Jökull - 01.01.2021, Side 53

Bedrock and tephra layer topography within the Katla caldera
were reviewed. Most observed mismatch was due to
the limitation of the 2D migrated RES-profiles. When
profiles are not driven parallel to the maximum slope
of a steep bed, the traced bed reflection may origi-
nate from cross-track bed reflections up-slope from
the measurement location, hence the obtained bed re-
flections appear higher than the actual bed directly be-
neath the profile (e.g. Lapazaran et al., 2016). This is
clearly demonstrated by comparison between traced
reflections from 2D and 3D migrated RES-profiles
around K6 (Figure 5). At locations where this ex-
plained the mismatch, the profile more closely match-
ing the bed slope direction was kept unchanged while
the data from the crossing profiles was either omitted
or shifted between 5 m and 130 m cross track in up
slope direction to fit the bed elevation of the cross-
ing profile (Figure 3). In the few cases where neither
profile followed the bed slope direction, both profiles
where shifted (< 50 m) cross-track to obtain a match
at the crossing point. At a few locations where the
mismatch could not be related to the shortcoming of
the 2D migrated RES-profiles, the bed-tracing was re-
vised. This usually revealed discrepancy in the inter-
preted bed reflections for the crossing profiles. In such
cases the tracing that seemed more likely to be correct
given the surrounding data was kept. The revision of
the 2D migrated bedrock data reduced the maximum
difference of crossing profiles from 108 m to 8 m.
There are various ways to construct bedrock DEM
from RES-profile data. In recent studies on Ice-
landic glaciers (Magnússon et al., 2012; 2016) the fi-
nal bedrock DEMs have been produced by manually
modifying elevation contours of a preliminary DEM
obtained with kriging interpolation of the RES-data
and then interpolating the final product using the mod-
ified contours as input for another kriging interpola-
tion. This has been done to reduce artefacts of krig-
ing interpolation from discrete RES-profiles as well
as artefacts caused by the shortcoming of the 2D mi-
grated RES-data described above. Applying this ap-
proach for a large and dense data set requires time-
consuming manual work. The artefacts of the kriging
methods are also substantially less prominent than in
the case of more discrete RES-profiles, particularly
after taking care of mismatch at crossing profiles as
explained above. Alternatively, more sophisticated in-
terpolation schemes, constrained by physical models
and other input data such as glacier surface topogra-
phy and velocity observations (e.g. Morlighem et al.,
2011; Fürst et al., 2017), could be adopted. However,
for a large portion of the study area these schemes
would fail without substantial improvements, due to
strong basal melting beneath ice cauldrons, which
strongly affects the surface topography and motion.
Due to the complications of the above interpola-
tion methods it was decided follow a relatively sim-
ple approach in the creation of the final bedrock
DEM (with 20×20 m cell size). Kriging interpolation
was applied in Surfer 13 (©Golden Software, LLC),
with input data (coordinate list of easting, northing,
bedrock elevation) consisting of the filtered and re-
vised bedrock traces from the 2D migration, bedrock
traces from the 3D migration, elevation of nunataks at
their edges, elevation of the previous bedrock DEM
(Figure 2b) at the edges of the study area and a few
manually created elevation contours, drawn to obtain
realistic landforms where RES-data is lacking (Figure
3). The bedrock DEM obtained with this interpola-
tion was then mosaicked with the previously existing
bedrock DEM outside our survey area and lidar DEM
(Jóhannesson et al., 2013, subsampled to 20× 20m
cell size) of nunataks, resulting in the final product
(Figure 6). The ice thickness map (Figure 7a) is calcu-
lated as the difference between a glacier surface DEM,
obtained from Pléiades images in 28 September 2019,
and the presented bedrock DEM.
Extraction of tephra layer data
The 2D migrated data from May 2016 and February
2017 along with 3D migrated data from 2017, (around
K6, K7, K10, K11 and K16) and 2018 (K1 and K2)
was used to map the ice thickness above the 1918
tephra layer. To compensate for the time difference
between observations, data from 2017 and 2018 were
shifted upwards by 3 and 6 m, respectively, to rep-
resent the year 2016. This is based on the crude as-
sumption that the layer depth increases linearly with
time; in 2016, almost a century after the eruption, the
tephra layer was on average at ∼300 m depth. Given
the variable depth of the layer and that the vertical mo-
tion is expected to decrease with depth (see discussion
JÖKULL No. 71, 2021 51