[ncl-talk] EOF Reconstruction of mode 2 larger than mode 1

[Dennis Shea]

I think the only NCL examples of *reconstruction* are based on a simple
text book:

  https://www.ncl.ucar.edu/Applications/eof.shtml

Unweighted:
  Example: eof_0.ncl and eof_0_640   [use 6.4.0 if available]

Weighted:
  Example: eof_0a.ncl   [modify for 6.4.0 if available]

---
FYI: NCL's EOF works on anomalies

Good luck

On Fri, Dec 15, 2017 at 2:52 AM, 蘇容瑩 <kikyourobin at g.ncu.edu.tw> wrote:

> Hello all,
>
> I input omega data(time, lev, lat, lon) and do EOF for every grid point to get characteristic vertical structure of omega.
>
> I used eofunc to get 3 modes of eigenvectors.
>
> However, I met a problem, when I reconstructed mode 1(w1) and mode 2 (w2).
>
> I tested 3 grid points and compared reconstructed omega and original omega.
>
> grid 1 : 0N, 150E in western Pacific
>
> grid 2 : 10N, 120W in eastern Pacific
>
> grid 3 : 10S, 110W in eastern Pacific cold tongue
>
> Result in grid 1 & 2 match my intuition. w1 >w2
>
> However, result in grid 3 against my intuition. w2>w1
>
> After examining data, I found ts_mean of mode 2 larger then mode 1 so that w2 larger than w1.
>
> I attached my code for EOF reconstruction below.
>
> Is there anything wrong with my conceptions or coding process?
>
> Thanks for help!
>
>
>
>     do mode = 0, 2
>
>         w_re = new((/nSeason,klev,nlat,mlon/),float,msg_val)
>         w_re = 0.
>         dimName = (/"time", "plev", "lat", "lon"/)
>         do dim = 0, 3
>           w_re!dim = dimName(dim)
>           if (dim .ne. 0) then
>             w_re&$dimName(dim)$ = w_eof&$dimName(dim)$
>           end if
>         end do
>         do i = 0, nlat-1
>           do j = 0, mlon-1
>             do time = 0, nSeason-1
>               if (num(ismissing(w_eof(mode,:,i,j))) .le. 5) then
>                 w_re(time,:,i,j) = w_eof(mode,:,i,j)*(w_ts(mode,time,i,j) + w_ts_mean(mode,i,j))
>               end if
>             end do
>           end do
>         end do
>
>
>
> --
>
> -------------------
> Rung-Ying, Su
> Department of Atmospheric Sciences
> National Central University (http://www.ncu.edu.tw/)
>
>
>
>
>
>
> _______________________________________________
> 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/20171215/94e5ee55/attachment.html>