Attached are plots (Pictures 1&2) of MICH, PRCL, SRCL, and DARM spectra (one plot is zoomed into 10-100 Hz):

References are before the earthquake (06/23/17 03:00:00)

Non-References are after the earthquake (07/20/17 03:00:00)

Seems like no real differences except for MICH so I made a TF between it and DARM  just to make sure (Picture 3)

it doesn't look like these DoFs are the root of our problem.