numerical noise

[VarunD21029]

Description

I am simulating earthquake simulation for Mw>7 using dip, rake and strike for the fault. Further, I am using a random slip model for the fault. My domain is 500 by 500 km Approximately, and I have defined 21 recording stations. I am getting noise in the output-displacement time histories. I know it will take many replies and time on someone's behalf, but I would be grateful if someone could help me in this regard. I am ready to share the files for further explanation of the problem.

Affected SPECFEM3D version

Specfem3d Cartesian

Your software and hardware environment

Linux Ubuntu 22.04 on highend 128 cores system

Reproduction steps

1. Go to '...'
2. Change settings '....'
3. Use data files '....'
4. See error

Screenshots

![DESCRIPTION](LINK.png)

Logs

No response

OS

No response

[jpampuero]

Numerical noise could be due to your source containing significant amplitudes at frequencies that are too high to be well resolved by your mesh, i.e. at wavelengths that are smaller than the element size. You may want to plot the spectrum of your source (the spectrum of the slip rate functions or the moment rate functions of your source model) and check if the spectral amplitude decays rapidly enough with increasing frequency. Especially, check if the relative amplitude is small enough at frequencies close to (S wave speed)/(element size).
If that is an issue, you can low-pass filter either your sources or your seismograms (it's equivalent because the problem is linear), or use a finer mesh (if you really want to resolve those high frequencies).

[VarunD21029]

T

Numerical noise could be due to your source containing significant amplitudes at frequencies that are too high to be well resolved by your mesh, i.e. at wavelengths that are smaller than the element size. You may want to plot the spectrum of your source (the spectrum of the slip rate functions or the moment rate functions of your source model) and check if the spectral amplitude decays rapidly enough with increasing frequency. Especially, check if the relative amplitude is small enough at frequencies close to (S wave speed)/(element size). If that is an issue, you can low-pass filter either your sources or your seismograms (it's equivalent because the problem is linear), or use a finer mesh (if you really want to resolve those high frequencies).

Thanks for the reply, I think it solved the issue for now. I will keep posting updates on the issue.