Segmented r_power_n

[krisvanneste]

Hello Claudio,

My best wishes for the new year!

I have created a new branch adding support for a segmented powerlaw spreading model, similar to Boore (2003), and implemented in simulation codes like EXSIM and SMSIM. I also think that we could move all the geometrical spreading related functions in ssp_util.py to a dedicated ssp_geom_spreading.py file. In principle, the original r_spread_n model can also be implemented with the r_spread_n_segmented model, but for now I left it, which may result in some overlap. This is a design decision that I prefer to leave to you.

I based this branch on v1.8 (because I want to apply it to my data and v2 is still experimental), but if you prefer, I could also base it on v2. I thought development of the v2 branch had stalled, but I now notice that you have continued developing. I will try to catch up as soon as possible.

[claudiodsf]

Dear Kris, happy new year to you too! 🌟

Thanks for the pull request, which looks nice. 👍

I rebased it on the latest main branch and added a few commits (including a Changelog entry 😉).
Let me know if that still works for you.

Concerning moving the geometrical spreading code to ssp_geom_spreading.py, I agree with that.

However, I prefer to proceed in this order:

1. merge this PR
2. rebase v2 over the current main (which will be already quite some work)
3. in the main branch, move the code to ssp_geom_spreading.py
4. rebase again v2 over main

[krisvanneste]

Claudio,

Thank you!

I force-pulled and ran it again on my example. The results are the same as before.

Here is a plot comparing a simple 1/R spreading model and the segmented Boore (2003) model, obtained by calling the r_power_n and r_power_n_segmented functions directly:

Of course I agree to proceed in the order you propose. Let me know if I can help.

[claudiodsf]

Great, I also wrote some code to compare the different spreading models. Once we create the ssp_geom_spreading.py module, we could add a utility function to plot the different models.

[krisvanneste]

Claudio, please wait before merging with the main branch.
I'm running into a problem with another example.

[claudiodsf]

Did it already! Just make another PR 😉

[krisvanneste]

OK, I first need to find the problem.

[claudiodsf]

However, I prefer to proceed in this order:

1. merge this PR
2. rebase v2 over the current main (which will be already quite some work)
3. in the main branch, move the code to ssp_geom_spreading.py
4. rebase again v2 over main

Points 3. and 4. are done! @krisvanneste, could you please double-check? Thanks!

[krisvanneste]

Points 3. and 4. are done! @krisvanneste, could you please double-check? Thanks!

I ran the example in my notebook with a segmented spreading model and it seems to work as expected!

[claudiodsf]

Great, thanks!

I'm currently working on adding to the module a plot_geom_spread_models() which should be useful for debug.
I'll let you know when it's done.

[claudiodsf]

Plotting function added to the mainbranch! (still have to do for v2).

Here's an example code and resulting plots:

import numpy as np
from sourcespec.ssp_geom_spreading import plot_geom_spread_models

source_depth = 10
epi_dists = np.arange(10, 1000, 10)
models = [
    ('r_power_n', {'exponent': 1}),
    ('boatwright', {'cutoff_dist_in_km': 100}),
    ('boatwright', {'cutoff_dist_in_km': 200, 'freqs': 0}),
    ('boatwright', {'cutoff_dist_in_km': 200, 'freqs': 10}),
    ('teleseismic', {'phase': 'P'}),
    ('teleseismic', {'phase': 'S'}),
]
plot_geom_spread_models(
    source_depth, epi_dists, models=models,
    plottype='linlog', xaxis='epi_dist')

source_depth = 0
epi_dists = np.arange(1, 1000, 1)
models = [
    ('r_power_n', {'exponent': 1}),
    ('r_power_n_segmented', None),
]
plot_geom_spread_models(
    source_depth, epi_dists, models=models,
    plottype='loglog', xaxis='hypo_dist')

[krisvanneste]

Nice!