[Thread Prev][Thread Next][Index]

Re: [ferret_users] FFT.



Ryo is quite right.

I would add that sometimes an axis is intended to be regularly spaced but is stored in the netCDF file in such a way that Ferret detects it as irregular. Ferret reads the coordinates and looks at the spacing, to the precision of the coordinate data - that is if they're stored as DOUBLE, or FLOAT, it looks at the spacing relative to the data type. However sometimes an axis that seems to be intended to be regular is in fact not - quite.

It's worth looking at the coordinates themselves to see if they should be treated as regular, and if so then when opening the dataset you could tell Ferret to treat the axis as regularly-spaced. For the dataset containing the relevant time axis,

yes? use/REGULART  dataset.nc

Ansley

On 8/14/2015 12:37 AM, Ryo Furue wrote:
Hi,

I am trying to calculate the FFT for a data set which has irregular axis
I'm no expert, but I think the FFT algorithm works only for
evenly-spaced gridded data.  If so, what you need is to map your data
on to an evenly-spaced grid.

Mapping inevitably introduces leaks (aliasing) in the frequency domain
and if your problem is sensitive to aliasing, you would have to
carefully choose an appropriate mapping.  If, on the other hand, you
aren't concerned about a bit of aliasing, you can just use Ferret's
capability of regridding.  Look at the Ferret manual for regridding.

For general information about Fourier transforms of unevenly-spaced
data, google-search for "FFT uneven grid" without the quotation marks.

Hope this helps,

Ryo



[Thread Prev][Thread Next][Index]
Contact Us
Dept of Commerce / NOAA / OAR / PMEL / Ferret

Privacy Policy | Disclaimer | Accessibility Statement