In the field of seismic tomography, a large difference
in resolving power exists between high resolution
regional studies and long wavelength global studies.
Regional studies tend to resolve features on a scale
of 100 km or so (e.g. Spakman et al., 1993), whereas
global studies usually image structure on a scale of
1000 km (e.g. Su et al., 1994). Hence, global studies
have been unable to map important small scale lateral
heterogeneity such as subduction zones, rift zones and
hotspots.
To improve on the global modeling done so far, we
concentrate on two aspects of the tomographic
modeling: the data set and the model parameterization.
The applied data have been derived from relocated
hypocenters of the ISC and NEIC data sets (Engdahl
et al., in press). This resulted in a data set of
nearly 11 million arrivals. We perform simultaneous
inversions for slownesses, event locations and station
corrections using depth phases to increase global
coverage and constrain the hypocenters. The earth is
parameterized with a 'flexible cell model' in which the
chosen cell size depends on the hitcount, thus
allowing for more detailed solutions in high hitcount
regions.
Inversions are performed with single as well as
composite rays (Spakman and Nolet, 1988) to test the
validity of the 'average' ray approach. Several
synthetic tests are used to visualize the model's
resolving power. In general, small scale structure is
resolved well in densely sampled regions, notably near
Tonga, Japan, Indonesia and the Mediterranean. In these
areas a qualitative comparison shows good agreement
with regional models (e.g. Widiyantoro and Van der
Hilst, 1996 and Spakman et al., 1993). In the low
hitcount parts of the model the solution resembles
existing global models such as the IASP3D model
(Vasco et al., 1994).