Padding zeros and filtering issues with MTUQ synthetics

[ammcpherson]

Hello,

I have been making and saving synthetics using MTUQ and SPECFEM3D-generated Green's functions for later use. These GFs are 23 seconds of data and should contain 2 main arrivals, and potentially some very small reflections off my mesh boundaries.

However, I noticed that the MTUQ synthetics end up with about 140 seconds of data, with approximately 60 seconds of zeros added to both the beginning and end of the trace. At the beginning of the trace, this is fine because the GF traces start at 0 amplitude. However, the end of the GFs is not 0 and this introduces an enormous artifact in the records after filtering.

The image above shows an example of a synthetic that I created and saved from MTUQ. The top panel shows the raw synthetic, then the filtered synthetic with an abnormal spike right at 23 seconds after the origin time. The third panel I have trimmed the trace to +/- 10 seconds before and after the start and end times of the SPECFEM GF, and the fourth panel shows the trimmed trace filtered. The final 2 panels show the synthetic trimmed the the same length as the GF, and then filtered.

I believe there needs to be some control when saving the synthetics from MTUQ in how long of a record is produced, and some sort of tapering at the end of a record before adding zeros to avoid these artificial spikes.

Thanks,
Amanda

[rmodrak]

Hi Amanda,agreed this can be important. mtuq’s process_data.py was designed with this in mind.  if you could, please search for “padding” options in that module and see if you could get it to work, or implement your own approach. it’s a useful feature, but it’s been a while since it’s been tried out. can touch base further after I get back from travel. thanks -Ryan