2.4 Tomography inversion
An optimal initial velocity model is often necessary for the success of
tomographic inversions. In this study, we invoke the global 3-D
lithospheric model LITHO1.0 (Pasyanos et al., 2014) as the starting
model. Our tomography region spans the area from 20°N to 30°N in
latitude, 93°E to 111°E in longitude, and from the surface down to 155
km in depth. We parameterize our model by a 3-D grid with grid spacing
of 0.1° horizontally and 2.5 km vertically, which enables us to capture
small-scale anomalies in regions where the ray paths are dense.
In this study, we adopt the inversion scheme FMM of Rawlinson and Urvoy
(2006) which uses the multi-stage fast marching algorithm (de Kool et
al., 2006) to compute the P-wave travel times in 3-D models and the
subspace technique (Rawlinson et al., 2006) to iteratively solve the
inverse problem of finding the model parameters that minimize the
residual between the observed and model-predicted travel times with
appropriate regularization parameters. We have conducted extensive
inversion tests to find the optimal regularization parameters, and the
final values of 0.1 and 5.0 are determined for the damping and smoothing
constraints, respectively, leading to results with sufficient
small-scale structural features but without unwarranted strong localized
anomalies and sharp contrasts in velocities.