Qbo wavelet #836
Qbo wavelet #836
Conversation
Adjust the plot panel to accommodate
e3sm_diags/driver/qbo_driver.py
Outdated
| def get_psd_from_wavelet(data): | ||
| deg = 6 | ||
| period = np.arange(1,55+1) | ||
| freq = 1/period | ||
| widths = deg / (2*np.pi*freq) | ||
| cwtmatr = scipy.signal.cwt( data, scipy.signal.morlet2, widths=widths, w=deg ) | ||
| psd = np.mean(np.square(np.abs(cwtmatr)),axis=1) | ||
| return ( period, psd ) | ||
|
|
There was a problem hiding this comment.
I'm not sure what the syntax guidance is for the diags - but I think it would be good to add a comment section for these routines. This one especially seems like it would benefit from a description. Something like this:
def get_psd_from_wavelet(data):
'''
Return power spectral density using a complex Morlet wavelet spectrum of degree 6
'''
There was a problem hiding this comment.
@whannah1 thanks for the review. I agree that inline documentation about the new function will be very helpful!
|
Thank you for this PR @justin-richling and @whannah1 |
|
It looks like I need to update the
Also the 4th plot I got seems problematic: |
|
@chengzhuzhang Thanks for pointing this out, the code was getting the power spectra level based on the test case only, not the reference case so I updated this to the |
* Add vertical line to indicate wavelet spectra peak and x-axis label for corresponding periods * Use square root values for clarity
|
@chengzhuzhang @tomvothecoder After discussion with @whannah1 it seems for now we will not be including the external wavelet analysis package until a further date. However, I added some clean up and small enhancements to the plot. Here is a snippet of the QBO diags with updated wavet plot:
With that, this PR should be good to review again, but I seem to be failing the auto checks. I can't seem to narrow down why these small changes are failing but I'm happy to fix them if there easy solutions. |
|
@justin-richling and @whannah1 thank you for the nice enhancement to identify the value of the peak period. I think we can do something similar to the original method. One thing I would suggest is that the x-axis is also updated and, now the wavelet method has different ticks compare to the original method, as shown below: Can we consider to revert this part of change to align x-axis for both plots. This makes it easier to compare both methods visually. Let me know what you think...Thanks! P.S. don't worry about the auto checks, I can fixes those formatting issues after we finalize the figures. |
That seems like a good idea. |
|
I slightly tweaked the plotting code, so that 3rd and 4th panels have same axis range, and with period of maximum power/variance labeled. The figure is shown below. @justin-richling @whannah1 and @jjbenedict let me know if you'd like any changes. Many thanks for this nice enhancement!
|
|
@chengzhuzhang It's great to see the peak align so closely, thanks for adding this to E3SM Diags! My only comment: I think there's a typo in the wavelet y-axis label. And, variance should be m2 s-2...? |
|
@jjbenedict thank you for catching the error. The units are now fixed with 3fd8aa5 @justin-richling the auto checks are also passed now with some formatting and after merging fixes from |
|
@tomvothecoder as a heads-up. We will need to bring this PR to the cdat-migration branch. |
Description
Add additional QBO diagnostics for power spectral density wavelet calculations and plot, see comment from #816 (comment)
Checklist
If applicable: