Jökull - 01.01.2019, Page 105

Tussetschläger et al.
south. This overall regional pattern is supported by
temperature data from four shallow boreholes located
in the central highlands and the mountains in eastern
Iceland (Farbrot et al., 2007b).
Due to its character as a mostly invisible phe-
nomenon on the land surface, area-wide permafrost
detection and quantification is not well developed
(Klug et al., 2014) and thus most knowledge of the
regional permafrost distribution is based on proxy in-
dicators (e.g. Haeberli, 1975; Haeberli, 1987; Stöt-
ter et al., 2012). Perennial Snow Patches (PSPs) can
be related to local permafrost occurrence (Haeberli,
1975; Haeberli, 1987; Stötter et al., 2012; Furrer
and Fitze, 1970) and are often associated with in-
tact rock glaciers (Haeberli, 1975; Damm and Langer,
2006; Rolshoven, 1982). The relationship between
PSPs and permafrost is twofold: firstly, the relative
high albedo of snow protects the permafrost beneath
from solar radiation, and secondly, the underlying
permafrost prevents the snow patches from the influ-
ence of the ground heat flux (Haeberli, 1975; Stöt-
ter et al., 2012; Rolshoven, 1982; Furrer, 1955). To
prevent the ground from cooling due to decreasing
air temperatures, the snow pack must be between 80
and 100 cm thick (Zhang, 2005). Zhang (2005) de-
scribes the effect of seasonal snow on different per-
mafrost appearance (continuous, discontinuous, and
sporadic) due to its influence on the temperature and
thickness of permafrost as well as on the distribu-
tion. It is also mentioned that permafrost can only
develop under a thin snow pack (less than 40 cm),
which allows the penetration of low temperatures into
the ground. Before considering PSPs as permafrost
indicators, different topographic conditions must be
identified to exclude snow patches which are not re-
lated to permafrost occurrence, e.g. avalanche slopes
and windward-leeward situations (Damm and Langer,
2006).
The detection and classification of PSPs and
changes in their distribution must be based on time
series of images, because one image reflects the snow
patch distribution at only one point in time (Stötter et
al., 2012). Various studies utilize high resolution re-
mote sensing images to detect the distribution of snow
patches using different techniques (e.g. Damm and
Langer, 2006; Kääb, 2008). The techniques are of-
ten used for quantifying temporal variations in snow
covered areas in polar environments (Eveland et al.,
2013; Kargel et al., 2005; König and Sturm, 1998;
Macander et al., 2015; Sturm and Benson, 2004) or
to analyze the influence of snow cover on the bio-
sphere (Green and Pickering, 2009; Rosvold, 2016).
In contrast, only little is known about the temporal
variation of PSPs in Iceland. Lewis (1939) analyzed
snow patches and their influence on periglacial phe-
nomena in Iceland and divided the PSPs into classes.
The interaction between the (winter) snow cover and
ground surface temperatures as well as permafrost
distribution in Iceland is discussed in (Etzelmüller et
al., 2007; Farbrot et al., 2007a; 2007b).
The aim of this study is the detection of peren-
nial snow fields based on aerial and satellite classifi-
cation and an investigation of the effects of climatic
and topographic factors on their occurrence and evo-
lution, in an effort to provide insight into the dis-
tribution of permafrost in Iceland. We use freely
available optical satellite data to identify and classify
the snow patches by calculating a Normalized Differ-
ence Snow Index (NDSI), including different thresh-
old values and using a high resolution digital eleva-
tion model. Avalanche-induced snow patches are ex-
cluded by using a simple semi-automatic model. Dif-
ferent time periods were used to show the develop-
ment of the snow patches in six different study ar-
eas. The mapped snow patches are compared to aerial
images, orthophotos and field photos to evaluate the
PSPs classification.
THE STUDY SITE
The study is conducted on the Tröllaskagi Peninsula
(65◦50’N, 18◦45’W) in the north of Iceland (Fig-
ure 1). The peninsula is located between Eyjafjörður
in the east and Skagafjörður in the west. The high-
est summits of the peninsula are between 1400 m and
1550 m a.s.l. The main bedrock of the peninsula is
basalt of Late Miocene age of 5.3–11 Ma (Hjart-
arson and Sæmundsson, 2014). The present land-
scape of the mountain massif is mostly formed by
glacial and fluvial erosion during Quaternary glacia-
tions (Jóhannesson, 1991). In 2007, 111 named
104 JÖKULL No. 69, 2019