Jökull - 01.06.2000, Qupperneq 47

Comparison of tomographic crustal models with gravity data
or in error sources 4-6 listed above, which were not
quantified for inclusion in the error budget, but which
could account for several tenths of a mGal locally. An
error in detrending the observed gravity field would
produce a systematic effect and cannot explain the
isolated misfits. The two areas of highest residual cor-
respond to the topographic highs of Skarðsmýrarfjall
and Hrómundartindur which rise to 600 m and 520
m respectively. An error in the Bouguer density of
about 900 kg/m3 would be required to explain all of
the residuals - an unrealistically high value.
The most likely explanation for the misfits are
variations in the velocity : density relationship. Scatter
in the data used to derive Equation (1) shows that
for a given seismic velocity, the density may vary by
up to ±100 kg/m3 from that predicted. The Reyðar-
fjörður drillhole is sited in Tertiary rocks, whereas
the Hengill-Grensdalur and Krafla areas are in the
neovolcanic zone, and the rock types are thus rather
different. In addition, the velocity : density relations-
hip is constrained by data only down to velocities of
3.8 km/s. Densities corresponding to velocities lower
than this were determined by downward extrapolati-
on and are probably associated with larger errors.
Densities of 100 kg/m3 lower than those predicted
in the material between sea level and 1 km depth
could explain the zone of residual low gravity, as
could densities 50 kg/m3 lower, extending down to
3 km depth. The zone of low gravity residual cor-
responds to the zone of greatest recent volcanism and
much of the present geothermal area, and relatively
low densities in this zone are thus to be expected
from high- porosity volcanic rocks and geothermally-
altered material.
Poor resolution in the LET model in the form
of a systematic underestimate of the amplitude and
extent of low-velocity volumes is also a possible
explanation. LET may be poorer at imaging low-
velocity volumes than high-velocity volumes because
rays follow minimum time-paths which by-pass low-
velocity material.
Krafla
Comparison of the LET model with the gravity field
was poorer for the Krafla area, probably because
the observed gravity field is dominated by anomalies
that originate from density variations at very shallow
depth where LET is insensitive. Small, shallow anom-
alies such as these will not be detected by the
LET, which will return a smoothed, average velocity
field for volumes of low resolution. Other methods,
e.g., shallow refraction profiling with closely-spac-
ed seismic recorders, are more suited than LET
to detecting small-scale, shallow anomalies. Furt-
hermore, only the seismic structure inside the well-
resolved volume was modelled, and thus unresolved,
neighbouring bodies might contribute to the gravity
field. It is interesting to note that the primary feature
of both the gravity field and the LET model are high
density/velocity anomalies beneath the caldera ring
fault.
The average elevation of the Krafla area is ~50()
m above sea level, and thus the Bouguer slab is about
500 m thick. The Bouguer correction assumes this
slab to have a uniform density. However, rock types
are in reality highly diverse as a result of intensive
volcanic activity and this may result in errors in the
Bouguer anomaly of up to ~ 1 mGal. These errors will
result in errors in the calculated density variations for
the material below sea level. In circumstances such as
these, gravity data are clearly a feeble test for LET
models, but they can contribute to revealing structure
where LET is insensitive, i.e. at shallow depth.
Nevertheless, it is clear that there are substantial
lateral variations in density in the Krafla area. The
exact shapes and depth extents of the bodies ima-
ged are not well resolved because localized averag-
ing of velocity and smearing of spatial characteristics
is a feature of the inversion process. Some compari-
sons with the observed geology may be mentioned,
however. The volume beneath the Leirhnjúkur area in
the centre of the caldera contrasts with the caldera fill
material that surrounds it in being of slightly higher
seismic velocity and density. There is lso evidence
from boreholes, located 2.5-3 km southeast of Leir-
hnjúkur, that the basaltic intrusives that underlie the
caldera fill may be elevated to a depth of 400-500 m
b.s.l. within the caldera, compared with ~1000 m b.s.l.
just south of the caldera (Ármannsson et al. 1987).
This could explain the velocity/density high observed
in the centre of the caldera.
JÖKULL No. 48 45