An explanation for the dark region in the western melt zone of the Greenland ice sheet

An explanation for the dark region in the western melt zone of the Greenland ice sheet I. G. M. Wientjes and J. Oerlemans Institute for Marine and Atmospheric Research (IMAU), Utrecht University, Utrecht, The Netherlands Received: 1 February 2010 – Accepted: 3 February 2010 – Published: 11 February 2010 Correspondence to: I. G. M. Wientjes (i.g.m.wientjes@uu.nl) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
When temperatures rise, the Greenland ice sheet will experience enhanced melting in the ablation zone.Especially in the western part of the Greenland ice sheet, the ablation zone is very wide.Different studies have been carried out to get more insight into the melting behaviour of this area (e.g.Ambach, 1972;Braithwaite and Olesen, 1993).An extensive meteorological field campaign along the K-transect, 67 • N, showed that in this part of the ice sheet solar radiation makes a large contribution to the total melt energy, and that albedo variations play an important role (Oerlemans and Vugts, 1993;Van de Wal, 1992; Van de Wal and Oerlemans, 1994).More recently, Van den Broeke et al. (2008) investigated the high spatial variability of the surface radiation of this area with the help of automatic weather stations.Already in the early nineties it was realized that there is a dark region in the ablation zone, being potentially important for the melt rates in the area ( Van de Wal, 1992).This dark region stretches from 65 • N to 70 • N, at a longitude of around 49 • W (Fig. 1).
A dark appearance implies low radiance.This means that much light is absorbed, involving low spectral albedos and thus enhanced melting.The first explanation of the dark region suggests accumulating of melt water at the surface.The hypothesis is that the water does not drain subglacially due to the cold ice, and that it runs off slowly because of relatively small surface slopes (Knap and Oerlemans, 1996).Greuell (2000) developed this idea further and found a relationship between the amount of melt and the albedo lowering, confirming this hypothesis.Zuo and Oerlemans (1996) showed that a model with meltwater-albedo coupling predicts the albedo pattern and mass-balance profile along the K-transect better than models without this coupling.
Even though meltwater accumulation could be an explanation for the dark region, new investigations of recent satellite images indicate that there might be another reason for the darkening, which probably induces the accumulation of meltwater.In this paper, we examined the dark region with the help of MODIS (Moderate Resolution Imaging Spectroradiometer) and ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite images.On the basis of these images we will show that outcropping of ice containing old dust is a plausible cause of the dark region.
First, we describe the satellite products that we used.From the MODIS images, time series of the position of the dark region were made and the spectral information of the images was analyzed.Finally, we discuss the results and look at the region in more detail with the help of an ASTER image.

Satellite images and techniques
The MODIS instrument provides satellite images with a spatial resolution of 250 m to 1 km and a temporal resolution of 1-2 days.We used Level 1B products, which contain sensor-and geometry-corrected data in 36 spectral bands, of which 20 in the solar spectrum (405-2155 nm).From these products, we studied geolocated radiances.The (almost) cloud-free scenes were selected manually.It became apparent that for some of the reflective solar bands the ice and snow parts are saturated.The remaining 10 bands are used for further analysis.Figure 1  To get more insight into the character of the dark region, we used RGB (Red Green Blue) colour composites.The composites consist of a combination of three different wavelength bands, where each single band is plotted in another colour scheme.For example, a true colour composite is obtained, if wavelength Band 1 (620 to 670 nm, partly covering the red light) is assigned to a red colour scheme, Band 4 (545 to 565 nm, partly covering the green light) to a green colour scheme and Band 3 (459 to 479 nm, partly covering the blue light) to a blue colour scheme.In Fig. 2a, such a composite is shown for the satellite image of 9 August 2007.For image enhancement, a linear contrast stretch is applied to each colour plane of this image.This means that a transformation is performed, whereby the range of intensity values of each colour plane is linearly expanded to make full use of the complete range of available intensity values.Also, to make the dark region stand out even more clearly, a decorrelation stretch is made.This is a method that maximizes the difference between the colour planes by decorrelating their colour values.The result of this stretch is shown in Fig. 2b.Note that these colour composites have no quantitative meaning, but are only designed for qualitative visual interpretation.
In addition to the Level 1B product, we used a daily surface reflectance product, MOD09GA, which had been computed from the MODIS Level 1B product.A correction had been made for the effects of atmospheric gases and aerosols.Again, images with little or no clouds were chosen visually.This MODIS product contains only reflectance values for 7 different wavelength bands; therefore we used it only for the comparison with values from literature.
Finally, we used ASTER images to obtain more information about spatial details.
ASTER delivers sensor-registered radiances in 14 different spectral wavelength bands with a resolution varying between 15 to 90 m.The instrument operated only on request, and ASTER images cover a much smaller spatial area than the MODIS images, so there are only a few images that contain a part of the dark region and are (almost) cloud free.After projection to the Geographic Coordinate System, the latitudes and longitudes of the ASTER images clearly differ from the MODIS images.Therefore, we chose recognizable points visible in both images, and used their MODIS coordinates to define the latitudes and longitudes of the ASTER images ourselves.As a result, the latitudes and longitudes from the MODIS and ASTER images used in this paper coincide with each other.

Time series
For a better understanding of the temporal behaviour of the dark region, we constructed time series.For this purpose, the radiances between 67.75 • N to 68 • N were averaged, along a transect from 50 • W to 48 • W. The profiles of Band 2 (841-876 nm) that were generated in this way, are presented in Fig. 3, for different days during the summer season of 2007.The dark region developed between 49.5 • W to 49 • W. On 25 June the snow was still melting, creating a zone with dark water patches surrounded by brighter snow, making the whole profile more irregular.On 27 August the radiance profile was a bit lower along the whole ice part.This can be explained by the fact that radiance is a quantity that is measured in one direction.Because ice reflects in a non-isotropic way, a slightly different position of the satellite may be the cause of the lower radiance profile.
Figure 4 shows the same kind of profiles (again from Band 2) for different years.For each year, a day in the summer season has been chosen in late July/early August.
Note that varying cloud conditions made it impossible to use the same dates for every year.Again, the dark region showed up between 49.5 • W to 49 • W for the different years, although the lowering of the radiance varied from year to year, especially for the years 2001 and 2004, where no clear minimum can be observed.This may be due to different amounts of melt in different years.

TCD Introduction Conclusions References
Tables Figures

Back Close
Full

Spectral information
To get more insight into the spectral signature of the dark region, radiances for different bands were calculated.These radiances were averaged for the area of the dark region, marked with a blue contour in Fig. 1.These averaged values were calculated for each possible wavelength band.We chose a reference area of brighter ice, closer to the margin, namely between 67.5 • to 67.75 • N and 49.5 • to 49.75 • W, indicated in Fig. 1 with a red contour.For this reference area the radiances were also averaged.In Fig. 5, the results for the reference and the dark region are plotted against the median wavelength of each band for 9 August 2007.For other days the results were quite similar.In the visible part of the spectrum, the dark region clearly has much lower radiances than the reference area, which results in a darker appearance.At wavelengths larger than 800 nm, however, the difference between the values of the dark region and the reference area become much smaller and for wavelengths larger than 900 nm the values for both areas are the same.
In the literature, reflectance curves for different types of ice are often shown instead of radiance curves; therefore, the surface reflectance product of MODIS was also considered.Although this product contains only useful information for a smaller number of bands, it still provides a good indication of the spectral signature.For the reflectance data, we made the same kind of figure as for the radiances.This time, we averaged over an area inside the dark region, from 67.5 • to 68.25 • N and 48.75 • to 49.25 • W, because the reflectance images do not cover the whole dark region as outlined in Fig. 1.We averaged the reflectance values for the same reference area and presented these values against the median wavelength of each band.In Fig. 6 this is shown for 9 August 2007.Again, the major difference is observed in the visible part of the spectrum and becomes smaller with increasing wavelength.Introduction

Conclusions References
Tables Figures

Back Close
Full

Discussion
From Sect. 3 we can conclude that the appearance of the dark region follows the same cycle every year.It becomes visible in the early summer season when snow begins to melt, but it remains at this position during the whole summer and does not become wider even when melting proceeds.Greuell and Knap (2000) determined positions of the slush line, defined as the boundary between the uniformly snow covered area and the area covered with patches of snow, slush and ice.For a year with much melt, 1995, they find the positions for the slush line lying more to the east when summer passes.
They also indicate the location of the dark region in their results, although they explain this feature as meltwater accumulation.However, when we compared their findings with the position of the dark region, we conclude that the dark region remains on the same location, even if the slush zone migrates further to the east than the dark region.This behaviour indicates that the darkening is not simply caused by melt water, but that it reflects a property of the ice.
The spectral signature of the dark region (Sect.4) shows a pattern of lower reflectance (radiance) in the visible part of the spectrum and less or no difference in the infrared part, when it is compared to the reflectance (radiance) of the reference ice.Zeng et al. (1984) found the same kind of pattern for the reflectance of dirty ice compared to that of clean ice.These measurements are also shown in Fig. 6.Takeuchi et al. (2001) reported a lowering of the spectral albedo of ice with cryoconite compared to clean bare ice for wavelengths till 950 nm, where the lowering already becomes smaller around 700 nm.Both these reflectance curves have more or less the same shape as our reflectance pattern, suggesting that debris in the ice is a possible cause for the darkening.The true colour plot in Fig. 2a confirms this hypothesis.The brownish black appearance of the dark region is typical for debris.
There are two possibilities how this debris could have reached the dark region.First, it could be wind-blown material from the tundra area.However, this seems unlikely, because wind conditions vary from year to year, in contrast to the dark region.Moreover, the dark region becomes visible at some distance from the margin, whereas windblown dust would settle closer to the margin.Although an counterargument is that due to lower elevations and steeper slopes, dust is washed away more efficiently nearer to the margin by melt water runoff, this does not explain why the dark region is limited on its eastside on the same position every year.In addition, the slope and elevations seem to change only gradually in this area (Fig. 7), whereas the transition between the dark region and the brighter ice is quite abrupt.Therefore, it is unlikely that wind-blown material causes the dark region.
The second possible cause for the debris in the dark region is old dust which settles on the higher parts of the ice sheet, travels through the ice sheet and crops out again in the ablation zone.During colder periods, more dust is deposited on the Greenland ice sheet (e.g.Ruth et al., 2003).Due to the typical flow pattern in an ice sheet, this dust will be transported to the melting zone.The ice surface in the ablation zone can be seen as a horizontal representation of a vertical ice core (Reeh et al., 1987).The closer to the margin, the older the ice is.The ice in the dark region may thus originate from a time period when more dust was deposited on the ice sheet.This would explain the appearance of an abrupt transition and the discontinuity in dust content, which then represents an isochrone.Bøggild et al. (1996) found that in Kronprins Christian Land in north-eastern Greenland, ice of Wisconsin origin is significantly darker than ice of Holocene origin, due to the increased dust content of the ice.Reeh et al. (2002) and Petrenko et al. (2005) also found bands of darker and dustier ice near the western margin of the Greenland ice sheet, which originates from a colder glacial period.However, they found the transition between ice from the last glacial maximum and younger ice at some hundreds of metres from the margin, whereas our dark region is much wider and lies in the order of tens of kilometres away from the margin.Hence, our dark region probably originates from a dustier colder period within the Holocene.
For a more detailed look on the dark region, we have also investigated ASTER satellite images, because they have a much higher spatial resolution.Figure 8 shows a part of the dark region in wavelength Band 2 (630-690 nm) from an ASTER image acquired Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion on 2 August 2004.A part of the MODIS image (Band 1, 620-670 nm) taken on the same day is also shown.The ASTER image clearly shows that the dark region contains a wavy pattern.Such patterns are typical for the outcropping of tilted stratified ice layers.However, this pattern is less clear in the brighter ice.Therefore, the surface morphology of the dark region supports the assumption that ice containing more dust from colder periods surfaces in this region, causing the dark appearance of this region.

Conclusions
Time series reveal that the dark region in the western ablation zone of the Greenland ice sheet appears every year during the summer season and can always be found on the same location.For this reason, accumulation of melt water seems unlikely to be the only cause for the darkening.Spectral information shows that the darkening of the surface is strongest in the visible part of the solar spectrum, when it is compared to the surrounding brighter ice.The differences become smaller with increasing wavelength, till no differences remain in the infrared zone.We conclude that the dark region has a spectral characteristic that is significantly different from that of the brighter ice closer to the ice-sheet margin, and that is typical for ice containing dust.
It is unlikely that wind-blown material is the origin of this dust, because the dark region appears on the same location every year, at some distance from the margin.The assumption that the dust is washed away closer to the margin due to more melt seems also unlikely, because of the sharp transition between the bright ice and the lower edge of the dark region.A detailed ASTER image shows wavy patterns in the ice of the dark region.These patterns are typical for outcropping tilted layers of ice.Therefore, we conclude that outmelting ice layers that contain a relatively large amount of dust causes the darkening.The dust is released during the melting process and accumulates at the surface, and this will increase the melting of the ice and subsequently the amount of external forcing.For this reason, it is important to understand more about the origin of the dust.When we know where the dust originates from, we may predict how it will behave in the future and how it will affect the mass balance of this part of the Greenland ice sheet.Therefore, further analysis of the dust in the dark region is required.Introduction

Conclusions References
Tables Figures

Back Close
Full meltwater.Varying amounts of dust adds to the memory of the ice sheet, implying that outcropping of this dust can enhance the melting of the Greenland ice sheet without

Fig. 1 .
Fig. 1.MODIS image from 9 August 2007, band 2 (841-876 nm), with the appearance of the dark region indicated with a blue contour.S is Sukkertoppen Iskappe and J is Jakobshavns Isbrae.The red contour indicates the reference area as described in Sect. 4. (Latitudes and longitudes along the axes).
shows an image of radiances measured Introduction