HR: 16:15h
AN: S32E-02 [PDF]
TI: 3D mantle P-velocity model based on uniform
tetrahedronization of the Earth and the natural-neighbour
interpolation method
AU: * Pan, J
EM: pan@seismology.harvard.edu
AF: Harvard University, 20 Oxford St.,
Earth & Planetary Sciences, Cambridge, MA 02138 United States
AU: Dziewonski, A M
EM: dziewons@seismology.harvard.edu
AF: Harvard University, 20 Oxford St.,
Earth & Planetary Sciences, Cambridge, MA 02138 United States
AB:
Traditional parameterization of three dimensional (3D) Earth models consists of radial and horizontal basis functions with a
constant lateral grid size, independent of depth. As a
result, the horizontal basis functions become more densely distributed with increasing depth. The nominal resolution thus
becomes finer. For example, in the lower mantle and D'' region, current whole-mantle models appear to be characterized by
particularly short-wavelength structures and high gradients, which might not be real features.
In this study, we use instead a 3D parameterization generated by an ``advancing front method'' (NETGEN), given an initial
triangulation on the surface of the Earth. The result is a number of tetrahedrons of optimal and roughly similar size making
up of the Earth's volume. We then utilize the natural neighbour interpolation technique to set up the whole mantle 3D
inverse problem. The functional value (in our case, perturbation to the P-velocity) at any point within the Earth can be
found through interpolation of the values of its natural neighbours (on the order of 10). The number of nodes decreases with
depth as the volume decreases. The basis function at each node is smooth, and dependent on the distances of other nodes
around it. No separation of radial and horizontal basis functions is needed.
Synthetic tests with 3D S-model S362 show that most of significant structures of S362 are recovered. Two major differences
exist: (1) at 2,500 km and 2,800 km depth, the recovered model is smoother and has less short-wavelength signal than S362;
(2) at the surface and CMB, the recovered S-velocity perturbation tends to be larger than S362. Currently we are deriving a
whole mantle P-model. The preliminary result is promising.
Compared to the traditional approach which employs both radial and horizontal basis functions, this parameterization allows a
true 3D representation of the Earth's seismic velocity perturbations. It is also straightforward to build norm and
roughness damping matrices and apply them to the target model. Because of the ease of the grid generation and the
grid-dependent basis functions, it is also possible to introduce multigrid (finer resolution in a particular region) into the
3D seismic tomography to achieve multi-scale resolution.
DE: 7200 SEISMOLOGY
DE: 7260 Theory and modeling
DE: 8180 Tomography
SC: Seismology [S]
MN: 2003 Fall Meeting