Harmen Bijwaard and Wim Spakman
Vening Meinesz School of Geodynamics, Utrecht University,
The Netherlands
E-mail: bijwaard@geo.uu.nl
In the last few years results obtained by global travel time tomography
have been greatly improved by refinement of the parameterization scales
toward those often employed in regional tomography studies (Zhou, JGR,
1996; Van der Hilst et al., Nature, 1997).
Recently, Bijwaard et al. (JGR, 1998) constructed a high-resolution
global tomography model, mapping upper mantle structure with detail
(60-100 km) directly comparable with that obtained in regional
studies. The improved imaging allowed among other things for the
identification of the previously unrecognizable Mongol-Okhotsk
slab (Van der Voo et al.,
Nature, 1999) and a whole mantle signature of the Icelandic plume
(Bijwaard and Spakman, EPSL, 1999).
This improvement was achieved through the exploitation of an accurate
global data set of 7.6 million P and pP phases
(Engdahl et al., BSSA, 1998) and
the implementation
of a model parameterization with cell sizes adapted to the amount
of ray sampling.
However, the in principle nonlinear global tomography
problem has up to
now always been linearized about the ray paths in the 1-D reference
Earth model used. The bending of rays due to lateral heterogeneity
has therefore been ignored, although this may be very important when
imaging such small-scale structure. To investigate the influence of
ray bending on the global tomography done so far, we have extended
the linearized tomographic imaging to nonlinear imaging.
Here, we shall present new results obtained
from the nonlinear approach. For this purpose a new and fast ray
tracing algorithm was developed (Bijwaard and Spakman, submitted to
GJI) to be able to trace the entire global data set, consisting of
7.6 million rays, accurately and within a feasible amount of computation
time through the model obtained from linearized tomography. After
the computation of new travel-time residuals, this new data set has
been inverted in a similar way as for the linearized tomography.
The nonlinear results obtained from this inversion indicate
significant differences between linear and nonlinear inversions
in regions with high-amplitude heterogeneity, i.e. mainly in the
upper mantle. In these regions, we observe
enhanced focussing and higher amplitudes.