Airborne observations of summertime surface features and their effect on surface albedo during SHEBA

Mark A. Tschudi, J.A. Curry, J.A.Maslanik
Dept. of Aerospace Engineering, University of Colorado, Boulder, Colorado

Abstract. Aircraft observations of the arctic surface were obtained during the recent Surface Heat Budget of the Arctic Ocean (SHEBA) and FIRE Arctic Clouds experiments. A series of images were created from a downward-looking video camera that was mounted on the underside of a research aircraft as it flew in the vicinity of the SHEBA camp, which was on the pack ice in the Beaufort and East Chukchi Seas. These data are processed to determine the distribution of melt ponds and open water during five flights in July 1998. Melt pond and open water coverage vary between 25-34% and 5-9%, respectively. Coincident observations of albedo derived from upward and downward-looking broadband shortwave radiometers indicate that the albedo decreased through July until the last week, when the albedo began to rise. Pond fraction was observed to vary inversely with albedo through July. The albedo can be crudely constructed by applying a linear combination of pond, open water and ice fraction to their respective average albedo.

1. Introduction

Modifications to the arctic surface albedo that result from surface melting may contribute to the sea ice-albedo feedback [e.g. Curry et al., 1995]. During the arctic summer, the absorption of solar radiation by sea ice, and consequently its melt rate, is accelerated by the formation of melt ponds and open water. Measurements of melt pond characteristics [Perovich, 1998; Morassutti and LeDrew, 1996; De Abreu et al., 1995] have shown that ponds serve to substantially lower the albedo from typical values relative to bare or snow-covered ice. This reduction in albedo was also observed during the Surface Heat Budget of the Arctic Ocean (SHEBA) experiment [Perovich et al., 1999a]. Pre-melt (May 1998) surface albedo at SHEBA ranged from 0.67 to 0.85, depending in part on overhead sky conditions. By July, melt ponds were ubiquitous and influenced the reduction of surface albedo to a range of 0.42 to 0.56 [Curry et al., 1999a].

At the onset of melt over sea ice, water from melting snow can either evaporate, run off into the ocean directly, or collect in depressed areas on the ice to form melt ponds. Younger ice is relatively flat compared to multiyear ice, so the meltwater tends to spread over a larger area than it would over older deformed ice, where the surface melt collects in pockets within deformed areas. On first-year ice, the ponds may align their major axis with melting sastrugi, in contrast to the seemingly random pattern of ponds observed on multiyear ice. As the melt season progresses, the ponds deepen and decrease in perimeter, since the underlying melt rate is 2-3 times faster than the surrounding ice [Fetterer and Untersteiner, 1998]. This process continues until late summer, when the ponds begin to freeze and are eventually covered by snowfall.

To properly represent the surface energy budget in the Arctic, models may need to include formulations for the treatment of melt ponds. A comparison between observed and modeled shortwave radiation in the Arctic by Briegleb and Bromwich [1997] concludes that the absence of melt ponds in the model studied results in an excess of 10-20 W/m2 shortwave absorption during early summer and a deficiency of 20-40 W/m2 during late summer.

The sea ice model by Ebert and Curry [1993] was the first to include an explicit treatment of melt ponds. The sea ice model by Schramm et al. [1997] distinguishes between ponds on first-year ice and multiyear ice. Each ice type has a varying pond fraction over the melt season, beginning with a maximum value and decreasing to a minimum value over the course of 30 days. The albedo of each pond is a function of ice age, pond depth and ice thickness. General circulation models (GCM’s) typically do not treat the effect of albedo from pond cover specifically. Instead, albedo over sea ice is typically parameterized as a function of ice surface temperature and snow accumulation [Barry, 1996; Curry et al., 1999b].

In addition to the role of melt ponds on surface energy flux, ponding also affects the accuracy of remotely-sensed sea ice concentration. Passive microwave data cannot differentiate between ponds and true open water, thus leading to errors in estimates of ice concentration [Cavalieri et al., 1984; Fetterer and Untersteiner, 1998]. Knowledge of melt pond fraction may be used to reclassify the ponded ice and lead to a more accurate determination of concentration.

Melt pond areal coverage has been measured in several Arctic locations during field experiments [Tucker, 1999; Eicken, 1994; Maykut, 1992; Nazintsev, 1964] and has been observed from aircraft [El Naggar et al., 1998; Tschudi et al., 1997] and high-resolution satellite imagery [Fetterer and Untersteiner, 1998]. Albedos of individual ponds have been observed from surface-based measurements [Perovich, 1998; Morassutti and LeDrew, 1996; De Abreu et al., 1995]. The effect of varying pond and open water fraction on the areal albedo has been addressed by Langleben [1968], but these observations were performed from a fixed platform, limiting the area of coverage.

In this paper, we investigate the summertime evolution of the areal coverage of melt ponds and open water in the Arctic by examining video obtained during five low-altitude flight patterns over pack ice in the Beaufort and East Chukchi Seas during July 1998. The effect of ponds and open water on albedo is addressed by comparing coincident hemispheric shortwave upwelling and downwelling radiation data with the distribution of these surface features.

The SHEBA and FIRE Arctic Clouds experiments provided several opportunities to observe the arctic surface from the National Center for Atmospheric Research (NCAR) C-130 research aircraft. A suite of remote sensing instruments was mounted on the aircraft during 16 flights during May and July 1998. Selected data from these flights are available at the University of Colorado SHEBA homepage ( and will be available via the NCAR Joint Office for Science Support (JOSS). This study utilizes observations from a downward-looking video camera as well as upward- and downward-looking NCAR-modified Eppley PSP hemispheric shortwave radiation pyranometers ( and Total Solar Broadband Radiometers [Valero 1982], which were mounted on the NCAR C-130.

2. Video data analysis

A Sony XC-999/999P downward-looking color video camera continuously recorded during each flight. The camera has a 4.8 x 6.4 mm field of view, creating an image size of 160 x 213 m and resolution of 34 cm at an altitude of 200 m. Flight patterns during five days in July 1998 were flown specifically for video acquisition in the vicinity of the Des Groseilliers at an altitude of approximately 200 m. The patterns were designed to cover a 20 x 20 km grid and to make use of the relatively fine resolution of the video camera at this low altitude. Figure 1 illustrates the typical pattern that was flown for video observation of the surface. The "X" denotes the position of the ship. Start and end positions for the pattern are also shown. Overcast conditions overhead prevailed during the five days that video patterns were flown.

JPEG snapshots of the video are created using standard commercial software that interface a video cassette recorder to a personal computer. The software has an automatic frame-grabbing feature that creates a new image in a random fashion. This is done using the software’s time-delay between snaps, which varies through the run of the program. Thus, any location along the flight pattern has an equal chance of being sampled, resulting in a statistically random sample. Table 1 summarizes information about each image set. The latitude and longitude given in the table is the location of the center of each pattern. All patterns were centered over the Des Groseilliers, except on July 24, when the grid was moved a few miles southeast of the SHEBA ship to avoid intervening clouds.

A computer program processes the video images to calculate melt pond and open water areal fraction as described by Tschudi et al. [1997]. To summarize, the program reads each image into a 3x480x640 array that contains the red, green and blue digital values for each pixel. The ponds are discriminated from the surrounding pack ice because the albedo of ponds is substantially lower than pack ice. In addition, the pond reflectance is greater in the blue portion of the spectrum than the red [Grenfell and Maykut, 1977], compared to the relatively flat spectral signature of the surrounding ice. Using these criteria, pond pixels are identified, and pond areal fraction, as well as individual pond area, is computed. The program also calculates the areal coverage of open water by using its low reflectance to separate it from first-year and multiyear ice, and its flat spectral return to distinguish it from the ponds.

A despeckle filter is applied to the data during processing, which eliminates small (one or 2 pixels in a single dimension) areas that initially pass the pond or open water threshold tests. These effects appear to be insignificant at the low flight altitude used, but we are investigating this issue further. Since the video camera’s resolution at 200 m altitude is about 34 cm, the minimum pond diameter after applying the filter is about 1m.

Threshold values for pond, open water, and bare ice identification are automatically determined based on the average scene brightness. The image-to-image threshold adjustment is necessitated by the changing overhead sky conditions during each flight. The ability of the computer program to separate features with differing reflectance permits identification of portions of melt ponds that have melted completely through the underlying ice. These "melt holes" have a reflectance similar to open water, and are therefore included in the open water statistics. The individual pond areas for each image are computed, using the number of pixels the pond covers and the resolution of the video camera at the scene altitude. A "check image" is produced to visually verify the identified surface features, with coloring for the ponds and open water. Figures 2 and 3 show a sample image before and after processing. Note that the portions of the image that are near the edge and the area containing the time stamp are not processed, since this would lead to erroneous pond size and areal coverage statistics. In Figure 3, ponds are colored white, open water orange, and the surrounding ice is colored green. The program erroneously identifies lead edge pixels as melt pond pixels, but analysis has shown that the resulting error in pond fraction is less than one-tenth of one percent.

3. Melt pond and open water distribution

3.1 Areal fraction

Table 2 depicts the melt pond and open water coverage for the five SHEBA flights analyzed. The column in Table 2 entitled "melt pond percentage" represents the total areal percentage of ponds, not including the melt holes. Since the melt holes are spectrally similar to open water, they are included in the open water fraction, as noted earlier. The ‘± ’ values indicate the standard deviation of the areal fractions for each flight. The standard deviation for the two surface types should serve as a limit to observe areal fraction differences between days. For instance, the "increase" in pond fraction from July 8 to July 15 is a small fraction of the standard deviation of pond percentage for either day, so this change is statistically insignificant. However, the increase in pond fraction between July 8 and July 24 exceeds the standard deviation. The relatively large standard deviation for melt pond fraction reflects the inhomogeneity of the ice surface.

The melt pond percentage increases as the melt season progresses through July, until July 26, when a decrease is found. Note that the July 24 observations were not over the same area as the other flights, since low clouds obscured the surface in the vicinity of the ship. Instead, the flight pattern was performed two miles southeast of the ship. However, the pond fraction for July 26 still shows a decrease when compared to the July 18 analysis. The continuing melt process provides more water to create new ponds in pack ice depressions and/or enlarge existing ponds. Surface reports indicated that more channels between ponds were being formed. These channels are included in the melt pond coverage, as they exhibit the same spectral properties as the ponds. By the last flight in July, the breakup of the pack ice may have contributed to the reduction of pond fraction, due to increased lateral drainage.

The open water percentage increases through July (with the exception of July 18), which is probably due to the divergence and thinning of the ice pack. The large values for the standard deviation reflect the large image-to-image variability in open water. Some images consisted of areas where the pack had melted through, exhibiting a relatively large percentage of open water, while other images showed no such areas. Reports from the surface indicated that many small ice floes were melting as they traveled through the relatively warm (2-4 ° C) open water and that ice on the perimeter of leads was beginning to break up, with cracks melting from the bottom up. The warmer water was also beginning to spread under the ice, further accelerating the melt process.

During the same period of observation, Perovich et al. [1999b] examined the evolution of pond fraction and other statistics from surface-based observations along a 200 m survey line, referred to hereafter as the "albedo line." The weekly observations were performed every 2.5 m along this line, which was located within the grid flown by the C-130. Pond fraction was found to increase from 23% to 37% during July along the albedo line, and the maximum pond fraction reached 38% on August 8, in contrast to the July 24 maximum observed from the C-130.

These airborne and surface-based observations are generally consistent with previous measurements. Perovich et al. [1998] examined helicopter images during the height of the Beaufort Sea melt season in 1994 and found ponds to cover 13% of the ice area, which would be reduced somewhat if this value represented the fraction of the complete viewing area (i.e. including the open water). Lapp [1982] classified pond coverage according to ice type in the Beaufort Sea and Canadian Archipelago during early August 1980, with pond coverage averaging 28% on multiyear ice and 36% on first-year ice. El Naggar et al. [1998] approximated pond coverage through the melt season in the Arctic using airborne observations from a line scanning camera over the North Water Polynya. A polynomial fit through a pond fraction time series yielded a maximum pond fraction of 22% at the peak of the melt season. Pond coverage at the North Pole drifting stations [Nazintsev 1964] reached a maximum of 25% in mid-July, with local coverage as high as 45%. Maykut [1992] found pond coverage to reach 50-60% on multiyear ice at the peak of the melt season. The statistics summarized above demonstrate the high temporal and spatial variability of pond coverage in the Arctic.

3.2 Individual pond size

Melt pond area is computed by multiplying the area within one video pixel by the number of pixels contained in each pond. Assuming a square pixel, the area of each pixel at 200 m altitude is 0.12 m2. By applying the despeckle filter, the minimum pond area detected is 0.24 m2. The error in measuring pond area is a function of the perimeter of the pond, since the pixels at the edge of the pond may be mixed (i.e. ice and pond).

A histogram of individual pond areas for the July 8 video flight pattern is illustrated in Figure 4. Interestingly, all of the flights had similar pond area distributions, with correlation between flight segments ranging from 0.95 — 0.98. Smaller ponds (< 100 m2) dominate the histogram, and the number of ponds decrease exponentially with increasing size. A relationship between the size of a pond (S, in m2) and the number of ponds (N) found with size S is given by:

N = 1800 / (1 + 0.3S1.3)

This relationship is plotted as the smooth curve in Figure 4 and correlates well with the observations (correlation = 0.945). To represent this melt pond distribution over a prescribed area, it is more convenient to determine the areal percentage of each melt pond size within the region. Since NS is the total area covered by a pond of size S, and the areal coverage for this flight was 7.95 km2, then the areal coverage (A) of the ponds having a size S is:

A = (2.26 x 10-4 S) / (1 + 0.3S1.3)

Tucker et al. [1999] also observed that most ponds were small (a median size of 14 m2) after analyzing helicopter photographs taken between 76° N and 84° N during a transect across the Arctic Ocean in August 1994. Several pond depth measurements were recorded during this cruise, and it was found that pond depth bore no relation to pond area.

4. Albedo vs. surface feature coverage

The lower albedo of ponds and open water may influence the broadband albedo observed from the shortwave radiometers. Detailed ice models calculate albedo for specific surface types, while less detailed models treat albedo as a "bulk" albedo representing a combination of surfaces. It is therefore interesting to test whether broadband albedo can be approximated by applying a linear combination of the albedos for pack ice, ponds and open water.

The upward- and downward-looking hemispheric shortwave Eppley PSP pyranometers enabled determination of the surface albedo from the C-130 aircraft. These radiometers receive shortwave radiation integrated between 0.285 and 2.80 m m through a cosine filter. These data were corrected for pitch and roll and released as 1 Hz datasets through NCAR’s Atmospheric Technology Division (ATD).

Another pair of broadband radiometers, the Total Solar Broadband Radiometers (TSBR’s) [Valero, 1982], were mounted on the C-130 to observe the downwelling and upwelling shortwave radiation during SHEBA. The TSBR’s integrate the solar irradiance between 0.224 and 3.91 micrometers. Preliminary comparisons between the Eppley and TSBR observations indicate that the Eppley may have some directional dependence when observing the downwelling solar radiation. The TSBR radiometers did not exhibit this dependence, but did not show as much along-track variability in upwelling radiation as the Eppley. Furthermore, the Eppley radiometers are not as well-calibrated as the TSBR’s. For these reasons, the Eppley albedos have been adjusted for this study to reflect the average albedo that was observed by the TSBR. The spatial variability within the Eppley observations are retained with this technique, allowing for comparison with the change in the surface type recorded by the video camera. Note that the TSBR responds to a larger spectral region than the Eppley’s, but the additional incoming radiation is minimal, given the low reflectance of the arctic surface in the ultraviolet and near-infrared wavelengths.

The area viewed by the Eppley and TSBR instruments (hereafter called the hemispheric radiometers) is much larger than the area viewed by the video camera. In terms of comparisons to the video observations, part of the problem this introduces is resolved by the hemispheric radiometers’ cosine response, since the influence to the radiometers is reduced as the look angle is increased.

4.1 Correspondence of video and albedo measurements

The area viewed by a snapshot of the video is responsible for a portion of the incoming radiation to the downward-looking hemispheric radiometers. The swath of the video camera, coupled with the hemispheric view and cosine filter response of the pyranometers, yields the video camera’s area of influence on the pyranometers. If the video camera had a hemispheric view, then coincident observations would yield a 100% area of influence. However, since a video image only accounts for a fraction of the Eppley’s and TSBR’s hemispheric view, the area of influence is reduced. Coupling this viewing area mismatch with the cosine response of the hemispheric radiometers yields an area of influence defined as:

Ic = sin [tan-1(0.53 hv/hE)] and

Ia = sin [tan-1(0.39 hv/hE)] ,

where Ic is the area of influence (in percent) in the cross-track direction and Ia is in the along-track direction; hv and hE are the altitudes of the aircraft for the video and hemispheric radiometer observations, respectively. The area of influence is 47% cross-track and 36% along-track when the radiometers and video camera are viewing the surface coincidentally (i.e. when hE=hv). Therefore the majority of the shortwave radiation reflected from the surface to the hemispheric radiometers is coming from outside of the view of the video camera.

To correct this viewing area mismatch, two flight segments that spanned approximately the same ground coverage at different altitudes were selected from the July 26, 1998 flight. A pass at 81 m is used for video observation, while the Eppley and TSBR observations are obtained from an altitude of 34 m. The Eppley and TSBR hemispheric observation is at a lower altitude, yielding a narrower view of the surface. The video observation area will thus influence a larger percentage of the hemispheric viewing area of the Eppley and TSBR. With the stacked flight segments, the influence on the shortwave pyranometers of the video viewing area is 78% cross-track and 68% along-track, significantly greater than coincident observations (47% and 36%, respectively).

Since the 34-m and 81-m flight tracks do not precisely coincide in lateral position, some additional geometry is needed to evaluate the cross-track influence as a function of the horizontal offset between the two flight tracks. The influence with no offset (i.e. the upper flight track is colinear with the lower track) is 78%. Track offsets ranged from 9 m to 66 m, reducing the across-track influence to a maximum of 77% and a minimum of 19% (see Figure 5). Note that the area of influence falls below 47% (the value for coincident observations) for a portion of this segment.

The along-track influence is also dependent on how well the video and shortwave hemispheric observations coincide in time, ranging from a maximum of 68% to a minimum of 3%. Unfortunately, fractions of seconds are not measured for video observations, so no measure of along-track influence can be made. However, an assessment of the range of influence is beneficial for comparison with coincident observations. Figure 6 depicts the along-track area of influence, based on the offset (in meters), using a typical ground speed of 100m/s. To exceed the 36% influence obtained from coincident observations, the stacked views must be offset by no more than 38 m, or 0.38 s of flight time. Note that, in order to obtain continuous video coverage, some coverage statistics were computed for 2 snapshots, if they both were time-tagged for the same second. This enhances the along-track area of influence and ensures that there is no gap in along-track coverage.

4.2 Relation between broadband albedo and surface characteristics

Figure 7 depicts the adjusted surface albedo, computed by dividing the downward-looking Eppley pyranometer observation by the coincident upward-looking observation and adjusting the average albedo to match the TSBR’s. The flight tracks occurred under cloudy skies; therefore diffuse-sky albedo is observed by the broadband radiometers. The C-130 flew this segment in 38 s at an altitude of 34 m. Continuous images were digitized from video as the C-130 traversed approximately the same flight line at 81 m. Pond and open water fractions computed using the video snapshots are plotted for each image. During one interval along the stacked flight segments, the track offset became large enough to yield an area of influence that was less than coincident observations (see Figure 5). Although the area of influence would be improved by using coincident observations, the albedo observed from a higher altitude is typically higher than at the lower altitude, due to the increase in diffuse radiation observed by the sensor. For this reason, the low-altitude albedo is still shown.

How does the albedo observed from the broadband radiometers relate to the areal coverage and individual albedos of ice, melt ponds and open water? Can a broadband albedo be constructed as a linear combination of surface feature albedo, i.e. according to:

a BB = S a i,

where a BB is the broadband surface albedo and a i is the broadband albedo for each surface type (melt pond, open water and pack ice)?

Using the derived areal fractions of these three surface types, an iterative algorithm varies the albedo of pond, water and ice and computes the resulting albedo curve. The process is continued until the maximum correlation with the observed albedo is obtained, producing an average albedo for the three surface types. To determine the validity of this technique, the constructed curve should correlate well with the observation by the radiometers, and the derived albedos for each surface type should be within the range of acceptable values. Note that the bounds of iteration intentionally exceed acceptable surface type albedos, to check that the routine derives albedos that are realistic. The algorithm computes a maximum correlation between observed and constructed albedo of 0.643, yielding pond, water and bare ice albedos of 0.26, 0.07 and 0.60, respectively. Note that this correlation improves to 0.712 if we exclude the portion of the time series when the across-track area of influence is below 40% (see Figure 5). This technique demonstrates that the broadband surface albedo may be crudely constructed as a linear combination of the individual surface type albedo.

A few approximations reduce the correlation between the measured and constructed albedo. As surface studies have shown, albedos for melt ponds vary, depending on the underlying ice type and amount of sediment in the pond and perhaps, to a lesser extent, pond depth. Multiyear ice, which makes up the highest areal fraction of the surface in this area, is not homogeneous either, as hummocks, ridges, differences in ice makeup and distribution of sediment all affect the reflectance. Albedo is also dependent on the overhead sky conditions, as cloudy skies tend to raise albedo due to the increased forward scattering for preferred wavelengths through the clouds. Overcast conditions prevailed during both flight segments, as evidenced by the consistent downwelling solar radiation observed by the Eppley and TSBR pyranometers. No attempt is made to account for pitch and roll with the video data, which will contribute to a reduction in the area of influence. Comparisons between the pond, open water and ice albedo derived here and literature values should also take into account what spectral region is being measured. The visible portion of the spectrum has the largest reflectance, while the near infrared (NIR) wavelengths exhibit minimal reflectance. Therefore, radiometers that integrate over more of the NIR region will observe a slightly lower albedo when observing the same area than a radiometer with a narrower bandwidth.

Two melt ponds under the flight path were examined in detail on the surface [Perovich et al. 1999b]. Both had albedos of about 0.17 on July 26, less than the average value of 0.26 extracted from the aircraft observations. It is interesting to note that one of the ponds analyzed on the surface had an albedo over 0.4 a few weeks earlier, demonstrating the high temporal variability of the pond. Spatial variability of pond albedo is also large, as observed by viewing the differing brightness of ponds on the C-130 video. White ice albedo from the surface observations was found to be 0.61, which agrees well with the ice albedo computed in this study (0.60).

During a summer 1993 cruise through the Eurasian sector of the Arctic, Eicken et al. [1994] found pond albedo to range from 0.14 (dirty) to 0.30 (clean, less than 30 cm deep). Except for extremely dirty ice, the typical ice albedo ranged between 0.55-0.59. Perovich [1998] cites other surface studies [Chernigovskiy, 1963; Langleben, 1971; Grenfell and Maykut, 1977] that observed albedos of 0.10 for leads, 0.2-0.4 for ponds, and 0.5-0.7 for bare ice.

The broadband hemispheric albedo for each flight is plotted in Figure 8, along with the melt pond and open water fraction derived from the airborne video camera. Note that the albedo plotted here represents the average for each date as recorded by the Eppley pyranometer, adjusted to the average TSBR value averaged over all of these patterns (to preserve the variability sensed by the Eppley’s). The evolution of surface albedo through July varies inversely with the change in the observed melt pond fraction. As the pond fraction increased from July 8 to 24, the albedo decreased; by July 26 the pond fraction had decreased, while the albedo had increased. Interestingly, the open water fraction was increasing during the entire period, but the area-averaged albedo variation was most responsive to changes in pond fraction. This is due in part to greater abundance of pond area over open water.

To investigate how the albedo varied across the 20 x 20 km area traversed by the video patterns, results from the grid patterns flown on July 8, 18 and 26 were examined. The Des Groseilliers was located in the center of each pattern. Each pattern was divided into 25 boxes, 4 km on a side. For each box, the adjusted Eppley albedo is averaged and plotted, yielding a spatial distribution of albedo across each pattern (Figure 9). For the July 26 flight, the grid is narrower that the two other flights, so the averaging is performed for only four columns instead of five, preserving the individual box dimensions of 4 x 4 km. Table 3 summarizes the adjusted Eppley albedo observed within these boxes for each flight. The difference across the 20 x 20 km grid was greatest on July 18. Interestingly, this date has the largest observed average albedo for the five July flights (see Figure 8). The minimum variability is found on July 8, when the average albedo is lowest for the three flights. The variability increase with a higher albedo results from persistent areas of low reflectance, where open water and/or ponds are abundant.

5. Conclusions

Utilization of video camera imagery from the NCAR C-130 during SHEBA and the FIRE Arctic Clouds Experiment has been effective in providing melt pond and open water statistics during the 1998 summer melt period in the Beaufort and East Chukchi Seas. Melt pond areal fraction was observed to increase from July 8 to July 24 from 25 to 34% before decreasing to 26% by July 26. This suggests that the pond coverage maximum in July may have occurred between July 18 and 26. Surface measurements by Perovich et al. [1999b] found the pond fraction to increase throughout July, attaining a maximum pond fraction on August 8, after the C-130 campaign was completed. Differences in pond fraction between Perovich et al. and this study may in part be attributed to the differing scales of measurement, since the aircraft was observing a 20 x 20 km area, while the surface measurements were made over a 200 m line.

The pond fraction increase is probably due to the interconnecting of ponds during the study period, as well as the expansion of ponds as they fill with additional meltwater and melt laterally. A portion of the melt ponds had melted through the underlying ice. The part of each pond that had melted through (the melt hole) exhibited a reflectance that is similar to open water, while the remainder of the pond retained its spectral signature. Ponds with melt holes allow a hydraulic connection to the ocean, and thus (if some mixing is assumed) contain water with a higher salinity than ponds with no melt holes. After the autumn freezeup, the more saline ice covering the pond with a melt hole would likely appear as first-year ice, leading to a possible misidentification of ice type from satellite observations. The portion of ponds containing melt holes will be investigated.

The fractional area of open water, which includes leads and melt holes, rose from 5.1 to 8.7% through the sampled July 1998 period. The transformation of portions of the surface from ponded to open water, along with the thinning ice pack, account in part for the open water increase and corresponding decrease in pond percentage in late July.

The melt pond and open water fractions obtained in this study should assist modelers in evaluating parameterizations of the evolution of the ice pack during the summer melt season. These statistics may also be applied to remove the effect of pond area from passive microwave observations, leading to a more accurate determination of total ice concentration and perhaps concentration of first-year vs. multiyear ice.

The areal albedo can be approximated by applying a linear combination of the albedo of each individual surface type, based on the fractional coverage of each type. This may be a useful tool for improving the representation of the polar surface energy balance in regional and global climate models. Moreover, the evolution of the arctic surface during the melt season may assist modelers with parameterizing the strength of the ice-albedo feedback.

The 1998 melt season in the Beaufort and Chukchi Seas was one of particular interest, since there was a record reduction in sea ice cover [Maslanik et al., 1999] in this region. Open water formed earlier than in previous years in this area. In the autumn of 1997, the SHEBA ice camp was forced to locate on multiyear ice that was thinner than anticipated [McPhee at al., 1998; MacDonald et al. 1999], due to the lack of relatively thick multiyear ice that is typically found in this location. The data described here, in conjunction with the wealth of other observations acquired during SHEBA and FIRE-ACE, should help in determining the causes and effects of these unusual ice conditions.

Acknowledgements. This research was funded by NSF SHEBA and NASA FIRE. We would like to thank Francisco Valero and his research group for making the radiation data available to us. We would also like to thank the NCAR Research Aviation Facility for the support of these flights, as well as all the field participants in FIRE and SHEBA, including the crew of the C-130 and the Des Groseilliers.

6. References

Barry, R.G., The parameterization of surface albedo for sea ice and its snow cover, Progr. In Phys. Geog., 20, 61-77, 1996.

Briegleb, B.P. and D.H. Bromwich, Polar radiation budgets of NCAR CCM3, J. Climate, 11, 1246-1269, 1997.

Cavalieri, D.J., P. Gloersen, W.J. Campbell, Determination of sea ice parameters with the Nimbus-7 SMMR, J. Geophys. Res., 89, 5335-5369, 1984.

Chernigovsky, N.T., Radiational properties of the central Arctic ice cover (in Russian), Tr. Arkt. Antarkt. Nauchno Issled. Inst., 253, 249-260, 1963.

Curry, J.A., J. L. Schramm and E. Ebert, Sea ice-albedo climate feedback mechanism, J. Climate, 8, 240-247, 1995.

Curry, J.A. et al., Bulletin Amer. Met. Soc., submitted, 1999a.

Curry, J.A., J.L. Schramm, D.K. Perovich and J.O. Pinto, Applications of SHEBA/FIRE data to evaluation of snow/ice albedo parameterizations, J. Geophys. Res., this issue, 1999b.

De Abreu, R.A., D.G. Barber, K. Mizurak and E.F. LeDrew, Spectral albedo of snow-covered first-year and multi-year sea ice during spring melt, Ann. Glaciol., 347-352, 21, 1995.

Ebert, E.E. and J.A. Curry, An intermediate one-dimensional thermodynamic sea ice model for investigating ice-atmosphere interactions, J. Geophys. Res., 98, 10,085-10,109, 1993.

Eicken, H., T. Martin and E. Reimnitz, Sea ice conditions along the cruise track, In Futterer., D.K., editor, The Expedition Arctic ’93. Leg ARK-IV/4 of RV Polarstern 1993, Berichte zur Polarforschung, 149, 42-47, 1994.

El Naggar, S., C. Garrity and R.O. Ramseier, The modeling of sea ice melt-water ponds for the High Arctic using an airborne line scan camera, and applied to Satellite Special Sensor Microwave / Imager (SSM/I), Int. J. Remote Sensing, 19, 2373-2394, 1998.

Fetterer, F. and N. Untersteiner, Observations of melt ponds on Arctic sea ice, J. Geophys. Res., 103, 24,821-24,835, 1998.

Grenfell, T.C. and G.A. Maykut, The optical properties of ice and snow in the Arctic Basin, J. Glaciol., 18, 445-463, 1977.

Langleben, M.P., Albedo of melting sea ice in the southern Beaufort Sea., J. Glaciol., 10, 101-104, 1971.

Langleben, M.P., Albedo measurements of an arctic ice cover from high towers, J. Glaciol., 7, 289-297, 1968.

Lapp, D., Study on ice meltponds, Toronto, Ont., Polar Research and Engineering for Atmospheric Environment Service. (Contract Report), 1982.

Macdonald, R.W, E.C. Carmack, F.A. McLaughlin, K.K. Falkner and J.H. Swift, Connections among ice, runoff, and atmospheric forcing in the Beaufort Gyre, Geophys. Res. Lett., 26, 2223-2226, 1999.

Maslanik, J.A., M. C. Serreze and T. Agnew, On the record reduction in 1998 Western Arctic sea ice cover, Geophys. Res. Lett., 26, 1905-1908, 1999.

Maykut, G.A., T.C Grenfell and W.F. Weeks, On estimating the spatial and temporal variations in the properties of ice in the polar oceans, J. Marine Systems, 3, 41-72, 1992.

McPhee, M.G., T.P. Stanton, J.H. Morison, and D.G. Martinson, Freshening of the upper ocean in the central Arctic: Is perrenial sea ice disappearing?, Geophys. Res. Lett., 25, 1729-1732, 1998.

Morassutti, M.P. and E.F. LeDrew, Albedo and depth of melt ponds on sea ice, Intl. J. Clim., 16, 817-838, 1996.

Nazintsev, Y.L., Heat balance of the surface of multiyear sea ice in the central Arctic (in Russian), Trudy Arkticheskii Antarkticheskii Nauchno-Issledovatelskii Institute, 267, 110-126, 1964.

Perovich, D.K. et al., Year on ice gives climate insights, EOS Transactions, 80, 481-486, 1999a.

Perovich, D.K., T.C. Grenfell, B. Light, J.A. Richter-Menge, M. Sturm, W.B. Tucker III, H. Eicken, G.A. Maykut and B. Elder, SHEBA: Snow and ice studies CD-ROM, 1999b.

Perovich, D.K., C.S. Roesler and W. Scott Pegau, Variability in arctic sea ice optical properties, J. Geoph. Res., 103, 1193-1208, 1998.

Schramm, J.L., M.M. Holland and J.A. Curry, Modeling the thermodynamics of a sea ice thickness distribution, J. Geophys. Res., 102, 23,079-23,091, 1997.

Tschudi, M.A., J.A. Maslanik and J.A. Curry, Determination of areal surface feature coverage in the Beaufort Sea using aircraft video data, Ann. Glaciol., 25, 434-438, 1997.

Tucker, W.B., III, A.J. Gow, D.A. Meese and H.W. Bosworth, Physical characteristics of summer sea ice across the Arctic Ocean, J. Geophys. Res., 104, 1489-1504, 1999.

Valero, F.P.J., W.J.Y. Gore, and L.P.M. Giver, Radiative flux measurements in the troposphere, Appl. Opt., 21, 831-838, 1982.



Table 1: Video Image Data.

Date (1998)

Time of pattern (UTC)

Latitude (° N)

Longitude (° W)

# Images

Area covered (km2)

Mean altitude (m)

July 8

22:47 — 23:47






July 15

22:39 — 23:04






July 18

21:40 — 22:23






July 24

00:18 — 00:41






July 26

20:37 — 21:25












Table 2: SHEBA Surface Morphology.


Melt pond %

Open water %

July 8

24.6 ± 6.2

5.1 ± 6.5

July 15

24.9 ± 5.5

5.5 ± 6.3

July 18

28.0 ± 5.5

4.8 ± 6.2

July 24

34.1 ± 8.4

5.9 ± 9.9

July 26

26.3 ± 6.6

8.7 ± 6.2

Table 3: Albedo variability across 20 x 20 km grid

Flight Date

Min box albedo

Max box albedo

Max — Min

Standard deviation

July 8





July 18





July 26





Figure 2: Snapped video image


Figure 3: Image after processing

Figure 4: Individual pond areas.

Figure 7: Adjusted and constructed albedo