Anisotropic Crustal Structure Inversion Using a Niching Genetic Algorithm.
Abstract
Seismic anisotropy in the crust can provide important constraints about past and present tectonic processes. The inversion of seismograms with anisotropic effects is a problem that is not easily linearized, and consequently linear inversion techniques are not suitable for the determination of anisotropic crustal structure. One global minimization technique that is applicable to highly nonlinear problems is that of genetic algorithms. We are using a niching genetic algorithm to invert synthetic seismograms with anisotropy in order to determine which model parameters can be resolved, how sensitive the parameters are to noise, and how many back azimuths are required to constrain the parameters. A niching genetic algorithm is useful for finding multiple potential solutions by determining various minima that are separated in the solution space be some pre-determined value. By using synthetic data we are able to compare the resultant model with the original model and determine which parameters have been correctly resolved. We are also able to introduce noise into the synthetic data and determine the effect that this has on the results. Finally, we can include a variable number of back azimuths in order to determine how many are required to constrain the model parameters suitably. We are using an algorithm that attempts to fit simultaneously the radial and transverse components of receiver functions for multiple back azimuths. It is necessary to include both components of the seismograms and multiple back azimuths in the inversion because the solutions for one component or one back azimuth are highly non-unique. By including multiple back azimuths more constraints are placed on the problem and many duplicate solutions are eliminated. So far, all our tests have been performed on models with one anisotropic layer over an isotropic half space. These tests have shown that thickness tends to be the best resolved parameter. We have also found that when multiple back azimuths are included the strike and dip of the anisotropy are well constrained, and that the percent anisotropy is the least constrained. The niching aspect of the genetic algorithm has proven to be important because we have found that in some cases a single solution test is unable to find a good solution while a multiple solution run is able to better fit the data and find a more correct solution. We hope to refine this method in the future in order to obtain better resolution of all the model parameters and apply the technique to real data from the Tibetan plateau and the Andes.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2001
- Bibcode:
- 2001AGUFM.S51E..05E
- Keywords:
-
- 3260 Inverse theory;
- 7203 Body wave propagation;
- 7205 Continental crust (1242);
- 7260 Theory and modeling