2.1 Snow sampling and sieve analysis

Several of the snow profiles discussed in this paper were located in open woodland of approximately 25% crown coverage. One profile was located on a small bog and one profile was sited on an open, wind-swept lake (Knob Lake). A Swedish snow-density sampler (Reference GranbergGranberg 1984, Reference Granberg and KingsburyGranberg and Kingsbury in press) was used to collect rectangular slabs, one litre in volume, at vertical intervals of approximately 8 cm throughout the snow profile.

The samples were brought back to the laboratory where sieve analysis was performed on the distintegrated snow skeleton at temperatures of −20°C or colder. The disintegration of the snow skeleton was accomplished by hand-kneading the sample bags until no lumps could be felt. This was done immediately before the sieving. Standard 8-inch sieves for soil analysis were used with an automated sieve shaker. The stack of sieves were 4, 2, 1, 0.5, 0.25 and 0.125 mm mesh respectively and the agitation time was five minutes on an automated sieve shaker. The sieve contents were weighed on a Mettler 4000 electronic balance. Microphotographs of different size fractions were taken at intervals along the vertical profile in the 1983 survey of the “time-profile” site which is a specially prepared site for interannual surveys of snow-cover stratigraphy.

2.2 Calculation of grain-surface area

To calculate the internal surface area per unit volume of snow, the partial densities of individual size fractions Dp were first calculated by

where W_p is the weight of the particular grain-size fraction in grams and V is the volume in cubic centimetres.

The number of grains per unit volume of each sieve fraction is given by

where Di is the specific weight of ice in grams per cubic centimetre, S (=1) is the effective grain size for the grains in the particular sieve fraction and Sf (=1) is a shape factor representing the ratio of the actual grain mass to the mass Vr of a sphere of the same effective grain diameter. The effective grain diameter is a function of the grain diameter at the smallest cross-section of the grain, the shape of that cross-section and the distribution of sizes of these cross-sections in a given grain aggregate. In the following calculations, the mean of the upper and lower mesh sizes for each size fraction has been used as the effective grain diameter. There are some obvious shortcomings in using this approach but they are thought not to distort the results too seriously in this application.

The surface area Apv attributable to the particular grain size fraction is calculated by

where Af (=1.5) is a factor accounting for the surface area of the grains as a decimal fraction of the area Ar of a sphere of the effective grain size.

The total surface area per unit volume Av is obtained by summation of the partial areas of the individual grain-size fractions. To convert this value to surface area per unit weight of snow Aw the following relationship is used

where D is the density in grams per cubic centimetre of that snow layer.

3.1 Repeatability of results

To assess the repeatability of the results, a few of the samples were re-mixed and analyzed several times in succession. The repeatability was in all cases excellent, far better than is normally obtainable in soil analysis. Figure 1 shows results of a triple sieving which illustrates the close repeatability of the sieve analysis results.

Fig.1

Result from repeat sieving of one sample.

The grain-size distributions from any given depth were always clearly distinguishable from the grain-size distributions obtained from any other depth within the particular snow profile. Figure 2 shows a typical set of grain-size distribution curves for one of the woodland sites.

Fig.2.

Grain-size distributions at time-profile site in 1983.

3.2 Grain-size variations

In the following graphs the variations in grain-size distributions with depth in the snow cover have been displayed in per cent of total weight on the left side of the graph. The isolines represent different sieves and the space is proportional to the weight percentage of the total sample that did not pass through the particular sieve while the space to the right indicates the amount that did pass through it. The graph on the right shows vertical lines representing the depth interval spanned by each sample (5 cm) and the horizontal position of each such line indicates the density of that particular sample.

In the 1983 survey the plots of grain-size distributions with depth showed a close correspondence between different open woodland sites although some of the sites were located several kilometres away from the others. Figures 3, 4 and 5 represent profiles obtained at different woodland sites in the Schefferville area. Although jags occur due to ice layers and occasional inhomogeneities in some of the profiles, the similarity between the different profiles is striking.

Fig.3

Open woodland site.

Fig.4

Open woodland site.

Fig.5

Time-profile site, 1983.

There was a similar pattern of variations in grain-size with depth for the small bog and lake sites (Figs.6 and 7). Although the actual grain sizes and total depths differed in these cases the trends of change in grain-size with depth were similar. In all cases a peculiar, as yet unexplained, minimum in the finer size fractions occurred a few decimetres above the snow/ground interface rather than at the interface as might, at least intuitively, be expected.

Fig.6.

Small bog near Barney.

Fig.7.

Site on Knob Lake.

The same general pattern was repeated in 1984 (Fig.8). The grain-size distributions show considerable similarity to those observed in 1983 and suggest a process which closely controls the relative proportions of different grain sizes at different depths.

Fig.8.

Time-profile site, 1984.

3.3 Variations in surface area

The surface area per unit volume or weight of snow is shown by Figures 9 to 14. In these figures the surface area per unit volume is plotted with a solid line while the surface area per unit weight is plotted with a dotted line. Four of the figures represent woodland sites (Figs.9 to 12) while Figures 13 and 14 represent a small sheltered bog and a lake, respectively.

Fig.9.

woodland site.

Fig.10.

woodland site.

Fig.11.

Time-profile site, 1983.

Fig.12.

Time-profile site, 1984.

Fig.13.

Site on small bog.

Fig.14.

Site on Knob Lake.

The surface areas per unit volume and per unit weight exhibit relatively similar patterns for the three sites in open woodland as shown by Figures 9, 10 and 11. The greatest variations occur in the near-surface layers where also the greatest uncertainties exist regarding the shape factor assumptions. In general, with the present assumptions, the surface area per cubic centimetre remains <20 cm² in the lowest 30 to 40 cm of the snow cover and is reasonably uniform throughout this layer. In fact, a slight minimum occurs 15 to 20 cm above the snow/ground interface. In the upper parts of the profiles, values are approaching 40 cm² according to the present calculations. The true values are likely to be higher; the present set of assumptions would produce a conservative estimate.

The surface area per gram of snow varies from about 60 cm² near the base to about three times that value near the surface. The value is constant or even declines somewhat with height in the lowest 30 cm of the snow cover.

There are no substantial differences between the areas per unit volume at the time-profile site in 1983 and 1984 respectively (Figs.11 and 12). The uppermost sample shows a very low value which can be ascribed to recent new snowfall in 1984. The stellar crystals from this snow fall were still intact when the analysis was performed which explains the fictitiously small surface area per unit volume. This layer obviously did not conform to the shape factor assumptions used in this study.

The bog profile (Fig.13) has, on account of the shallower snow, had a much steeper temperature gradient through it. There was an early winter melt in 1982–83 which incorporated the lowest part of the snow cover into a basal layer of solid ice. The snow cover was otherwise similar to the uppermost 50 to 60 cm of the woodland snow covers and was, similarly, unaffected by wind. The depth hoar was quite well developed in this snow cover as is shown by the small surface area per unit volume that prevails through most of the snow profile. The profile of surface area per gram is also different from those of the woodland sites. In its lower part the values are greater and in the upper part smaller than those for the forest sites.

As expected the wind-blown snow cover of Knob Lake shows a large surface area per unit volume (Fig.14). Its surface area is greater than for any of the other snow covers on the basis of both per unit volume and per unit weight.