Improving resolution of a signal in the frequency domain

I have ten years' worth of data and have reason to believe it contains a 4-year periodicity. I have never done any sort of signal processing before, but I examined the discrete Fourier transform of this data, and there is a peak at the frequency corresponding to about a five year period; this is the nearest point to four years (in period), so it is (I think) consistent with my belief. However, I would like to be able to get better resolution here. Is there any way to do this, short of waiting a few years for more data?

Someone told me that there is a method involving applying a windowing function and shifting its center to perform a sort of interpolation in the frequency domain, but I haven't been able to find any information on this. Then again, I'm very new to this and am probably using all the wrong words in my search. Does this ring any bells for anyone?

Also, I'm not sure if this is relevant, but this data is sampled every few seconds, but it's satellite data, so to try to remove any orbital effects, I average a day's worth of data so that my signal consists of 3652 points. If it would help, I could try to average over only a single orbit, but I don't think it. Am I correct in thinking that increasing the sampling rate or interpolating the time series will do me little good when trying to improve resolution at the very lowest end of the frequency domain?

Thanks in advance.

Answer 1

[Matt J]

Padding the data with trailing zeros will increase resolution in the frequency domain.

[Walter Roberson]

That would have the effect of convolving the data, which is probably not a good idea in this case.

[Matt J]

Zero-padding does convolve the spectrum with a sinc interpolator, but that's what leads to the increased sampling. Not sure what the hazard is.

[Jordan Rozum]

Thank you. That seems to have done the trick.