Some parameter wish to added in wavelet_coherence

[thelightonmyway]

I'm learning the method on wavelet_coherence in pyleoclim and I find it very convenient!

However, I'm facing some difficult that I found my plot is a little difference when I reproduce others work. I guess it's due to the resolution. How can I change some parameters to improve the resolution? For example, specify that mother wavelet as morlet wavelet, 2 months as the starting scale, 12 sub-octaves per octave and 84 as the total number of scales.

The left is my reproduction and the right is the result from the paper https://doi.org/10.1029/2023GL103323

By the way, the data series pre-processed could be download.
MonthlyAnomSWS_IP.txt
(The data from the paper https://doi.org/10.1029/2023GL103323)
and
pv.txt

Thanks, sincerely!

[CommonClimate]

Hello @thelightonmyway and glad you are finding the package useful!

Given that the paper you mention seems to use Aslak Grinstead's code, which was based on Torrence & Compo 1998, I believe you can achieve the results you want by playing with the frequency vector.

You can use this function to generate a frequency (scale) vector with the resolution you want and then pass it to Series.wavelet_coherence() as the freq argument. We show something similar in this tutorial though this is not exactly what you want. All you need to do to access the function I mentioned is
from pyleoclim.utils.wavelet import freq_vector_scale

then you'll be on your way. Please let us know what you get!

[CommonClimate]

PS: I am closing this issue as solved. In the future, questions of this kind are best asked in the discussion forum, as this is not an "issue" with the code, only a matter of usage.

[thelightonmyway]

Well, thank you for your kind guidance!@CommonClimate

I utilize the method freq_vector_scale following your teach, and I get the similar result as before. The difference is that, the scales number is decreased, which is what I want.

However, I'm confused that my result is still low-energy and the significant result is less than the result in paper. And when I set the confidence area not to show, I find my arrows is different compared with the paper, expect for the scales 8-16. I wonder if it could be the problem of the parameters from the function or others.

In addition, I find that the COI in my result is not complete. How can I complete it?
Besides, when I set the freq, I find the order is reversed , like this,

and I need change it to order. Otherwise, it appears error when I use coh=pv_series.wavelet_coherence(SWS_series,freq=freq). And the error shows, ValueError: negative dimensions are not allowed.

This is my new result.

[CommonClimate]

It's not easy reproducing your graph exact, so once PR #635 is incorporated (should be a few minutes), can you please run this again and make sure you are getting the results you should?