Jökull - 01.01.2019, Síða 119

Tussetschläger et al.
neath and around several PSPs was observed in the
Kerling, Búrfellsdalur and Sakka areas, but not stud-
ied in detail; therefore it is difficult to reach conclu-
sions regarding an accurate distribution of permafrost
around those snow patches. However, the occurrence
of at least sporadic permafrost in Brimnesdalur, Búr-
fellsdalur and Kerling is considered very likely, due to
the high occurrence of PSPs, observed frozen ground
proximal to the PSPs and their very high likelihood
of occurrence. Damm and Langer (2006) point out,
if the distribution probability of PSPs is low, espe-
cially in lower elevations, then external influences
have to be considered (depressions, avalanche chan-
nels, windward–lee site situations etc.). Results from
Etzelmüller et al. (2007) and Farbrot et al. (2007b)
and Lilleøren et al. (2013) suggest that permafrost is
widespread at sites where the MAAT is below -2 to
-3◦C which is at elevations above 800–900 m a.s.l. in
the Tröllaskagi peninsula. Recent slope movements
in Tröllaskagi also confirm scattered permafrost at
around 800 m a.s.l. (Sæmundsson, 2018; Morino et
al., 2019). Figure 8 shows PSPs plotted on the MAAT
from 1961 to 1990 (Icelandic Met. Office), many
PSPs occur on elevations where discontinuous per-
mafrost is most likely present.
The widespread snow patch distribution above
800 m a.s.l. in the Brimnesdalur, Kerling and Búrfells-
dalur areas is considered to be related with at least
sporadic permafrost. In the lower Almenningar and
Úlfsdalir areas only few snow patches occur for long
periods with lower distributions, but many are below
the MAAT -3◦C isotherm. In the Sakka region at 600–
700 m a.s.l. the snow patches only occur with lower
distributions and therefore, occurrence of permafrost
is not likely. However, at higher elevations some snow
patches have a higher likelihood to occur and per-
mafrost is more likely.
Based on the results of our study, we believe that
PSPs with a high likelihood of occurrence, which oc-
cur over long time periods and are widespread, are
good indicators for local permafrost distributions, es-
pecially used in combination with MAAT. Avalanche
channels should be ignored, because PSPs in those
channels do not represent the lower boundary of per-
mafrost in the research area.
CONCLUSIONS
Optical satellite images, especially Landsat-5/-7/-8
and Sentinel-2, have a great potential in mapping
PSPs due to an increased temporal resolution and spa-
tial resolution. This study presents a PSP detection
methodology, based on optical satellite images and a
digital elevation model, which is used to identify the
distribution and evolution of PSPs in six study areas
on the Tröllaskagi peninsula. The algorithm calcu-
lates the NDSI to derive a binary snow classification
excluding avalanche channels derived from a DEM.
An intercomparison with independent data shows that
snow is classified with confidence. Different periods
of PSP coverage are evaluated to show the evolution
of the snow patch distribution over a longer time pe-
riod. Due to frequent heavy cloud cover, six study
areas were selected to increase the number of suitable
satellite images. Still, in some cases the quality of
the images is not satisfying, due to cloud cover or ac-
quisition timing. Several datasets, i.e. aerial images,
orthophotos and field work photos, are used to evalu-
ate the quality of the classification method. The inter-
comparison is complicated by the fact that data is only
available for some years and when available, often not
acquired on the same day or even month.
Snow cover extent varies between the areas and
time periods due to different local characteristics,
e.g. elevation, aspect or topography (e.g. plateau/open
slopes vs. valley). The greatest extent of the snow
patches occurs during periods of high winter precip-
itation, while minimum extent occurs during periods
of low winter precipitation and relatively high sum-
mer temperature. In Brimnesdalur not only the largest
snow patches are detected but also the highest prob-
abilities of the snow patches occurring are identified.
However, the distribution pattern of mapped PSPs in
all study areas is similar in each time period. Conclu-
sively, the PSPs in each area and each time period are
often in potential permafrost areas (elevations above
800 m a.s.l.) and therefore, we suggest that PSPs can
be used as a local permafrost indicator, if the snow
patches reappear in multiple years (more than 2 years)
and not only a single snow patch is located in the area
of interest but several.
118 JÖKULL No. 69, 2019