Floe Sizes in the East Antarctic Sea Ice Zone estimated using combined SAR and Field data

Abstract

In the polar regions, sea ice is an important link between the coupled ocean/atmosphere system. The salt flux to the ocean as the sea ice grows, and the modification of the ocean/atmosphere heat flux due to variations in sea ice concentration and thickness are critical components of the global climate system. In the Southern Ocean, sea ice in mid-winter can extend over 18 million square km, of which over 4 million square kilometres may be open water. The size of these open water areas is frequently on the order of 10’s of metres, and may be difficult to resolve using even high resolution satellite imagery. Yet, small amounts of thin ice and open water can have a significant effect on the regional heat and salt fluxes.

ERS-1 SAR data were collected over a region of East Antarctica around 65S, and 140 E during winter of 1995. These SAR images are combined with data collected during an intensive field program during August, 1995 in the same region. These field data include extensive aerial photography with a resolution of less than 1m which allows us to accurately classify the ice type, particularly thin ice, brash and open water. We use these field data to classify a SAR image, with the goal of estimating the ocean-atmosphere heat fluxes. We find that thick, multi-year floes increase the average backscatter and can be an indication of increased open water regions.
Keywords: Sea Ice, Antarctica, heat flux

Introduction

Numerous studies have investigated the dependence of the backscatter signature of sea ice on variables such as temperature, ice type, and snow characteristics using synthetic aperture radar (SAR) data collected from satellites (e.g. Kwok and Cunningham, 1994, Comiso et al., 1992, Onstott, 1992, Carsey, 1985, Kwock et al., 1992). Stern et al. (1995) used sequential SAR images to estimate open water formation rates, and ice motion tracking is nearly routine in the Arctic (Kwok et al., 1990). These studies have relied largely on the ability to identify individual ice floes and/or leads, whether for ice type classification or for strain and drift estimation. Other studies (e.g. Comiso et al., 1992, Morris and Jeffries, in press) have shown the utility of satellite-based SAR in identifying different ice regimes, particularly near the ice edge, where floe sizes are often below the resolution of the image. Much of the Antarctic sea ice zone is subject to ocean swells, generated to the north, which persistently break up the ice into smaller floes, often 10’s of metres or less in diameter. The characterization of these regions, in terms of floes sizes, and amount of brash and open water is important in estimating ocean/atmosphere heat fluxes, ice growth rates in winter, and ice melt rates in summer. Here we present results which paramaterize the sea ice in the East Antarctic region in terms of floe sizes which are smaller than the resolution of available SAR images.

Data Collection and Analysis

We collected detailed measurements of snow and ice physical properties during a field experiment in August, 1995 aboard the Australian icebreaker R.S.V. Aurora Australis. A primary goal of the experiment was to characterize the ice growth rates and the ocean/atmosphere heat fluxes in the sea ice region (Worby et al., 1995). To measure ice drift and deformation, several drifting buoys were deployed on the sea ice. These buoys recorded their location routinely using the Global Positioning System (GPS), or the ARGOS positioning system, accurate to about 100m and 300m, respectively. A synthetic aperture radar (SAR) image from ERS-1 was collected at 2330 on August 5, 1995. The SAR image, centered at 63.957^oS and 140.542^oE is shown in Figure 1. Also shown, are the location of 3 buoys with GPS recorderss (Snow White, Happy, Sleepy), one buoy with an ARGOS transmitter (ARG663), and the ship’s position at the time of the SAR overpass. Approximately 2.1 days after the overpass, digital aerial photography was collected from a helicopter from 1.5 km elevation. The flight leg analyzed here is 50 km long, in a generally north-south orientation. The resulting digital black and white photographs are 730 m x 480 m in size, and have a pixel size of about 0.47 m. Based on the buoy drift data, the ice had drifted approximately 24.7 km to the northeast during this time. Consequently, we have corrected the location of the flight to coincide with the location of the same ice floes at the time of the satellite overpass. During the time between the SAR data and the photography data, the sea ice in the region showed a slight divergence, however, the location of the flight relative to the SAR image is estimated to within 5 km. This divergence was primarily in the north, and may result in some of the photos not being correctly located relative to the SAR data, particularly toward the northern part of the flight. This will not effect the results, as we are using the photos simply to indicate general ice conditions, rather than specific locations on the SAR image. The corrected location of this flight is indicated in Figure 1 by a white line in the center of the image. Photos were collected continuously along the flight. The letters A through E correspond to example photos from the flight, which are discussed later. The size of the black squares on the white line is approximately the size of the photos. From August 5, when the SAR image was collected, to August 8 when the photos were taken, the ship travelled northward from the location shown in Figure 1, and then eastward across the flight path. Hourly observations of sea ice type and floe size collected from the ship during this time are used in addition to the photographs to interpret the SAR image.

The SAR image has been geolocated and calibrated using a standard calibration algorithm (Laur, 1992). It (Figure 1) shows two distinct regions, a darker region (Figure 1, south of photo A), where the ice cover is near 100%, and individual leads and floe edges are readily distinguishable. Observations collected in the field indicate that the sea ice to the south was predominantly first year floes, with an undeformed thickness of 50-60 cm (Worby et al., in press). These were vast floes, with virtually no open water. To the north of photo A there is a transition region where there are some leads which can be seen, but in much of this region individual floes are below the resolution of the SAR and cannot be identified in the image. Estimating open water percent, or ice type, and the associated ocean/atmosphere heat fluxes in this region using the SAR image alone is difficult. Backscatter values increase to the north, from an average of about -12.4 dB in the southern portion, to about -9.8 dB in the brighter bands in the northern portion of the image.

Figure 2a shows the southern most image collected along this leg of the flight, and corresponds to the location A in Figure 1. This is the edge of a vast, snow-covered sea ice floe; the darker stripe on the south edge is a lead about 100 m wide, and contains newly forming ice. Figure 2b is the photo corresponding to the location B in Figure 1. The region here is primarily covered with ice floes on the order of 100 m in diameter. There are smaller floes and brash between the larger floes, and about 7% of the area is open water. The open water regions are about 15 m or less in width, less than the size of a single pixel in a SAR image. In the photo shown in Figure 2c, corresponding to location C in Figure 1, the floes continue to decrease in size, averaging about 20m in diameter. The open water area in this photo has increased to 14 %, and individual open water areas are generally less than 10 m across.

Further to the north, the floe sizes continue to decrease, however, a new ice type appears. Figure 2d, corresponding to location D in Figure 1, shows a combination of first-year floes, and thicker, multiyear ice. The multiyear ice can be distinguished by the long shadows cast by the increased height about the water, and the generally more ridged features on the surface. Field observations indicate that this ice is in excess of 5m thick, with a thick (about 2m) snow cover. More details on the structure and possible origin of this ice type can be found in Worby et al. (in press). During this experiment this ice type was never found in concentrations greater than 30%, and it usually covered 10% or less of the area. Similar floes were found by Wadhams et al., (1987) in the eastern Weddell Sea and probably has its origins in the fast ice regions near the coast. Because of it’s thickness, and increased water drag, this ice drifts at a different speed than the surrounding thinner ice. It often occurs in wide bands, with the thinner first year ice accumulating on the upwind side, and open water, or new ice, on the downwind side. An example of this can be seen in the photo, where the majority of the 6% open water and new ice, occurs to the east of the multiyear floes, due to the prevailing westerly winds during this time. The first-year ice floes in this image average about 8 m in diameter, although there are still a number of floes which are as large as 15 m in diameter. The thicker, multiyear floes are larger with an average diameter of 24m.

Figure 2e, corresponding to the location E in Figure 1, is at the northern end of the photo flight. The floe sizes continue to decrease, with only a few floes greater than 12 m in diameter, and an average diameter of 4 m. Much of the region consists of brash, or very small floes, below the resolution of the photography. A few of the thicker floes remain, and are distinguished by their shadows cast on the rest of the ice. There is less than 1% open water in this image, but the large amount of brash may still result in relatively large ocean/atmosphere heat fluxes.

The SAR image was classified first by identifying open water and thin ice areas where possible directly from the SAR image. The wind speeds measured from the ship when the SAR data were collected were relative calm at 5.2 m/s, and the air temperature was -16.2^o C and open water/thin ice shows up dark in the image. Backscatter values of less than -16.5 dB are classified as open water. This value is based on the backscatter value obtained from an obvious new ice or open water region in the SAR image collected during the same orbit, but one frame to the south of the image in Figure 1. For this study we do not distinguish between newly forming nilas (ice less than about 5 cm thick) and open water. We use the aerial photography to classify the remainder of the SAR image based on floe sizes, and the presence of the multiyear floes. We have done this crudely, by manually contouring along the banding seen in the image. There is not a clear demarcation in the aerial photography between floe sizes, rather a gradation in floe sizes, and we don’t expect to find a distinct boundary in the SAR image. Similarly, although there were distinctive bands of multiyear floes, sometimes with concentrations up to 30 %, on a regional scale, there was no clear boundary where multiyear floes were present. We identify regions corresponding to the photos shown in Figure 2. These include vast floes (>500m), medium sized floes (~100m diameter), small floes (~20m diameter), and cake ice floes (<10m diameter). We divide the last category into regions with and without multiyear floes. A small portion of the SAR image north of the photography flight has a lower backscatter. We have no direct observations of this region from the photography flight, however, the ship was in the area shortly before the photo flight took place. Based on observations from the ship , this region probably contains a mix of grease ice, brash and open water. For each of the regions classified we calculate an average backscatter value.

RESULTS

The final classification of the SAR image is shown in Figure 3. The percent concentration of each ice class, the average backscatter, and the percent open water is listed in Table 1. The small open water region (0.5%) estimated based on backscatter values alone agrees with ship based observations collected in the southern portion of the image. However, it is less than observations from the northern portion of the image. By mulitplying the concentration of each ice classification (column 2, table 1) with the open water estimates obtained from the aerial photography (column 4), we calculate the regional contribution of open water within each zone of different ice type (column 5). The total of 4.3% open water and thin ice is an order of magnitude greater than the estimate of 0.5% using backscatter values alone.

Aside from the small open water and new ice regions, the average backscatter is the lowest for vast, first-year floes at -12.4 dB. The backscatter generally increases, as the floe sizes decrease. The bright bands seen in the northern portion of the SAR image where the multiyear floes are found have the highest average backscatter at -9.8 dB. This is lower than the -7 dB average backscatter found by Drinkwater and Lytle (in press) for multiyear Antarctic ice in the Weddell Sea. It should be emphasised that the multiyear ice found in this study has very different physical characteristics. It is thicker, occurs in smaller floes, and has a much thicker snow cover than the multiyear ice found in the Western Weddell Sea. In addition, it was usually found at concentrations at 10% or less. As a result, the average backscatter value is a mix of primarily first year ice, with a relatively small percentage of multiyear ice. Consequently, we expect the multiyear floes alone would have a considerably higher backscatter value. The backscatter from the mixture of grease ice, brash and cake ice is -12.0 dB, similar to the value for vast floes, yet their physical characteristics, in particular, their influence on the ocean/atmosphere heat flux is significantly different. In the SAR image (Figure 1), they have obviously different textural characteristics, and their relative locations within the pack ice imply different ice types. The backscatter in the region with mixed ice types would increase with increasing wind speeds due to wind roughening of the open water surface; the backscatter from vast floes would not be expected to depend on wind speed.

Table 1. Classification of SAR image based on aerial photography.

Conclusions

While this is a tedious method for estimating ice types, it qualitatively uses texture as well as the absolute backscatter to classify the ice in terms of floe size. Over 1500 km of aerial photography was collected, along with SAR images whenever possible during the three week experiment. These cover a wide range of ice types, air temperatures and wind speeds. Although we realize that detailed digital photography is a luxury rarely available, we hope that the results from this data set can be extended to other SAR images where photography is not available, particularly for the identification of bands of multi-year ice. The final estimates of new ice and open water are increased significantly by the addition of sub-pixel regions identified in aerial photography, as compared to estimates from the SAR image alone. The final estimate of open water and thin ice concentration compares well with the ship based observations. In the East Antarctic region where the sea ice is a narrow band, seldom greater than 600 km wide, much of the ice is broken up by the ocean swells generated by storms to the north. Consequently, a large proportion of the ice consists of floes which cannot be individually identified using the SAR data. The amount of open water and brash between theses floes can significantly increase the ocean/atmosphere heat flux. In addition, the identification of regions containing the thicker, multiyear floes will help determine their total ice mass in the region, and their influence on the overall dynamics of the pack ice.

References

Comiso, J.C, T.C. Grenfell, M. Lange, A.W. Lohanick, R.K. Moore and P. Wadhams, 1992:
Microwave Remote Sensing of the Southern Ocean Ice Cover, in Microwave Remote Sensing of Sea Ice, Geophys. Monogr. Ser., 68, edited by F. D. Carsey, pp 243-259 AGU Washington D.C.

Drinkwater, M and V. I. Lytle, in press:
ERS-1 radar and field-observed characteristics of autumn freeze-up in the Weddell Sea, J. Geophys. Res.

Kwock, R., J.C. Curlander, R. McConnell, and S.S. Pang, 1990:
An ice-motion tracking system at the Alaska SAR Facility, IEEE J. Oceanic Eng., 15, pp 44-54.

Kwok, R. and G.F. Cunningham, 1994:
Backscatter Characteristics of the winter ice cover in the Beaufort Sea, J. Geophys. Res. 99, pp 7787-7802.

Laur, H.:
Derivation of Backscattering Coefficient in ERS-1 SAR.PRI Products, European Space Agency Report, ESA ESRIN, 16pp.

Morris, K. and M. O. Jeffries, in press:
Sea Ice characteristics and Seasonal Variability of ERS-1 SAR backscatter in the Bellingshausen Sea, J. Geophs. Res.

Onstott, R.G., 1992:
SAR and scatterometer signatures of sea ice, in Microwave Remote Sensing of Sea Ice, Geophys. Monogr. Ser., 68, edited by F. D. Carsey, pp 73-104, AGU Washington D.C.

Stern, H.L., D.A. Rothrock and R. Kwok, 1995 :
Open water production in Arctic sea ice: Satellite measurements and model parameterizations, J. Geophys. Res. 100, pp 20,601-20,612.

Wadhams, P., M.A. Lange and S.F. Ackley, 1987:
The Ice Thickness Distribution Across the Atlantic Sector of the Antarctic Ocean in Midwinter, J. Geophys. Res. 92, pp 14,535-14,552.

Worby, A.P., N.L. Bindoff, V.I. Lytle, I. Allison, and R.A. Massom, 1996:
Winter Ocean/Sea Interactions in the East Antarctic Pack Ice, EOS, 77(46) pp 453&456-457.

Worby, A.P., R.A. Massom, I. Allison, V.I. Lytle and P. Heil, in press:
East Antarctic Sea Ice: A Review of its Structure Properties and Drift, J. Geophys. Res.

Figure Captions

Figure 1

SAR image collected August 5, 1995 at 2330, orbit 21217 frame 4941. Locations of drifting sea ice buoys and the ship are indicated. The aerial photography fight is shown by a white line in the northern portion of the image. Letters A through E indicate locations of the photos shown in Figure 2.

Figure 2

Aerial photos collected over the sea ice on August 8, 1995 at approximately 0200. Each image is 730 m x 480 m in size, and north is towards the top of the page. The locations of the photos corresponds to letters indicated in Figure 1. a) Vast floe with a lead containing newly forming ice in the south. b) medium floes, 100 m in diameter, c) small floes, 20 m in diameter, d) first-year cake ice, floes less than 10 m in diameter, and thicker multiyear ice 24 m in diameter. Newly forming ice and open water can be seen in the northeast corner and e) first-year cake ice, floes less than 10 m in diameter, with a few multiyear floes.

Figure 3

Classification of the SAR image. The color codes and ice types are indicated.