Unable to[BUG] - Unable to get power in the signal upto resolvable period.

[VarunD21029]

Description

Is there any problem with SPECFEM3d for finite fault simulations? I am running a job for 6.9 Mw event. My minimum resolvable period from the mesh is 0.5s, but when I evaluate the spectra from the output signal, for the majority of stations, it dies down before 1s. My model is a simple layered medium with no topography included. I have tried a number of trials, but all failed. I am attaching a screenshot for an explanation of the problem. Red one is the simulated and black one is recorded signal.

Affected SPECFEM3D version

SPECFEM3D 4.0.0

Your software and hardware environment

Ubuntu 22.04

Reproduction steps

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

Screenshots

![DESCRIPTION](LINK.png)

Logs

No response

OS

Linux

[jpampuero]

What do you mean by "simulated signal" and "recorded signal"?

[VarunD21029]

@jpampuero Simulated signal is the spectra from the specfem output and recorded signal is the spectra recorded at station.

[jpampuero]

What do you mean by "the specfem output"? The seismograms "recorded at station" are also "specfem outputs", so I am confused by your statement.

[VarunD21029]

@jpampuero Yes it the accelration corresponding to that station

[VarunD21029]

One is from the specfem .sema file and other is from the recorded value for that earthquake.

[jpampuero]

Oh, "recorded signal" is a real seismogram from a real earthquake! Got it, sorry.

The shape of the black and red curves look similar to me at frequencies below 1 Hz (if we ignore their amplitude difference; in other words, if we shift them vertically). They have a similar decay. Can you point out in the figure what is the feature that worries you?

A plot of the source time function used in the simulation and its spectrum might also help.

[VarunD21029]

@jpampuero Thankyou for your reply. My concern is regarding the red signal i.e. if we look into the rightmost point of the red signal and comes towards 1hz the signal dies down. That is the part which I got from SPECFEM simulations.

[jpampuero]

Oh, the horizontal axis is period, not frequency! Thank you for adding axis labels to your figure. And your issue is not that the spectrum "dies down before 1s", but that it increases with increasing periods above 1 s. That's the main difference between black and red curves in your figure. Can you plot the spectrum of the source time function used in the simulation? Can you also plot the SPECFEM accelerogram, and indicate in the plot the time window used to calculate the red spectrum?

[VarunD21029]

@jpampuero Sure T=[0 0.01 0.02 0.03 0.04 0.05 0.06 0.075 0.09 0.1 0.15 0.2 0.3 0.4 0.5 0.6 0.7 0.75 0.8 0.9 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.8 3.0 3.4 3.8 4.0];

[jpampuero]

Can you indicate in the figure the time window that you used to compute the spectrum?

[VarunD21029]

@jpampuero I used the entire signal from 0 to 100s.

[jpampuero]

Then can you replot the spectrum up to a period of 100 s?

[VarunD21029]

@jpampuero Sorry for the misunderstanding. The period window of the spectra is from 0-4s

[jpampuero]

Can you compute the spectrum of the whole time window of the accelerogram, from time t=0 s to t=100 s, and then plot it up to the period T=100 s?

[planetarianPKU]

Your result has a strong numerical dispersion from Pphase to Sphase and in coda wave, indicating that the maximum frequency of your setting is not enough to support the simulation.