On Fri, Apr 15, 2022 at 6:29 AM ashwin .D <winash12@gmail.com> wrote:

Robert,
Thanks. I have enclosed the results for your interest. So x.png is the untrended plot of periodogram vs. frequency. Note the spike near the low frequency area of the x axis(close to zero). Those I believe are the atmospheric tides(https://patarnott.com/atms411/pdf/class2018/atmosTides.pdf__. But the detrended plot has a richer zoo of frequencies including some higher frequency ones while at the same time reproducing the atmospheric tides.

I had one final question. In this peer reviewed paper (figure 4) https://journals.ametsoc.org/view/journals/apme/57/10/jamc-d-17-0250.1.xml they calculate the phase of the Lomb Scargel periodogram. I checked the scipy API for that but it does not seem to be available. It is available in the gatspy/astropy packages https://stackoverflow.com/questions/49859075/lomb-scargle-phase. Is that the better way to go then ?

They don't implement the Lomb-Scargle DFT as presented in that paper. They do have some facility for helping you find the important peaks from the periodogram and then help you do a least-squares fit to a sinusoidal model with those fixed frequencies to get the phases. Which may be all you need. The NFFT that I mentioned might be closer (if not identical; I haven't read too deeply) to the LS-DFT.

Robert Kern