Jökull - 01.01.2019, Qupperneq 5

Guðmundsson et al.
also obtained from the ArcticDEM (Porter and others,
2018). The georectification was carried out by iden-
tifying 30–50 common control points in the images
or maps and the lidar DEMs for each glacier and us-
ing ArcGIS tools for reprojection and warping. The
AMS maps are in general fairly accurate in compari-
son with the lidar DEMs, with an estimated horizontal
accuracy of ∼5 m after correction with respect to the
lidar DEMs. The DGS maps often needed horizontal
corrections by tens or even a few hundreds of metres
but are estimated to be accurate within ∼20 m after
the correction in the glacier forelands where the lakes
studied here are located. This accuracy estimate is
intended to represent the overall location and size of
substantial features such as terminus lakes. The un-
certainty of smaller undulations in the shape of the
shoreline around the lakes or similar-size features is
larger. The shapes of lakes determined from the DGS
maps are in many cases corrected based on the lidar
DEMs. Then, an estimate of the water level in the
lake at the time of the DGS survey is used to derive
a shoreline consistent with the DEM elevation around
the lake (in case one could assume the surface eleva-
tion in the proglacial area had not changed since the
DGS survey).
The number of available maps and images of the
foreland varied between the glaciers, and the temporal
resolution of the history for each glacier lake there-
fore varies. Since satellite images became available
in the 1970s, much more information about variations
in lake extent can be found, and in later years, the
extent of lakes can in most cases be estimated every
year, and even several times a year in some cases. The
most recent lake extent was in all cases determined
from Landsat 7–8 imagery from 2018. The ice mar-
gin and the boundary between the lakes and surround-
ing ice-free terrain is hard to distinguish in places in
some of the satellite images, in particular in the Land-
sat 1–3 and 4–5 imagery with 60 and 30 m pixel res-
olution, respectively. This is less of a problem in the
Landsat 7–8 imagery with its 30 m (spectral) and 15 m
(panchromatic) pixel resolution.
The areas of the lakes at different times are
determined from the digitized shorelines. Written
accounts, photographs and variations in termi-
nus position with time from the database of
the Iceland Glaciological Society (available at
“spordakost.jorfi.is”; Sigurðsson, 1998a) provided ad-
ditional information that is used to bracket the time of
the formation of the lakes in some cases and for aid-
ing the interpretation of the satellite imagery. They
are never used to derive quantitative estimates of lake
area.
Information about the volume of several lakes was
derived from a DEM of the bed below Vatnajökull
that has been produced by radio-echo sounding since
the 1990s (Björnsson and Pálsson, 2008; Björnsson,
2009a, 2017; Magnússon and others, 2007, 2012) and
from point measurements of water depth for three
lakes where such measurements are available. The
bed map is derived from radio-echo sounding profiles
of glacier thickness with spacing on the order of 1 km
(Björnsson, 2009a) or scattered point measurements
(Magnússon and others, 2007, 2012) so that there is
considerable uncertainty in the short-scale details in
bed geometry near the glacier termini. The derived
lake volumes may, therefore, be uncertain by several
tens of per cent but should give a good estimate of the
overall magnitude of the water volume stored in the
lakes. The bed map also provides important informa-
tion about the subglacial landscape upstream of the
current terminus and therefore about the future geom-
etry of larger lakes that will form if the glaciers con-
tinue to retreat as a consequence of the expected future
warming of the climate. The lakes for which point
water-depth measurements are available are Breiðár-
lón (ca. 50 points), Fjallsárlón (ca. 50 points) and the
northern part of Svínafellslón (ca. 20 points). The
measurements were carried out with a GPS and a rope
with a 5 kg weight on the end of the line. The line was
lowered to the lake floor and the depth measured with
a line-depth counter (Sæmundsson and Margeirsson,
2016; Sæmundsson and others, 2018).
RESULTS
Terminus lakes by S-Vatnajökull have grown rapidly
since the end of the 20th century, reaching a combined
area of ∼60 km2 in 2018 (Figure 2). The total area of
terminus lakes from Skeiðarárjökull to Hoffellsjökull
increased at an average rate of ∼1.7 km2 a−1 in the
4 JÖKULL No. 69, 2019