Jökull - 01.06.2000, Page 33

Comparison oftomographic crustal models with gravity data
THE HENGILL- GRENS DALUR AND
KRAFLA AREAS
The structures of the Hengill-Grensdalur and Krafia
areas have been described in detail previously (e.g.,
Sæmundsson, 1978; 1982; 1995; Foulger, 1988; Foul-
ger and Toomey, 1989). Both areas contain active
central volcanoes, fissure swarms and the Hengill-
Grensdalur area contains at least one, and possibly
two, extinct central volcanoes. The Krafla central
volcano has a caldera ring fault with a throw of ~ 1 km
(e.g., Björnsson et al. 1977) and is underlain by a
magma chamber with its top at ~3 km (e.g., Einars-
son, 1978). It is well supplied with magma and is
volcanically active at intervals of ~500 years (Björns-
son et al. 1977). The Hengill system is poorly supp-
lied with magma and is only volcanically active on
average about every 1,500 years. No evidence has yet
been found for a contemporary magma chamber in the
area.
Both areas contain high-temperature geothermal
areas that generate continuous, small- magnitude
earthquake activity with stable, known spatial distri-
butions. This circumstance, along with the known
substantial three-dimensionality of structure, made
them ideal targets for LET.
TOMOGRAPHIC MODELS
The experiments
Several hundred small earthquakes were recorded
on temporary seismic networks deployed in the two
areas in 1981 (Hengill-Grensdalur) and 1985 (Krafla).
Excellent datasets were obtained as the stations were
sited with a foreknowledge of the spatial distribution
and seismic rate in each area. LET inversions were
performed using the method of Thurber (1983) which
parametrises the seismic velocity field by values at the
nodes of a three-dimensional grid, in between which
velocity is linearly interpolated.
Hengill-Grensdalur
In the case of the Hengill-Grensdalur study, a
volume 14 x 15 km in area and 5 km in depth was
studied (Figure 2) (Toomey and Foulger, 1989; Foul-
ger and Toomey, 1989). The primary structures ima-
ged were high-velocity bodies beneath the Grensdal-
ur, Ölkelduháls and Húsmúli extinct volcanic sites.
There was a notable lack of low-velocity bodies,
with the exception of a small body, near the lim-
it of resolution, a few cubic km in size beneath
the northern flank of Hengill and tentatively inter-
preted as partial melt. Seismic velocities beneath the
fissure swarm in general were average. Resolution
was good in the centre of the study volumes and deter-
iorated towards the perimeters and the surface. The
high-velocity bodies were interpreted as high-density
gabbroic intrusives.
A second LET study of the area was conducted in
1991 as part of an experiment to study focal mechan-
isms of the earthquakes there (Foulger et al. 1995).
The repeatability of the velocity models between
experiments was about 0.19 km/s and most of the
difference is attributed to the use of different damp-
ing values. This illustrates well the trade-off between
extremity of the model and damping in LET inversi-
ons. The spatial pattern of anomalies was very similar
for both studies.
Krafla
The distribution of earthquakes measured in this stu-
dy was inferior for LET to those of the Hengill-
Grensdalur area. Many of the earthquakes lay in a
narrow zone running NNE-SSW along the centre of
the fissure zone, the locus of dyke intrusions a few ye-
ars earlier, and the hypocentral depths did not exceed
~3 km b.s.l. (Arnott, 1990; Arnott and Foulger, 1994).
A volume 13 x 18 km in area and 4 km in depth
was studied (Figure 3). The well-resolved volume was
limited to the upper 2-3 km b.s.l. of the fissure swarm,
and narrowed with depth. It is instructive to compare
the results to recent shallow refraction profiling in
the Krafla area by Brandsdóttir et al. (1997). In the
LET experiment, the very shallowest layers, between
sea level and the surface (which is at 500 m a.s.l.)
were only sampled in narrow cones beneath individual
stations, defined by the upgoing rays. The stations
were all deployed on outcrops of massive bedrock and
thus the lavas, tuffs and very low-velocity unconsoli-
dated material that fills the caldera were essentially
unsampled. For this reason, it is not valid to extra-
polate the LET model laterally away from the stati-
JÖKULL No. 48 31