Preprocessing Issues with synthetic traces

[ammcpherson]

Hello,

I have discovered that the default preprocessing steps in MTUQ can introduce long-period spurious signals in synthetic data depending on the frequency content. This applies to both SPECFEM3D forward simulations, and synthetic data created in MTUQ from SPECFEM3D Green's functions.

To start, here is an example of what I expect my data to look like. These are SPECFEM3D radial component synthetics that have had a 2-pass filter applied to them (minimum frequency for all plots is 0.1 Hz):

When I then apply the same preprocessing steps that MTUQ applies to data, in the same order that MTUQ applies them, a long-period signal is introduced mainly at the beginning of the traces:

I then systematically 'turned off' different features of the MTUQ pre-processing, such as removing the demean step but keeping the linear detrending, or keeping both detrending steps but changing the filter to a 2-pass filter:

As we can see, all these options still have the long-period spurious signal. However, when I completely remove any detrending steps, the signal no longer appears:

I believe that implementing more control over the detrending steps would be ideal for users and increase the flexibility of the code overall. For most things it appears that these preprocessing steps should be the default option though.

Thank you, Amanda