Jökull - 01.12.1989, Page 9

In order to calculate the thickness of the ice-layer
from the reflection times, the interval velocity of
layer 2 is needed. The RMS velocity for an ima-
gined reflection from the bottom of layer 1 was cal-
culated numerically using the velocity model from
the refraction profile, giving Vms=2390 m/s. The
uphole time (13 ms) was used as the interval time
for layer 1. Using the observed Vrms for the bottom
of layer 2, Dix formula was used to calculate the
interval velocity of layer 2. The results are given in
Table II and Fig. 4.
No analysis was carried out to extract a velocity
for the water layer from the data. The velocity of
fresh water is a slowly varying function of tempera-
ture and the value used here is 1420 m/s, which
corresponds to a temperature of 4 °C (Zemansky and
others, 1966, p. 258). No measurements of the tem-
perature of the subglacial lake have been carried out
nor have there been any theoretical studies of its
temperature profile (Björnsson and Kristmannsdótt-
ir, 1984; Björnsson, 1988). The value of 4 °C is
chosen here as it is the temperature at which a body
of water is most dense and gravitationally stable.
PROCESSING
The processing carried out on the data included
frequency filtering, normal moveout correction, spa-
tial filtering and summing of adjacent traces. The
frequency filtering was carried out in the time
domain and the passband used was 80-150 Hz on the
largest part of the data. On some parts of the data
two passbands (50-125 Hz and 160-200 Hz) were
used. The purpose of the frequency filtering was to
reduce random noise and cut off spurious resonance
frequencies present on some traces due to an
improperly placed or slightly defective geophone.
Normal moveout correction (NMO) was carried
out using a velocity function based on the results of
the velocity analysis. The velocity values used as
input for the function were V=3570 m/s for the ice
and V=1420 m/s for the water.
In addition to frequency filtering, spatial filtering
(mixing) was applied to the seismic sections after
moveout to further reduce the noise content. A
lowpass filter with a cutoff wavelength of 30 m
turned out to give the best results. As the distance
between reflection points along the seismic lines is
10 m (half the geophone spacing) the filter cuts off
only the highest spatial frequencies present. In this
set of data such spatial frequencies are only caused
by random noise and perhaps by diflractions.
Despite the loss in resolution and a slight danger
of aliasing, horizontal summing of adjacent traces
turned out to be advantageous, as the signal to noise
ratio was improved. It was therefore applied to all
the data after frequency filtering, NMO and spatial
filtering.
Parts of the data were contaminated by random
noise caused by a faulty tape recorder. In addition
to the processing steps described above, consider-
able editing of noisy traces had to be carried out. A
large part of this process involved treating the noisy
parts of the data with a gain function which reduced
the amplitude of those parts of the traces which evi-
dently only contained random noise. The original
hard copies from the survey were used to define
these areas on the seismic traces.
RESULTS
The processed seismic lines (Figs. 5a-c) show
clearly the reflections from the ice-water interface
(’a’) and the lake bottom (’b’). Deeper reflections
appear to be present in the northem part of line 2
(’c’-’e’) and in places on lines 1 (’c’) and 3 (’c’-’e’).
Multiples are mostly absent from the data, but a
ghost trails the ice-water reflection by about 25 ms
in places. Some parts of the sections are seriously
downgraded by noise. These parts are 3.5-3.7 km in
line 1 and 2.0-2.1 km, 3.3-3.4 km and 4.2-4.3 km in
Iine 3.
The ice-water interface reflection is everywhere
strong, especially in the interior of the shelf where it
is regular and continuous. Furthermore, the
reflection from the lakefloor can be seen on all three
seismic lines. Only in one place do the two
reflections merge, at 2.4 km in line 3 (Fig. 5c). In
other places the lakefloor reflection disappears near
the edges of the ice shelf. The absence of reflections
near the end of the lines indicates that the slopes
bordering the relatively flat caldera floor are fairly
JÖKULL, No. 39, 1989 7