoclimatic information if we can understand the system sufficiently well to interpret them.

The paleoclimatic data in the lakes are particularly welcome in the Antarctic because the more usual methods of dating climatic-events, for example, carbon-14 dating, are not often applicable in this region. This is because carbonaceous remains are rarely found.
FIG. 1: Lake Vanda in the Wright Valley: note high lake levels. (U.S. Navy Photograph)

FIG. 1: Lake Vanda in the Wright Valley: note high lake levels.
(U.S. Navy Photograph)

page 156

There are two ways in which the chemistry of the lakes helps us to estimate the dates of past climatic and glacial events—firstly the total amount of salt in the lake and secondly salt gradients that may be present.

Before we can understand the chemistry of the lakes we must understand the chemistry of the whole drainage system. When snow falls on to the snow fields it contains small quantities of salts, the principle cations being sodium and calcium, and the principle anion being chloride. The chemical composition of atmospheric precipitation has been the subject of much study (see for example Ref. 2 and 3). It is well established that as far as the above cations and anions are concerned, the chemical composition of rainfall (and snow fall) depends on the position of the collecting position with respect to the sea. If two areas have the same chemical composition for their atmospheric precipitation the total salt deposited will be directly proportional to the precipitation.

The snow that falls on the snow fields that feed the glaciers of the McMurdo Region contains a fraction of a part per million of inorganic salts. It can be seen that if we measure say the sodium or chloride content of the snow in the névé of a glacier and also the sodium or chloride content of the stream leaving the snout of the glacier it is a simple calculation to determine how much water has been lost from the glacier by sublimation. Such measurements indicate that only a fraction of one per cent of the water that falls as snow in névé survives to flow as a stream from the snout of the glacier.

These salts then are carried down the stream and eventually end in the lake. Since chloride is a very minor constituent of the rocks of this area and since the glaciers do almost no cutting and carry very little moraine, it is concluded that the chloride in the lakes is the result of the contraction of large quantities of snow. It is interesting in this connection to mention the effect of permafrost on the streams of the McMurdo Dry Valleys. Because of the extremely low mean annual temperature (ca.-20°C) any water that percolates into the ground immediately freezes so that the streams all flow on ice bottoms and the percolation of ground water through sediments, which the in temperature regions can contribute large quantities of soluble salts, is quite impossible.

Thus it is reasonable to conclude that the total chloride in a lake represents the total chloride which has fallen in the drainage area feeding the lake since the lake first formed. Since the lakes in this area have a permanent ice cover many feet thick, chloride can only be removed from the lake in one of two ways. Either the lake will rise until it overflows into another drainage basin (or the sea), or a large glacier can advance through the valley and push the lake contents into the sea. Thus every lake can be considered page 157 to have a chloride age which is obtained by dividing the yearly mean chloride inflow rate into the total amount of chloride in the lake. This age is really a minimum age since it is possible to conceive periods in the past when the total precipitation was much lower than at present. For example, the edge of the Ross Ice Barrier might have been further to the north.

It is difficult to imagine a situation when the precipitation was as much as twice that at present. This is because such an increase would itself lower the snow line converting part of the glacier which now has a negative nett precipitation into a region of positive nett precipitation. The glaciers would of course advance and offset this effect but the real problem is that of rising lake levels that would result. Consider Lake Vanda for example. At present for every 100 parts of water that are deposited on the snow fields less than 20% is evaporated from the lake surface. This value is deduced from determining the chemical composition of inflow waters and snow on snow fields feeding the lakes and calculating the water lost during the movement from the snow field to the lake. If the nett precipitation for the area were to double we would be faced with disposing of two hundred parts of water and the lake area would have to increase its area at least ten fold. It would have to increase much more than ten fold in fact because its surface would be much closer to the snow line and evaporation would not be efficient. It follows from this line of reasoning that any increase in precipitation by say a factor of two or more would lead to the filling of the whole valley with lake and expanded glacier and there is no evidence for these very high lake levels.

Thus some picture of past glacial events emerges. Perhaps the most startling result is that some of these lakes, for example Lake Bonney and Lake Vanda, have chloride ages in excess of 60,000 years. It has generally been assumed by most workers in the Antarctic that these areas had been extensively glaciated during the last ice age (i.e. up until about 10,000 years ago). The salt ages for some of the lakes in this area makes this appear very unlikely.

Since the salt in Lake Vanda is largely calcium chloride whose freezing point is approximately -50°C it is almost certain that any through glacier would have pushed the salts in Lake Vanda into the sea.

Since this is a very important conclusion, let us examine carefully the assumptions made in this estimate. Firstly that all the chloride did in fact come from snow melt water. The total chloride content of Lake Vanda is in excess of 1,000,000 tons. One would need an extensive salt deposit to make any significant contribution to this quantity. The salt is calcium chloride which is so deliquescent that it would be unlikely to occur as a solid and to have survived previous glaciations.

page 158

The second assumption is that the nett precipitation to this region has not averaged six times that found at present. From the argument presented above it follows that a six fold increase in the precipitation to this area would increase the surface area of Lake Vanda by thirty fold and this would overflow into the sea and lead to a reduction in the ‘salt age’. Further as will be seen below the evidence is that the last 2000 years have been on the average drier than at present.

This idea that the Antarctic was not extensively glaciated during the last Ice Age led to the development of a new theory for the Origin of the Ice Ages (4).

Turning to the problem of the origin of chemical gradients in the saline lakes of the McMurdo Oasis — an extreme example is Lake Vanda.

A brief summary of the physical chemical structure of the lake is given in Fig. 2. The top is covered by 12 feet of ice. The region 11-55 feet and 125-160 feet has a salt gradient and must be weakly density stratified. The region 55-125 feet has a uniform temperature and uniform chemical composition, and is believed, from heat transfer considerations, to be a layer of strong convection. The lower region of the lake below 160 feet is strongly density-stratified saline water which is considered, from heat flow considerations, to be non-convective. Detailed chemical analysis of the water in this region showed that it was principally a solution of calcium chloride.

It is interesting to speculate on the origin of the salt concentration gradient in the lake. The only reasonable explanation seems to be that at some period in the past the climate was such that the Onyx River did not supply appreciable water to Lake Vanda. Under these conditions the lake-level would have dropped until only a few feet of concentrated calcium chloride remained. When the climate changed, the Onyx would flow during the summer and fresh water would have flowed on top of this strong salt solution. Since that time the calcium chloride has been diffusing upwards. If such a model is assumed, it is possible to calculate the time (5) in the past when this climatic change occurred.

If it is assumed that Lake Vanda has a flat bottom and vertical sides and that at zero time all the calcium chloride is concentrated in a layer of negligible thickness on the bottom, then the concentration profile at time t is given by:

C= M/(Dt)1/2 · e−h2/4Dt

where C is the concentration of calcium at distance h from bottom after an elapsed time t; h = distance above bottom; D = diffusion coefficient of calcium chloride = 0.68 cm2/day at 10° M = total page 159 mass of calcium chloride per unit area. The calculated profiles for t = 50, 1,000, 1,500 and 2,000 years are given in Fig. 3, together with the experimental data. No corrections have been made for bathymetry. The bottom of the lake is remarkably flat, and the depression which the lake occupies has very steep sides. Bathymetric corrections would raise the values in the upper part of the lake. It is quite clear that the climatic change occurred about 1,200 years ago. Accurate bathymetric data could increase the precision of this calculation.
FIG. 2: Density and temperature profiles for Lake Vanda.

FIG. 2: Density and temperature profiles for Lake Vanda.

page 160

In the calculation the initial depth before the inflow is taken as being negligible. If this had been taken as any significant depth the gradient near the bottom of the lake would have been flatter. The experimental results suggest that the lake was indeed of small depth at the time of change in climate and also that this change was relatively sharp and definite.

The heat balance of the lakes in this cold arid region also presents some interesting problems.

The lakes fall into two types:

Type I. Many of the lakes contain water on which floats a permanent ice cover ranging from 12 to 22 feet in thickness. Each summer water flows into the lakes and under the ice. This inflow water must be equal to the total evaporation from the lake surface or the level will change with time. We will call this class of lake ‘perenially ice covered lakes’ and although they are common in the McMurdo Oasis area they have apparently not been described in other parts of the world.

Type II. Ice Block Lakes. These ‘lakes’ are really solid blocks of ice with a flat top. They are frozen to the bottom. The yearly inflow water which makes up for evaporation losses flows on top of the ice and freezes in early winter.

The problems are, (1) why should some lakes be of one sort and some of another? (2) What controls the ice thickness on the lakes that are not frozen to the bottom? (3) How can liquid water survive in lakes in a region whose mean annual temperature is -20° C? To understand these problems let us first consider the perennially ice covered lakes (i.e. Type I above). Each winter ice freezes on to the bottom of the floating ice. The amount that freezes must exactly match the total ice lost by evaporation during the whole year and melting during late summer (i.e. total annual ablation). The amount that freezes is determined by the winter temperatures, the thermal conductivity of the ice and the ice thickness. It therefore follows that the ice will be thinner where the ablation is greatest and will increase in thickness as the nett ablation decreases approaching infinite thickness as the snow line is approached. The lakes in this region are rare in the world because few places are so cold and so arid. Although there are many perennially ice covered lakes in the McMurdo Oasis region this type of lake is extremely rare in the world as a whole because summer melting in other regions can ablate a greater thickness of ice than can be produced by winter freezing.

Thus to summarise, the lakes of the McMurdo Oasis Region are of interest from a limnological point of view because in this cold and arid region are found perennially ice covered lakes which are the extreme class of lakes. These lakes present the apparent paradox of containing liquid water in a region whose mean page 161 annual temperature is -20° C. The ice thickness of these lakes is controlled by the ablation rate of the ice cover which is a function of the distance of the lake below snow line. If the depth of the lake is less than the ice thickness the lake is frozen to the bottom and we have perennially frozen lakes containing no water. Examples of these are Lake Vida and Lake Vaska in the Victoria Valley which is considerably closer to the snow line than the Wright or Taylor Valleys are.

Returning now to the problems presented by the heat balance of these lakes:

How does liquid water survive in a lake in a region where the mean annual temperature is -20°C? Firstly, it is heated by the inflow water which may be one or two degrees above freezing and which sinks to a depth in the lake appropriate to its density. The maximum density of water is at a temperature of 4°C.

Secondly, most of the heat lost by conduction during the winter is provided by the latent heat of fusion of the water that freezes into the bottom of the ice to replace that lost by ablation.

Some of the lakes have temperatures considerably higher than 4°C and an alternative source of heating must be found. In Lake Vanda, for example, the bottom waters are 26°C. It was in attempting to find the source of this heating that V.U.W. scientists first became interested in Antarctic lakes. The author together with H. W. Wellman studied the physics of Lake Vanda on V.U.W.A.E.5 and showed that it was solar heated6. Because of the arid climate in this region the lakes are snow free during the summer. The ice
FIG. 3: Time since Lake Vanda rose calculated from chemical diffusion.

FIG. 3: Time since Lake Vanda rose calculated from chemical diffusion.

page 162 which grows on to the bottom of the floating ice cover every winter consists of very clear ice (6% solar energy transmission through 12 feet of ice). The high light transmission of this ice is due to the fact that for hundreds of years the ice has been growing on the bottom of the ice and ablating from the top. The ice now contains large crystals several square cms. in cross section and 12 feet long with the X-axis arranged vertically. These crystals act as light pipes. The light energy from the sun passes through the ice and into the water where it is absorbed. Because of the very strong density gradient (produced by dissolved salts — see Fig. 2) in the bottom of Lake Vanda. convection is completely suppressed so that heat can only escape by conduction. Since water is a very poor conductor of heat, the solar energy which is being absorbed from the sun every summer heats the bottom water of Lake Vanda to 26°C.

Let us consider the quantitive aspects of solar radiation and, assuming that the radiation through the water is attenuated exponentially, we have


where Q is the amount of energy/unit area/unit time, being radiated down past a horizontal plane at some distance x below some arbitary zero depth: Q0 is the energy reaching the depth x = 0; and a = 0.693/x1 2. where x1/2 is the distance in which the radiation Q is converted into heat being absorbed either in the lake water below depth x or on reaching the bottom. Assuming no convection (because of strong density stratification). the amount of radiation energy passing downward past a depth x, plus the amount of heat conducted through the lake water, must equal the amount of heat conducted out the bottom of the lake (C).

Q0eax − k dT/dx = C


dT/dx = Q0/k e ax −Dx+F

which integrates to:

T = −Q0e-ax/ka − Dx + F…(1)

where T is the temperature, K the thermal conductivity of water, and D and F are constants, a is obtained by bolometry. Taking Lake Fryxell as an example (7) if we fit the appropriate plot of equation (1) to give dT/dx = 0 at the right depth and to give the right temperature gradient and temperature at x = 0, a calculated curve such as that in Fig. 4 is obtained.

page 163
Since the initial work on Lake Vanda6 the water of all the lakes in the McMurdo Oasis area which contain water have been shown to be solar heated to some degree7,8,9 — even Lake Joyce which has over 22 feet of ice cover. Lake Vanda is by far the most spectacular naturally solar heated lake in the world. However in recent years the Israelis10 have constructed artificial solar
FIG. 4: Temperature and salinity profile for Lake Fryxell.

FIG. 4: Temperature and salinity profile for Lake Fryxell.

page 164 heated lakes 2 meters deep in which they have obtained temperatures of 90° C. They hope to use this method to trap solar energy for industrial purposes. This is a development which would be of considerable use in New Zealand. Such solar energy traps could be used in the manufacture of salt from sea water which is always difficult to do in a humid climate such as our own unless a cheap way can be found for heating the brine to increase its water vapour pressure.

What of the future? Lake Vanda and the other lakes of this region provide a very large experimental system for studying many problems, for example, heat transfer, diffusion isotope separation. They provide one of the few naturally occurring non-convective systems in the world. Almost certainly the lakes of this region will provide topics of scientific study for some considerable time to come.


1. Charlesworth, J. K., The Quaternary Era (London, 1957).

2. Junge, C. E., Air Chemistry and Radioactivity, International Geophysics Series, vol. 4, Academic Press, New York, 1963.

3. Wilson, A. T., Nature, 184 99-101, 1959b. Wilson, A. T., Nature, 186, 705-706, 1960.

4. Wilson, A. T., Nature, 201, 147, 1964.

5. Wilson, A. T., Nature, 201, 178, 1964.

6. Wilson, A. T., and Wellman, H. W., Nature, 196, 1171, 1962.

7. Hoare, R. A., Popplewell, K. B., House, D. A., Henderson, W. M., Prebble, W. M., and A. T. Wilson, J. Geop. Res. 70, 1555, 1965.

8. Hoare, R. A., Popplewell, K. B., House, D. A., Henderson, W. M., Prebble, W. M., and A. T. Wilson, Nature, 202, 886, 1964.

9. Shirtcliffe, T. G. L., and Benseman, R. F., J. Geop. Res., 69, 3355, 1964.

10. Tarbour, H., Large area solar collectors solar ponds) for power production, in U.N. Conf. New Sources of Energy, E/Conf. 35/S/47, 1961.