White ice: surface scattering layer

Melt ponds are the most eye-catching feature of the summer ice surface, and their formation and evolution strongly impacts the ice-albedo feedback (Reference Grenfell and MaykutGrenfell and Maykut, 1977; Reference Ebert, Schramm, Walter and CurryEbert and others, 1995; Reference Fetterer and UntersteinerFetterer and Untersteiner, 1998; Perovich and others, in press b). A more subtle but equally important observation, which also impacts the ice-albedo feedback process, is that the ice stays white. During summer much of the ice surface is covered by a thin (1−3 cm) granular layer. Figure 2a shows a close-up photograph of this surface layer. The footprint on the lefthand side of the photograph shows that this surface layer is soft and is a few centimeters thick. At first glance this ubiquitous layer appears to be a thin layer of melting, well-metamorphosed rounded snow grains, each a few millimeters across. However, we know that this is not the case, as virtually all of the snow cover had melted by late June. Mass-balance measurements also showed that, on average, there was 50−100 cm of surface ablation at unponded sites during the summer. The surface granular layer is deteriorated ice. Our observations indicate that as surface ablation progresses, the ice fragments, developing a surface granular layer that continually renews itself during melt.

Fig. 2. (a) Cclose-up photograph of the surface scattering layer of white ice. the footprint on the left hand side shows the depth of this layer. the ring in the photograph is used to measure bidirectnal reflectance, (b) photograph of an ice core removed from multi-year ice. there is a 7 cm thick drained layer at the top of the ice core that evolves into the surface scattering layer.

The deterioration of the surface layer is a direct result of the structural and optical properties of sea ice. Structurally, sea ice is a matrix of individual ice crystals containing an amalgam of ice platelets, brine pockets and air bubbles (Reference Weeks and AckleyWeeks and Ackley, 1982). Sea ice is also a translucent medium, where solar energy penetrates, resulting in internal melting preferentially along the grain boundaries. The surface layer of melting multi-year ice is porous and above freeboard, so meltwater can readily drain. A 10 cm diameter ice core taken from the top of the ice is displayed in Figure 2b. The top 7 cm of this core were drained and appeared white. All of the ice cores we examined had a drained surface layer ranging from 5 to 30 cm in thickness. Densities in this drained layer were 0.6−0.8 g cm"³, compared to values of about 0.9 g cm"³ in the underlying ice. This drained layer is the breeding ground for the surface granular layer. Once the meltwater drains, the surface layer becomes friable and is easily broken apart into small fragments. The process of drainage and fragmentation results in a surface granular layer that continually renews itself as surface melting progresses. The surface layer has a white appearance because of its low density and high air volume. These air-filled pores scatter the incoming light, creating the white appearance of the surface.

To confirm the hypothesis that the white ice was the result of a perpetually draining surface layer, we repeated an experiment performed by Reference UntersteinerUntersteiner (1961). We scraped away the surface granular layer down to bare ice. Indeed, by the next day a surface granular layer had re-formed on this site. Such layers have been reported in the past (Reference UntersteinerUntersteiner, 1961; Reference Grenfell and MaykutGrenfell and Maykut, 1977; Reference ZubovZubov, 1979; Reference WilliamsWilliams, 1994; Reference Eicken, Lensu, Leppäranta, Tucker, Gow and SalmelaEicken and others, 1995; Reference Tucker, Gow, Meese, Bosworth and ReimnitzTucker and others, 1999). At SHEBA we determined that these layers continually renewed themselves during the melt season.

While the layer of coarse, loose ice grains was ubiquitous and pervasive, it did vary in thickness. Observations made at mass-balance sites (Reference PerovichPerovich and others, 1999a) showed that the layer typically was 1−3 cm thick. Qualitative observations indicated that this layer tended to become thicker on sunny days and thinner on cloudy days. This was a result of two factors: penetrating solar energy and surface ablation. Penetrating solar energy thickens the layer, and surface ablation thins the layer. Mass-balance studies (Perovich and others, in press a) indicated that there tended to be more surface ablation on cloudy days than sunny days, thinning the granular layer. In contrast, the incident, and penetrating, solar energy is greater on clear days, causing more internal melting and greater development of the granular layer.

Because of the self-renewing surface scattering layer the white-ice albedo typically remained in the 0.65−0.70 range throughout the summer. Reference Grenfell and MaykutGrenfell and Maykut (1977) reported an albedo of 0.5 for bare ice without a surface scattering layer. The effect of the surface scattering layer on areally averaged albedo is significant. For instance, if there were no surface scattering layer on 7 August the white-ice albedo would have been 0.5 instead of 0.67, giving an areally averaged albedo of only 0.35, half the 15 June value.

Break-Up

During summer, lateral melting around the perimeters of ice floes results in an increase in the amount of open water. Increases in open water can also result from ice dynamics. At Ice Station SHEBA, there was a sharp increase in open water in early August due to ice dynamics; wind stresses on the ice cover caused divergence of the ice floes (Reference Richter-Menge, Perovich and PegauRichter-Menge and others, 2001). Because air temperatures are consistently near 0°C and the incoming solar radiation is at a maximum during the summer, any open water produced by either melting or dynamics will not freeze over. This explains why the amount of open water increases, but it does not account for the break-up of the ice floes. Observations and models have determined that in the central Arctic for most of the summer, internal ice stresses are near zero, and the movement of the ice cover becomes more strongly correlated with wind and ocean forcing (Reference McPheeMcPhee, 1980; Reference Steele, Zhang, Rothrock and SternSteele and others, 1997). This is consistent with measurements of the ice drift and internal ice stresses taken at SHEBA (Reference Richter-Menge, Overland and McNuttRichter-Menge and others, 1998, Reference Richter-Menge, Perovich and Pegau2001; Reference PerovichPerovich and others, 1999a). Under these conditions, the ice is in a state of free drift, and forces that can break up the ice cover due to the dynamic interaction of ice floes are minimal. Hence, we believe that the summer floe break-up resulted from thermodynamic weakening of the ice.

There was dynamic activity in the ice pack throughout the winter. In the immediate vicinity of Ice Station SHEBA there were major stress-deformation events in the period January-March, resulting in the formation of cracks, leads and ridges (Reference Richter-Menge, Overland and McNuttRichter-Menge and others, 1998). The cracks and leads rapidly froze and were covered by drifting snow, effectively disappearing. When the snow melted in early June, a winter’s worth of cracks suddenly reappeared and acted as the starting-points for break-up. The photographs in Figure 3 demonstrate this. In January, multi-year ice adjacent to our "Quebec" site (Reference PerovichPerovich and others, 1999a) broke apart forming a 0.5 m wide crack. By April this crack was solidly frozen and snow-covered (Fig. 3a). As the melt season, which began in late May, progressed, the crack preferentially melted and was the first place to melt through to the ocean (Fig. 3b). There are several reasons for the preferential melting of cracks. First, the ice in cracks is thinner than the adjacent multi-year ice. The cracks also melt faster because they are first-year ice and, once the snow has melted, have a lower albedo than the more highly scattering multi-year ice (Reference Grenfell and MaykutGrenfell and Maykut, 1977). The thinner ice in the cracks also had a lower freeboard, so, unlike the multi-year ice with the self-renewing white surface layer, it could not develop much of a surface scattering layer due to drainage. Finally, being topographically lower, the cracks tended to collect meltwater, further reducing their albedo.

Fig. 3. Photographs of a crack in the ice that formed in january 1998. the photographs were taken in (a) april 1998 and (b) july 1998 there was preferential melting in the crack as the melt season progressed.

By the end of July, thermal deterioration of the cracks was well advanced and the floes were primed to fall apart. Walking surveys on the ice showed that ice in the cracks had almost completely melted and in many places the floes were barely held together. In addition, some melt ponds had melted through to the ocean, further weakening the ice. Winds were light during most of July and there was only modest ice drift. The wind increased at the end of July and the ice drift increased as the pack moved in concert. At the beginning of August the wind decreased, and the floes began to slow down, diverge and break apart (Reference Richter-Menge, Perovich and PegauRichter-Menge and others, 2001). This was the major divergence event of the summer, resulting in an increase in the amount of open water from 5% to 20% in the area around Ice Station SHEBA (Perovich and others, in press a). Figure 4 shows before and after photos illustrating the impact of this divergence event at the Quebec site. The area around Quebec fragmented into several pieces along the flaw lines of the melting cracks. We observed similar behavior at the other mass-balance sites. Three of our sites (Quebec, Seattle and the Mainline) were located within a few hundred meters of each other. After the divergence event these sites were on individual floes and were separated by as much as a few kilometers. The aerial photographs in Figure 5 show this break-up on a larger scale. The photographs were taken on 15 June and 7 August 1998 from an altitude of approximately 1800 m. The differences are obvious: on 7 August there is much more open water and the large plates of ice from 15 June have fragmented into many more smaller floes.

Fig. 4. Helicopter photographs taken of the site pictured in figure 3 (a) before and (b) after the divergence event in early august.

Fig. 5. Aerial photographs taken at an altitude of 1800 m on (a) 15 june 1998 and (b) 7 august 1998

To quantify these changes in the characteristics of the ice cover between 15 June and 7 August, we analyzed aerial photographs from these days taken along a 50 km × 50 km box surrounding the ice station (Perovich and others, in press a). Photographs covering a total area of 180 km² on 15 June (164 photographs) and 153 km² on 7 August (139 photographs) were processed to determine the area fractions of ice, ponds and leads (Fig. 6). For each day we selected a subset of images and determined the total floe perimeter in those images. Since each photograph was only 1.1km², and many floes touched the border of the image, it was not possible to accurately determine the number of floes and the floe size distribution. However, it was straightforward to determine the total floe perimeter in the images. As floes break up, the total perimeter of the ensemble of floes increases. Thus the perimeter provides an indirect proxy of the floe statistics and is also a key parameter for determining the amount of lateral melting (Reference Maykut and PerovichMaykut and Perovich, 1987; Reference Perovich and MaykutPerovich and Maykut, 1990; Reference SteeleSteele, 1992). In a simplistic sense, for a given set of atmosphere and ocean conditions, the greater the floe perimeter, the greater the amount of lateral melting. On 15 June the average floe perimeter was 5.5 km per km². By 7 August the amount increased by almost an order of magnitude to 44.2 km per km², even as the ice concentration decreased due to divergence. To provide perspective for this increase, consider a simple example of the perimeter of a chess board. For a comparable increase in perimeter, the board would have to break into its 64 individual squares.

Fig. 6. Histograms showing changes in the icepack and in partitioning of incident solar energy between 15 june and 7 august 1998.

We believe that there was a significant increase in the total amount of lateral melting associated with the increase in floe perimeter. Observations indicated that lateral melt rates varied considerably throughout the summer, but in general were on the order of 0.05−0.1 m d−¹ and were relatively high during the divergence event (Reference PerovichPerovich and others, 1999a). Using the observed perimeters, and assuming a melt rate of 0.1m d⁻¹, the daily change in ice concentration due to lateral melting would be 0.06% d−¹ for 15 June and 0.4% d−¹ on 7 August. This assumes, of course, that there is sufficient heat in the upper ocean to support these melt rates.

Impact on partitioning of solar energy

Understanding the ice-albedo feedback was a major concern of the SHEBA program. To examine the impact of changes in ice morphology on the partitioning of solar energy we will compare results from 15 June 1998 and 7 August 1998 (Fig. 5). We shall simplify and assume that the ice cover on those days consisted of three components: bare ice, melt ponds and leads, with each component represented by a single albedo obtained from observations (Perovich and others, in press b). This is a reasonable simplification since we know that there was little change in the albedo for bare ice, which remained white, and for leads (Reference Pegau and PaulsonPegau and Paulson, 1999). Some error is introduced by the variability of pond albedos. Observed albedos for 15 June were 0.70 for bare ice and 0.35 for melt ponds. For 7 August, the albedo for bare ice was 0.67 and 0.15 for melt ponds. A lead albedo of 0.06 was used for both days (Reference Pegau and PaulsonPegau and Paulson, 1999). The average albedo for each day was determined by combining the fractional areas of ice, ponds and leads with their respective albedos (Fig. 6).

Estimating the areal average of the solar energy absorbed in the ice and transmitted to the ocean was more difficult. The transmittance is the fraction of the incident solar energy transmitted to the ocean. Not only were there fewer observations of light transmittance, but transmittance is dependent on the ice thickness. Based on observations, we selected transmittance values for ponds and white ice for 15 June (7%, 2%) and 7 August (12%, 5%). This analysis did not include any thickness differences. This approach is an oversimplification, but will still provide a rough estimate of the partitioning of the incident solar energy. Radiative transfer modeling and thickness distribution data are needed for a more accurate appraisal.

The areally averaged albedo decreased from 0.67 on 15 June to 0.46 on 7 August. There was no change in the lead albedo and little change in the white-ice albedo during this time. Therefore, this 30% decrease in albedo was due to a decrease in the melt-pond albedo and an increase in the amount of open water. Associated with the decrease in albedo was an increase in the amount of solar energy input into the ice-ocean system. There was a modest increase in the energy absorbed by the ice from 28% to 31 %, but the major change was a more than four-fold increase in the energy deposited in the ocean from 5% to 23%. This increase was a direct consequence of the increase in the pond fraction and in the amount of open water.