Re: How to do 1D objective analysis (ie gridding irregular timeseries)?

[kessler@xxxxxxxxxxxxx (Billy Kessler)]

You can get a good deal of control by choosing how to define the
regridding time axis. For example, if you make it fine enough that 
it provides enough points to resolve (or nearly resolve) all of the 
original (irregularly-spaced) data points, then the various ferret 
smoothers should do what you want. Or it would be relatively easy to 
write your own using [L=@SHF:n]. I would guess that this will come 
very close to what the gaussian scatter2grid will do (and I wrote 
the original fortran code to scatter2grid). There is nothing magic 
about scatter2grid; it is just a gaussian-weighted running mean that 
includes points within twice the e-folding scale specified. Once you 
have regridded very finely, the equivalent of this could be written 
in ferret.

Billy K

> Mark Verschell wrote:
>
> 	If you don't want exact control over the regridding, you can just 
> use the built-in ferret axis regridding routines. All you do then is 
> define the new time axis.
> 
> Assume that your variable on the uneven time axis is TEMP, and the newly 
> defined time axis is regtime:
> 
> plot TEMP[gt=regtime]
> 
> For a list of regridding transformations (v5.22) see the ferret user's 
> manual:
> Ch4 Sec2.3.1.  Regridding transformations
> 
> > On Wednesday, May 23, 2001, at 11:33 AM, Lev Tarasoff wrote:
> > 
> > I have an irregularly spaced timeseries that I want to put onto a
> > regular chronology. How do I do objective gridding on the sequence of
> > (time,value) pairs? Is there a 1D version of scatter2grid around?
> > (I couldn't find anything useful from the archives)
> >
> > thanks
> >
> > Lev