Inconsistent Beachball Diagram with Same Strike, Dip, and Rake

[Rinku-274]

I successfully reproduced the solution for the example event shown here. However, when I repeated the process using only a few stations, fixing the strike, dip, and rake values while allowing for a time shift, the resulting beachball diagram changed. Specifically, the fault type switched from left-lateral to right-lateral, even though strike, dip, and rake remained the same. Please review this issue.

Expected solution:

Solution after fixing the orientation:

[rmodrak]

Hi Rinku, Thank you for the report. It's possible this is related to the change from GMT to matplotlib-- we will need to investigate

[rmodrak]

This was a very useful report that identifies a rounding or numerical issue that arises when strike and dip are close to 90 degrees. Hoping it can be investigated...

[thurinj]

Hi @rmodrak, @Rinku-274 ,

I have made a quick sanity check to see if the plotting function could be an issue:

from mtuq.graphics.beachball import _plot_beachball_matplotlib
import numpy as np
import matplotlib.pyplot as plt

_plot_beachball_matplotlib(None, np.array([0.49995996217394867, -615579475417923.2, 615579475417924.2, -3019123387859.885, 97992717618959.02, 9980553602963706.0]))
plt.show()

where I basically pulled the values out of the json solution file from the example you mentioned above (20171201023244DC_solution.json).

I get the following beachball:

Given that the matplotlib functions directly projects the radiation coefficient onto the surface of the sphere from the input Mij parameters it's unlikely there is something wrong here. Could the error come from something more internal (kappa/sigma/phi conversion to Mij parameters for instance)?

It would be helpful to compare the Mij values you get out of your input grid and compare with the json file @SeismoFelix attached to his test. That may give me some leads to understand what's going on here. Thanks in advance for your help.

Cheers.