Try the following:
(1) First subtract the mean from each of the 3 trends. They look reasonably stationary, so that should be OK.
(2) Take the fft, (using the fft command) of each of the 3 columns individually. You can actually do that in one go if you want.
(3) Plot the abs(fft(x)) for each of the 3 signals. This will be the correct shape, but the units will be normalised. You should also note the mirror line in the middle. I'm assuming you are most interested in the power within each frequency band as opposed to the phase.
Now since you have neglected to both give us the data, and the sampling time, we can only roughly guess from the plot what the sampling rate, and hence Nyquist frequency will be.
Lots more detail here: https://au.mathworks.com/help/matlab/math/fft-for-spectral-analysis.html