Re: [ferret_users] lanczos function - retaining annual cycle signal in monthly time series

["William S. Kessler" <william.s.kessler@xxxxxxxx>]

I'm not sure if the Lanczos function is appropriate for the annual  
cycle. It might not be doing what you need. The usefulness of a  
Lanczos filter is for working with precise frequencies. That is not  
usually the case for geophysical annual cycles, which typically  
include 1cpy, 2cpy, 3cpy, etc signals, and are complicated by leap  
years.

Perhaps try the regridding:

set data climatological_axes    ! bring in predefined annual (and a  
few other) modulo axes
let ann = original_variable[gt=month_irreg@mod]

The newly-created variable ANN will be the average annual cycle  
(monthly), correctly taking into account leap years. If you want  
another interval, look at the definition of month_irreg (or  
month_reg); analogies can be defined for daily or others.

If you want to check the result of this regridding, try:

let ann_on_orig=ann[gt=original_variable]    ! put the annual cycle  
onto the original time axis

plot original_variable, ann_on_orig, original_variable-ann_on_orig   !  
plot the original data, the annual cycle, and the anomalies

BK

On 21 Sep 2011, at 6:02 PM, Torben Mueller wrote:

> Dear Ferret Community,
>
> I want to use the lanczos filter to remove any signal from a time  
> series that is not the annual cycle.
> Regrettably, the documentation for the function is not very clear to  
> me, especially in which units the low and high frequency cutoff  
> points need to be given.
> My time series has monthly data and I just want to keep the annual  
> cycle. Which values should I choose for F1, F2 and N?
>
> Thanks!
>
> http://ferret.pmel.noaa.gov/Ferret/documentation/users-guide/appendix-a-external-functions/LANCZOS/