Possible issue with window rejection

[chenyu-lii]

Hello!

My name is Chenyu, and I am using hvsrpy to compute HVSR curves for my dataset. Thanks for this great open package.

I applied this script to my data: example_mhvsr_traditional_window_rejection.ipynb. I tried to read in multiple mseed files, and it works well for some stations. Unfortunately, for a few other stations, the window rejection seems not working, they didn't reject any curves, and the error seems quite large. The parameter is always n=2.

I was wondering if there are any reasons for not rejecting windows, such as the waveforms or HVSR curves are too noisy?

Thank you so much for your time!

Regards, Chenyu

[jpvantassel]

Hi @chenyu-lii,

The full algorithm is described in the Cox et al. (2020) paper. There is no exclusion of windows, the only issue may be if the HVSR does not have a peak. It’s hard to say what the issue is, without more information. Can you share some examples of what you mean?

[chenyu-lii]

Thanks for your reply! Here are some examples.

Probably it's because the HVSR didn't have a clear peak?

[alangi]

Hi @chenyu-lii, You could try selecting only the lower peak by changing the line below, which can be found in

Apply Window Rejection Algorithm

search_range_in_hz = (None, None)

Change the second None to be 3, so:

search_range_in_hz = (None, 3)

This is the high cutoff for the search range, and will stop the algorithm looking at any of the peaks above 3 Hz. You may still have issues though because the lower frequency peak has a broad range of peaks around it. I can see that the rejection algorithm is working in the second image though because the is one peak at about 0.7 Hz that is red.

[chenyu-lii]

Thanks for your suggestions. I could see an improvement after setting the search_range_in_hz = (None, 3).

Here is an example after changing the search range:

The HVSR before looks like this:

Thank you so much again!