mode-kissing

Efficient Computation of Dispersion Curves in Low-velocity Layered Half-spaces

A recurring challenge in practical scenarios involves models incorporating low-velocity layers, where the occurrence of ``mode-kissing'' and missing roots presents difficulties. In response to this issue, we present an algorithm designed to generate supplementary samples between neighboring roots at points of mode-kissing. By incorporating additional samples, our proposed algorithm demonstrates remarkable efficacy and stability, establishing itself as a valuable tool with broad applicability in related fields.

It is important to highlight that the proposed algorithm herein is equally applicable to models without low-velocity layers.

Dependencies

Install these external libraries according to the operating system, e.g. aptitude, pacman

• Eigen3
• HDF5

Complie

> git clone https://github.com/pan3rock/mode-kissing.git
> cd mode-kissing
> git submodule init
> git submodule update
> mkdir build
> cd build
> cmake ..
> make

Examples

There are an instance located in the directory demo corresponding to Model 1.

> # the position here is the main directory of the program
> cd demo
> ls .
config.toml model.txt

The configuration file config.toml serves as the configuration file for the executable, containing easily understandable parameters. On the other hand, model.txt functions as the model input file providing descriptions of model parameters. Each line in model.txt includes parameters such as the "index of layers", "depth of the top interface of the layer", "density", "Vs" and "Vp". It is important to note that the numbers in the first column have no significance during the program's execution.

1. Calculation of dispersion curves

   > ../../bin/forward
   > ../../python/plot_disp disp.txt

   The information of every command can be inquired by adding "-h", and more functional arguments are provided.

   

2. Scanning the dispersion function along phase velocities

   > ../../bin/secfunc 10
   > ../../python/plot_secfunc.py secfunc.h5 --sign --disp disp.txt

   The command is to compute the dispersion function at the frequency 10 Hz.

   

3. Calculation of dispersion curves if the proposed algorithm not applied

   > ../../bin/forward --raw
   > ../../python/plot_disp disp.txt

   

4. Calculation of dispersion curves if the iteration ends at the LVL

   > ../../bin/forward 
   > cp disp.txt disp_full.txt
   > ../../bin/forward --raw 
   > ../../python/plot_disp disp.txt --file_ref disp_full.txt

   

Parameters

For certain models, the algorithm may occasionally miss some roots. To reduce the likelihood of this occurring, you can increase thenfine parameter in disp.hpp to a larger number. While this adjustment can help ensure that fewer roots are missed, please be aware that it will result in longer running times. Especially for derivative-free inversion alogrithm that models are genereated randomly, e.g. MCMC, we suggest set a relatively large nfine if the program is used as the forwarding solver.

References

• Pan, L. & Chen X. (submitted). Efficient Computation of Dispersion Curves in Low-velocity Layered Half-spaces. Bulletin of the Seismological Society of America.
• Buchen, P. W., & Ben-Hador, R. (1996). Free-mode surface-wave computations. Geophysical Journal International, 124(3), 869-887.
• Fan, Y., Chen, X., Liu, X., Liu, J., & Chen, X. (2007). Approximate decomposition of the dispersion equation at high frequencies and the number of multimodes for Rayleigh waves. Chinese Journal of Geophysics, 50(1), 222-230.