Jökull - 01.06.2000, Qupperneq 31

Comparison of local earthquake tomographic crustal models
with gravity data for the Hengill-Grensdalur and Krafla areas
Gillian R. Foulger and Paul R. Field
Department of Geological Sciences, University ofDurham, Science Labs, South Rd., Durham, DHl 3LE, U.K.
Abstract — A combination ofgravity data and three-dimensional seismic velocity fields calculated using local
earthquake tomography (LET) can provide more information on subsurface structure than either method alone.
In addition, gravity data provide one ofthefew methods ofchecking LET results. LET velocity models from the
Hengill-Grensdalur and Krafla areas are used to simulate density structures using a linear seismic-velocity :
density relationship from measurements made in the Reyðarfjörður drillhole. The results are used to simulate
the Bouguer anomaly fields, which are then compared with the observedfields. A fairly good correspondence is
obtainedfor the Hengill-Grensdalur area. Overestimation ofthe gravity in a zone corresponding to the area of
most recent volcanism is probably a result ofa different velocity : density relationship for the rocks there. In the
case ofthe Krafla area, the observed gravityfield was dominated by anomalies due to relatively small, shallow
bodies, to which LET is insensitive. Thus, tlie gravity data are poorly suited to test the LET results there, but the
two combined yield more complete information on subsurface structure than either method alone.
INTRODU CTION
Because it utilises naturally-occurring earthquakes,
local earthquake tomography (LET) can determine the
three-dimensional structure of the crust in seismo-
genic areas down to several kilometres depth. In or-
der to probe so deep in an active seismic experi-
ment, a proíile several tens of kilometres long and
large explosions are necessary. A useful by-product
of LET where it involves simultaneous inversion for
structure and earthquake locations is more accura-
te hypocentres than can be obtained using a one-
dimensional crustal model.
Like other geophysical methods, LET has limitati-
ons. The volume to be studied is typically subdivi-
ded into blocks. Within each block seismic velocity
may be modelled as a constant or a gradient (e.g.,
Thurber, 1983). The spacings of the earthquakes and
seismometer stations govern the dimensional limitati-
ons of the blocks, which may be 1-2 km on a side
in typical experiments. LET returns a smooth model,
and thus sharp velocity discontinuities such as layer
boundaries are imaged as velocity gradations on the
scale of at least one block. Another limitation is
that the earthquakes may not be uniformly distribu-
ted throughout the study volume and thus resolution
may be patchy.
A fundamental problem is that of assessing the
resolution of the results, i.e., how close the model
calculated is to the truth. Difficulties inherent in the
method include trade-off between earthquake locati-
ons and structure, and the selection of an appropria-
te damping value for damped least-squares inversi-
on. Different damping values give different results
and, in the presence of other errors in the data, that
which gives the lowest final RMS seismic wave tra-
vel time may not be the best result. A robust approach
is to compare the results with other information from
geology and geophysics, e.g., gravity data.
Gravity is sensitive to variations in rock density,
which is broadly related to seismic velocity, and in
this respect suitable to test the LET models. However,
although seismic velocity has been found empirically
to be related to rock density approximately linear-
ly, the data scatter is generally large, rendering the
JÖKULL No. 48 29