The fast flow of SKK, one of the main ice streams draining 6.5 % of the GrIS e.g. has been thoroughly studied during the last decades. The ice flows through a bedrock through reaching a depth below sea level of 1500 m and an ice thickness of 2500 m . The surface velocity reaches 8 km a⁻¹ in the trough and up to 14 km a⁻¹ at the terminus .

Deep boreholes were drilled through the ice stream to elucidate the mechanisms of fast flow . At the ice stream margins, temperate ice close to the bed was discovered at sites A, C and D (Fig. ) but not at central site B where drilling stopped far above the bed. From modeled extrapolation of measured temperature profiles , it was concluded that a very thick layer of temperate ice resides within the trough. Together with convergent ice flow the temperate ice contributes considerably to the observed fast flow. However, recent modeling studies demonstrated that the fast flow can also be explained by significant sliding .

Multi-spectral satellite imagery from the Sentinel-2 and Sentinel-3 platforms were retrieved from . We used Level-2A processing, an atmospherically corrected ortho-image representing the bottom of atmosphere (BOA) reflectance. Sentinel-2 provides 12 bands and has a revisit time of 1 to 2 d at SKK. It was used to derive reflectance spectra in the visible (VIS, 4 bands), near infrared (NIR, 6 bands) and shortwave infrared (SWIR, 2 bands) with a spatial resolution of between 10 and 60 m. Sentinel-3 passes every day and the true color images from the Ocean and Land Contour Instrument (OLCI) were used to determine the time of calving. However, its spatial resolution (300 m) was too coarse for the scope of this study.

Blue ice was identified on freshly calved icebergs using true-color Sentinel-2 images. To ensure the most uniform reflectance conditions, icebergs in shadow and those exhibiting specular reflection (very bright icebergs) were excluded. A catalog of 24 icebergs (or iceberg ensembles) featuring blue ice between 2017 and 2022 was created. For comparison, 24 white icebergs were chosen from many that could be found in the fjord. Finally, we also extracted mean reflectance from 4 shallow melt water ponds at the surface of icebergs, and the reflectance of a lake on land with no evidence of significant suspended sediments in the surroundings of the area (50.87° W, 69.09° N) on 17 August 2023.

From the 24 blue icebergs, 9 were tracked and the evolution of the reflectance spectra and blue color index was studied over 13 d. These 9 icebergs were divided into two groups: 6 of them calved between May and mid-August, when solar radiation was at its peak. Three others calved on the same day on 18 September 2021, when solar radiation had already significantly decreased.

The blue color in the visible spectrum was quantified with an index (referenced as blue color index) comparing the magnitude of the blue channel to that of red, green and blue channels (Sentinel-2 bands 4, 3, and 2).

1Bindex=bluered+green+blue=band2band4+band3+band2

To compare the reflectance spectra among each other, and with reference spectra, we quantify the magnitude differences between spectra using different measures: root mean square difference (RMSD), linear relationships using the Pearson correlation coefficient, and monotonic relationships using the Spearman's rank correlation.

Field measurements during 16 d were conducted on the south shore of SKK near the calving front (green ellipse on Fig. ) between 3 and 18 July 2023. Numerous RGB pictures of freshly calved icebergs were taken from the camp and its vicinity, using digital cameras (Samsung NX300, Canon EOS 7D, Nikon D5200). Calving sequences were directly observed, and images from a stationary time-lapse sequence with 10 s interval were used (Canon EOS 70D).

The dimensions of freshly calved icebergs were measured by triangulation with a theodolite from two tripods. This method also allowed us to locate the various ice layers, especially blue ice and sediments, with an accuracy of 15 m in the horizontal and 3 m in the vertical . The theodolite (Leica TCR1101) was mounted consecutively on two survey tripods spaced 240 m apart in the vicinity of the camp. The distance to the target icebergs was between 2.5 and 6.0 km. Theodolite measurements were performed within 15 min after the ice mélange stopped moving after a calving event.

Thermal images were acquired to measure the thermal emissions of freshly calved icebergs with a pixel resolution of 5.5 m at 4 km distance. The thermal camera was a FLIR T440bx, with an uncooled infrared detector sensing between 7.5–13 µm, providing pictures of 320×240 pixels. Processing was performed with the open source software IRimage . After alignment of thermal and RGB images, temperature and blue color index were extracted along a profile (presumably vertical before calving) on three icebergs with adjacent portions of blue and white ice.

Based on numerous images taken over the past 20 years and from observations in the field, blue ice is visible during most major calving events and appears to be independent of the position of the iceberg along the calving front, indicating that it exists across the entire width of the ice stream. There are two main ways to see blue ice appear during calving events. First, when an iceberg calves as a vertical slab and rotates backward, it horizontally exposes the formerly vertical cross-section of the ice stream, and blue ice is visible at its foot. Secondly, other icebergs calve but do not rotate immediately. First, they rise above the calving front due to buoyancy. Then, the part emerging from the water, which does not show blue ice, disintegrates. Next, still due to buoyancy, the deeper part of the iceberg rises to the surface and exposes the blue ice. The observations of these calving events allow us to affirm that blue ice was only observed in the bottom half of the glacier thickness.

The bottom structure of two icebergs is shown in Fig. . Iceberg B1 (Fig. , B1) exhibits a blue layer with a thickness of at least 200 m (the upper limit of blue ice cannot be deduced precisely due to white ice dust covering the iceberg). Immediately beneath the blue layer (left on the picture) lies a gray-greenish layer approximately 50 m thick, followed by a lighter layer of 120 m crossed by two thin layers of dark sediment or dust. The bottom consists of a 150 m thick layered structure, with sediment layers intertwined with blue or white ice. Iceberg B2 (Fig. , B2) shows an extended area of blue ice. A folded sequence including green-greyish ice, white ice and sediment layers is clearly observable, with an amplitude of around 100 m.

Surface temperatures from thermal imagery were determined along profiles on three icebergs (Fig. ). Icebergs B3 and B4 are small and stem from the bottom half of the calving front thickness, whereas B5 is a full-depth tabular iceberg (description in Sect. ). As seen by the thermal sensor, blue ice emits less infrared radiation than white ice. With emissivity set to 0.98 for both ice types, the inferred temperature of blue ice is colder than that of white ice. On iceberg B3, a clear temperature drop of 4 °C occurs between the white and the blue ice. This phenomenon is also visible on iceberg B4, with a temperature drop of 3 °C, but barely visible on iceberg B5, with a decrease of only 0.5 °C. Largest temperature variations are respectively 0.7, 0.3 and 0.1 K 10 m, all at the transition between white and blue ice.

While the reflectance spectra of white icebergs exhibit the classical shape of glacier ice found in the literature , blue icebergs show markedly distinct reflectances. They are more attenuated than the white ice, and also differ markedly from the BIA . In the NIR, blue icebergs show similar reflectances to those of a reflectance spectrum of water-saturated ice from Midtre Lovénbreen (Svalbard), which is certainly temperate . However, the saturated ice reflectance spectra shows a concave-downward shape, contrary to the concave-upward shape of blue icebergs from SKK. Shallow melt water ponds that form on icebergs and blue ice admittedly show similar reflectance values and concave upwards shape of the spectra, but the comparison does not lead further due to the distinct physical processes involved. On shallow melt water ponds, the light reflected on the submerged white ice surface passes through the water, which acts as a filter that blocks red light and transmits blue light, hence the blue color.

In what follows, we suggest reasons to explain why the spectral properties of blue ice differ from those of white ice. Several factors can influence the spectral properties of the ice and reduce its reflectance. Larger grain size or lower air bubble content can significantly reduce reflectance and could make the ice appear blue e.g.. Bi-directional reflectance also has an impact, especially on complex iceberg topographies . It however appears in these studies that the impact of larger grain size, lower porosity or lower viewing angle rather shifts the reflectance downwards, without altering the concave-downward shape of the spectra in the VIS and NIR. Contrary to the intuition that the BIA reflectance would resemble that of blue ice, the BIA reflectance obtained with an on-site spectroradiometer is surprisingly more similar to SKK white ice than to its blue ice. Because the color of BIA is attributed to large grain size and low air bubble content (both achieved through the overlying ice pressure over thousands of years, similarly to SKK white ice), this suggests that a different mechanism is behind the intriguing spectral properties of SKK blue ice.

Thermal emission temperature measurements contradict hypothesis 1 (high interstitial water content). Ice with a high interstitial water content must be at the pressure melting temperature, but cold blue ice cannot contain interstitial water. If the blue ice is indeed temperate, the thermal emissivity used in the processing the thermal infrared raw data has to be adjusted. The literature value of 0.98 was used to infer the temperature from raw sensor data, but it might be considerably lower than under laboratory conditions. For example, icebergs generally exposed flat surfaces and were seen with large viewing angles as a result of the small height-to-distance ratio. In the thermal infrared the directional ice emissivity drops to 0.82 with a viewing angle of 75° . Except for the viewing angle, various types of ice, such as sedimentary ice with variable air bubble content, grain size, clathrates or containing liquid water may influence the emissivity. However, no literature values on these characteristics in the thermal infrared could be found.

Emissivity ϵ is a crucial parameter in thermal imaging as it is used to account for the radiation from the surroundings reflected at the surface of the measured object (called reflectivity r) r=1−ϵ. The lower the emissivity, the higher the reflectivity. In our case, under clear sky conditions, decreasing the emissivity increases the computed object temperature. To investigate by how much this effect influences the derived temperature, the equation converting raw sensor values into temperature was inverted using a fixed emissivity.

We initially computed the ice temperature by processing the raw sensor values assuming an emissivity of 0.98. In the following, we assume that the blue ice is temperate (or at least, at the temperature of the adjacent white ice). Inverting the algorithm to compute the ice emissivity from the raw sensor values, the temperature was set to 0 °C. On iceberg B3, where the sensor values between the blue and white ice show the largest variation, the blue ice emissivity would drop to 0.70 if it were at the melting temperature (or to 0.80 for the temperature of the adjacent white ice). Given these calculations, it is possible that blue ice is temperate if it has a very low emissivity that makes it appear colder. Under these assumptions, the increase of ice temperature within a few hours (observed with fixed emissivity) can be translated into an increase of the emissivity (if the ice remains temperate), for example due to the drainage of the water from the temperate ice or the disappearance of clathrates.

In this study the properties of the blue ice observed on the freshly calved icebergs at SKK were investigated. These icebergs are particularly interesting because they offer a direct glimpse at the ice that is located deep within the ice stream and would otherwise only be accessible in boreholes. The blue ice layer is 200–250 m thick which amounts to one quarter of the total thickness of the ice stream.

The properties of the blue color and its time evolution after calving could not be unequivocally determined. From multi-spectral satellite imaging, attenuated reflectances are different from large-grained blue ice from Antarctica with low air bubble content and hint at a different cause. Clathrates or interstitial liquid water between ice grains are two possible explanations requiring further evidence. In thermal camera measurements the blue ice appears colder than white ice and suggests that it corresponds to the core of cold ice situated around mid-depth in the ice column, which contradicts the interstitial liquid water hypothesis. According to , the vertical position of the blue ice is not necessarily at the bottom of the ice stream, as the icebergs show a complicated layering structure of sediment-rich ice, white, green and blue ice. If blue ice containing clathrates lies above white ice, why would the ice below not contain clathrates despite favorable pressure and temperature conditions? If the white ice below were temperate, interstitial water within temperate ice should rather promote clathrates with an enhanced air diffusion to nucleation sites .

Conclusively answering the crucial question whether blue ice is temperate is not possible based on our measurements. If it were at the pressure melting temperature an implausibly low thermal emissivity of the blue ice would be required. On the other hand, such a thick layer of cold ice located near the bottom of the ice stream is difficult to explain regarding the high ice deformation rates, unless complex flow patterns from convergent ice flows have completely rearranged the original thermal structure.

In addition to the search for temperate ice, freshly calved icebergs reveal numerous structures (layers of different colors, large-scale folds) that are directly linked to complex and variable flow dynamics. The tributaries merging into the main ice stream likely create these complex structures, certainly accentuated by flow variability. These observations may confirm the numerical modeling of the fold formation in a convergent flow with a bottom layer of temperate ice . Understanding these observations hinting to complex flow patterns requires comparison of detailed numerical modeling results with complementary sets of field observations to better constrain the mechanisms of fast ice stream flow.

The limitations of our methods are due to the relatively coarse spectral resolution of Sentinel-2 (10 to 60 m depending on the bands), inaccurate processing parameters (air temperature, emissivity) and altered ice temperature (heat from the sea, air) in thermal imaging. To resolve the contradictions discovered in this study, and to better understand the fast flow mechanisms of SKK in relation to the blue ice layer, drilling a borehole down to the bedrock in the center of the ice stream is certainly the most desirable method. However, to remain consistent with the relatively simple methods used in this study, we recommend the following research. (1) Drill into freshly calved icebergs, measure temperature and take ice samples. Such work would be possible from helicopters, if readily available, and might soon be achievable with sophisticated UAVs (drones). (2) Use hyperspectral satellites such as EnMAP or PRISMA to derive reflectance spectra with high spectral resolution and focus on liquid water absorption bands to infer temperate ice. Alternatively, fly a multi-spectral or hyperspectral UAV to obtain better spatial resolution. The latter method would also allow to study the properties of the thinner green-grayish layers possibly resulting from a high dust content originating from the late Wisconsin period . Linking the properties and their evolution of blue and green ice would provide a better understanding of the underlying processes. (3) Track the streamlines from where blue ice appears on the fjord surface. If only central streamlines, where fjord depth exceeded 1600 m for long distances, show blue ice, but not adjacent streamlines, this would be a strong argument for the clathrates (hypothesis 3).