[ncl-talk] eofunc_ts

[Dennis Shea]

Umm, your semantics are confusing (to me).

"find" ==> reconstruct?

[1]
Generally, geophysical  based EOFs are quite "red" in nature. Most 'power'
(variance) is in the lowest few modes. For sure, 5 or fewer.

[2]
Further, *each* EOF should be significantly ("distinctly")  separated from
the surrounding EOFs to be used for reconstruction of 'significant'
patterns. See:

   http://www.ncl.ucar.edu/Document/Functions/Contributed/eofunc_north.shtml

If, say, your data show 2 or 3 distinct EOFs then the 27/28  other EOFs
that you calculated are  artifacts of sampling.

[3]
NOTE : *statistical-significance" is independent of "physical-significance"
. One  does not imply the other.
You should sum using *only* the contributions of the significant EOFs.










On Tue, Apr 3, 2018 at 3:36 AM, Tobias Kleine <tobias.kleine at gmx.net> wrote:

> Hi there,
>
> I am using eofunc to calculate the first 30 EOFs of detrended SST
> anomalies. I add these individual EOFs up to a "combined" EOF including a
> weighting, like so: eof_comb = dim_sum_n(eof*weights,0). This is
> pseudo-code obviously.
>
> I then try to find with which temporal variability the new combined EOF is
> present in my original SST anomalies:
> pc_new = enfunc_ts(sst_anom,eof_comb,False), where sst_anom is lat x lon
> x time and eof_comb is evn x lat x lon.
>
> My question is, since I can't find source code for eofunc_ts, is the
> routine capable of "finding" my combined EOF in the original data and gives
> me the correct PC or do I need to expect wrong results because I meddled
> with the EOFs?
>
> Regards,
> Tobias
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk at ucar.edu
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20180403/9b588cb3/attachment.html>