Wavelet Coherence application

[thelightonmyway]

Hi,

I could use wavelet_coherence to do some wavelet analysis thanks for Pyleoclim. And I get some trouble when I'm reproducing the results of the paper from https://doi.org/10.1029/2023GL103323. I'm grateful to @CommonClimate for help and I could set the parameters same as the paper now.

However, there still seems some difference between my reproduction and the paper. It could be seen as:

the left is from the paper and the right is my result.

The main character is similar in periods 8-16 months compared with the origin, but it appears difference in other periods, such as:

1. the energy of mine is lower;
2. the arrow of mine is inconsistent.
   This is my result without significance, in order to see the arrow more intuitively.

I'm sure that the method and data I use is correct, since it's not complex with the result of the coherence wavelet does not change with the data being anomaly or original.

For reference, the data used could be get from:
MonthlyAnomSWS_IP.txt
pv.txt

What troubles me is that:

1. How could I draw the COI completely? since my COI seems not complete now.
2. What causes the difference between my result and the paper?

Any reply will be appreciated sincerely!

[CommonClimate]

Hi @thelightonmyway

1. you can tweak the frequency vector so it shows very long scales. Naturally, that is a bit of a waste since any wavelet coefficient under the cone of influence is, by definition, suspect.
2. That is for you to find out! Different packages implement the same method(s) in subtly different ways, and much depends on the parameters being chosen. Are the results qualitatively inconsistent?
   The default frequency method used in wavelet analysis in Pyleoclim is log, so that is one first source of potential difference to explore.

[thelightonmyway]

I see...thank you very much! And how could I change the frequency method of log? I means, how to alter it when I have set the frequency vector as freq_vector_scale(time,dj=1/12,s0=2,j1=83,mother='MORLET',param=6)?

[CommonClimate]

Sorry, I meant in general that the defaults in Pyleoclim were different from Aslak Grinsted's code. In this case, your frequency vector should be getting somewhat closer, but again there is no way to know unless you do a side-by-side comparison, and that's not our job! Good luck and please let us know if you have other questions.

[thelightonmyway]

Well, I got it, thank you for your help!