4.1. Microclimate and Topographic Determinants of the Structure of the Snow Cover on Hornsund Glaciers and Its Melting Characteristics
The year 2014 was the second warmest year, after 2012, in the entire observation history in Hornsund, with a mean annual air temperature of −1.3 °C. The monthly average temperature for all months of the year was significantly higher than multi-annual records. This was particularly clear in winter, from January to March, with the deviation reaching 9.5 °C in February. It was also a dry year, with annual precipitation of 410 mm, which represented 90.6% of the multi-annual 1979–2013. The lowest precipitation was recorded in winter months, which are particularly important for snow cover formation. The monthly precipitation in February (7.6 mm) represented only 26.7% of the mean long-term values for this month [37
]. This resulted in lower than average snow depth, notably in the accumulation zones of investigated glaciers.
The local thermal conditions are strongly determined by the character of the surface. This is particularly evident in the summer, when tundra, which is not insulated by snow, is strongly heated and undergoes intensive evaporation. This causes the number of positive degree days (PDD) to vary significantly from site to site (Table 2
). The accumulated total of PDD ranged from 164.3 (H9) to 585.4 (HOR), and positive mean diurnal air temperature was recorded nearly twice as frequently in Hornsund (47% of the year) as in the accumulation zone of Hansbreen (25% of the year). On account of continuing low air temperature in April caused by northerly and north-easterly cyclonic and anticyclonic situations, April was the only month without a PDD at any of the measuring stations.
The melting period in the glacierized area started relatively late, i.e., only towards the end of the second half of May, and lasted until the beginning of October. An unusual meteorological situation was observed in the summer months, especially in July (on 16 days), when strong thermal inversions occurred, with the mean diurnal air temperature in the accumulation zone 3.6 °C higher than the temperature in the ablation zone (Figure 2
In recent years, warm spells during winter months have been occurring more frequently, accompanied by liquid precipitation, so-called rain-on-snow
]. They are an important indicator of on-going climate change in the polar regions [38
]. Occasionally, they even occur in highly elevated glacierized zones, and producing very compact layers in the snowpack—Melt-Freeze crusts (MFcr) and Ice Formations (IF), which slows down the meltwater percolation processes. This was the case on Hansbreen, where the share of these layers in the total snow depth exceeded 27% in the ablation zone, 20% at the equilibrium-line, and 12% in the accumulation zone (Table 3
). The mean snow density was similar across the sites and amounted to c.
400 kg m−3
. A noteworthy finding was the very low snow depth in the accumulation zones of the glaciers, on which the physical characteristics of the snow cover were measured in more detail.
The measurements taken at the mass-balance stakes in 2014 show that the melting vertical gradient along the Hansbreen axis depends on the elevation, reaching c.
−0.34 m w.e. 100 m−1
. Figure 3
presents mean values of the surface melt rate per day, calculated for periods between observations. The melt rate at the lowermost ablation zone was three times higher than in the upper areas of the glacier, resulting in the lowering of the glacier by c.
2.44 m at the terminus. The highest melt rate was recorded in July, when most of the ablation zone lost its snow cover, which protects the glacier surface against increased absorption of shortwave solar radiation, due to higher albedo. In July and August, the snow cover melt rate in the uppermost part of Hansbreen (500 m a.s.l.) was higher than in the 400 m a.s.l. zone, which can be attributed to frequent thermal inversions and wind activity. The measuring point is located at the ice divide, whereby the air masses moving southwards cause snow transfer and reduce the snow cover. The summer balance 2014 at the individual sites on Hansbreen ranged between −0.88 m w.e. (400 m a.s.l.) and −2.72 m w.e. (50 m a.s.l.), amounting to −1.21 m. w.e. for the glacier as a whole, while the Equilibrium-Line Altitude (ELA) ran at 342 m. a.s.l. For the other glaciers, it ranged between 231 m a.s.l. (Mendelejevbreen) and 400 m a.s.l. (Paierlbreen), with an average of 345 m a.s.l. for Hornsund region.
4.2. Snow Cover Melting Characteristics of Hornsund Glaciers Based on Satellite Data
According to the field measurements and meteorological data, the accumulation period lasted until mid-May (see: Section 4.1
). The first satellite image used in the analysis was from the end of May, when the surface melting processes had begun. At that time, the contribution of snow on the overall glacier surface reached 98%. The areas assigned to the ice
class (2%) mainly include the highly crevassed termini (Figure 4
A noticeable reduction in the Snow-Covered Area (SCA) was seen at end of June, when the SCA was changing at a rate of −0.86% day−1
. Consequently, the ice-covered area was gradually expanding, and by the end of August, the percentage of the ice
class was 11% higher than that of the snow
class (Figure 5
). The total SCA reduction during the studied period equalled −0.56% day−1
. When comparing the individual glaciers, the reduction of SCA varied considerably, ranging from −18% (Mühlbacherbreen) to −74% (Körberbreen) with an average of −52%. Mühlbacherbreen was the only glacier with insignificant SCA reduction. Previous papers focusing on snow distribution on Svalbard glaciers highlighted that interior areas are characterized by the local continental microclimate, where lower air temperature persists throughout the year [30
]. This dependency is visible both on the northern and southern coasts of the Hornsund Fjord (Figure 6
). The melting of the snow cover on Hornsund glaciers continued for several more weeks. However, the lack of later satellite data made further analysis impossible. The latest satellite image that was available, dated 21 September, had already been taken after the first snowfall episode.
From the beginning of August, a superimposed ice zone can be distinguished on the surface of some glaciers. During the classification, it was usually assigned to the accumulation zone (the snow class), with certain exceptions. However, it must be pointed out that separating superimposed ice exclusively by means of indirect methods is a highly subjective operation, requiring great interpretative experience and suitable in situ verification.
The setting that significantly improved the separation of snow from ice was the rejection of shortwave infrared bands (SWIR, band 6 and 7), even though they are widely used for the separation of cloud and snow [9
]. In addition, band 9 (Cirrus), which is mainly used for cloud detection [34
], was also disregarded in the classification since it generated an increased number of artefacts in areas characterised by low amount of clouds. However, rejecting the SWIR bands had a negative effect on automatic detection of the debris
class—fortunately, the class was relatively infrequent, and the number of pixels which represented it was strictly controlled by decreasing the bias for the class and its threshold.
All the material available was characterised by a low value of sun elevation (Table 4
), which highlighted the local topographic features of the glaciers, causing significant differences in the lighting of the surface and producing a high proportion of shadowed areas, reaching 12.5% in mid-July. Nevertheless, the classification allowed us to identify the type of the surface within those areas. Although, in general, shadowed snow
was classified correctly, shadowed ice
often included the debris
class, and differentiating them was a major challenge in the classification process.
The earliest discernible loose moraine material on the surface of glaciers (debris) appeared in early June. Throughout the melting period, the contribution of debris class was only several percent, but it contributed to melting in a significant way. The class consists of gravel and rocks that heat easily and have the lowest albedo in the study area. This generates positive feedback which accelerates the melting of the adjacent snow and ice cover. The largest accumulation of debris was observed in early August. However, debris is difficult to classify at times of advanced melting because areas of glacier ice with low image tonality, like wide glacier crevasses, were often included in the class.
During the melting period, temporary supraglacial water bodies may appear on the surface of the glaciers. Most of them are small melt ponds (<0.005 km2), only occasionally reaching larger sizes. The largest supraglacial pond was located on the surface of Paierlbreen (0.21 km2) and was formed after the snow cover started to melt intensively. The lifetime of such melt ponds mainly depends on the timing at which the glacial drainage system opens, followed by feeding with meltwater from the glacier surface. Small ponds were seen from the beginning of June until the end of August. Of all the glaciers, the greatest number of such water bodies was observed on Storbreen, where the unique relief of the glacier, i.e., a vast flat area with a small inclination, a low crevasse level, and numerous depressions, is favourable to their seasonal formation. The extra water class was added to the images dated 22 June and 15 July, when the supraglacial ponds were distinct and filled with meltwater, even though their share of the overall area was negligible (0.09%).
With a suitable image quality (no clouds, higher pixel intensity) and length of the glacier terminus, Landsat imagery can also be used for observing the evolution of subglacial drainage systems and for identifying the meltwater outflows, where the water, rich in mineral material, leaves the glacial basins. Such points are present in all Hornsund glaciers, most often at the side parts of their termini. Large amounts of suspended sediment are discernible through their much lighter image tone on the surface of the fjord. The earliest signs of the sediment in its initial form were observed at the beginning of June in front of the Mühlbacherbreen, Storbreen, and Hornbreen. As the melting advanced, the drainage channels were unsealed, and the migration of meltwater became even more noticeable. The greatest turbidity of Hornsund waters was observed in mid-July, especially in the bays facing the termini of the largest glaciers (Storbreen and Hornbreen). This could be seen in all images until the end of August. The largest amounts of sediment are transported from the glaciers in Brepollen, which can be explained by the presence of the long capes of Treskelodden and Meranpynten (in the north and south, respectively), which divide the fjord, preventing the Hornsund waters from mixing. Towards the end of September, the bays facing the termini of the glaciers slowly started to freeze over.
Clouds were a natural factor hindering correct identification of classes. In the summer, considerable cloudiness is typical for the Hornsund area [27
]. This is the factor that most significantly affects the quality of satellite data used in remote sensing analyses. For some glaciers, a cloud mask was applied: at the beginning of July (Samarinbreen) and in mid-July (Mendelejevbreen and Kvalfangarbreen). Due to extensive cloudiness, the remote sensing analyses in 2014 had to be discontinued in late August.
Of all the glaciers under study, the greatest problems with accurate classification were posed by those located on the southern side of the Hornsund Fjord. We assume that morphometric differences have an important effect on the quality of classification. The southern glaciers cut strongly into narrow valleys and landforms that are surrounded by higher mountain ranges, which generate more surface shadowing, and produce local cloudiness. In addition, they have a much smaller surface area. All this caused major interpretation difficulties, decreasing the accuracy of the classification results of those glaciers. However, the final results of each Landsat image have a very high kappa coefficient [44
] and overall accuracy of over 95% [35
], see: Table 4