Jökull - 01.01.2019, Page 115

Tussetschläger et al.
melted by September, the snow patch along the ridge
is 80% smaller. This makes the aerial photo less suit-
able for the validation, but still useful to visualize the
general development of the snow pattern.
Additionally, we resampled the aerial image to the
same resolution as the satellite image (30 m) and com-
pared the results again. In 1988, 11% less snow is
identified in the aerial image and the snow patch close
to the ridge is 13% larger in the aerial image compared
to the satellite image. The snow patch in the north east
is smaller in the satellite image as well (by 19%). In
1994, the total amount of snow is 65% smaller in the
satellite image compared to the aerial image and 72%
of the snow patch on the ridge has melted between the
two dates.
Four Brimnesdalur aerial images from the years
1988 and 1994 are available and are used to compare
snow patches to the optical satellite image from the
same years (Figure A3). The 1988 aerial photo only
covers the lower part of the valley, but acquired on the
same day as the satellite image. The snow identifica-
tion in the aerial image is good because there is al-
most no shadow in the image. Different snow patches
in the area are selected for comparison. A group of
snow patches on the eastern side of the area of inter-
est is 11% larger in the aerial image compared to the
satellite image and another snow patch close to the
debris-covered glacier is about 20% smaller. For the
1994 pair the time difference between the aerial im-
age and the satellite image is about a month, similar
to Kerling many snow patches have melted during this
time. Overall the PSPs are 23% smaller in the satel-
lite image compared to the aerial image. The snow
patches along the mountain ridge and the glacier are
identified in both images. However, during this month
the snow patch along the mountain ridge reduced by
13%, while two other selected snow patches in the
aerial image have completely melted. The aerial im-
age from 1994 is not suitable for comparing the results
from the satellite image, but it shows the snow pattern
during ablation season.
There is one aerial image from Úlfsdalir from the
year 1988 and in Búrfellsdalur and in Sakka there
is also one aerial image available from 1994 (Figure
A3). In Úlfsdalir the acquisition date is the same but
in the two other areas the time gap is more than a
month. Three snow patches are selected in Úlfsdalir
and compared to the snow patches identified in the
satellite image. All three snow patches are smaller
(between 10 to 20%) in the aerial image.
For Kerling and Búrfellsdalur orthophotos are
available and these are compared to the satellite im-
age based PSPs classification of the same year (2000)
in Figure 6. The resolution of the orthophotos is
Figure 6. Comparison of (a) orthophoto (Loftmyndir ehf), (b) satellite image from the 20.08.2000 (Landsat-7)
and (c) snow patch classification based on satellite image in the Kerling area (year 2000). Image (b) is the only
available satellite image of this area from the year 2000 which demonstrate well limited supply of quality data,
but at same time that this image is sufficient and the importance of using the data available at each time. –
Sjálfvirk greining fanna af gervitunglamyndum borin saman við háupplausnar flugmynd. Tunglmyndin er sú
eina af þessu svæði frá árinu 2000 en dugir til að greina fannir með okkar aðferðum.
114 JÖKULL No. 69, 2019