>I performed SVD analysis between annual forest cover and SSTs using svdstd
>and got the homlft and homrgt patterns for modes 1-4 and the percent variance.
>I plotted the maps of SST and forest cover patterns overlapped for each mode.
>Now I need time series for each mode. How do I get the time series,
>for example,
>for mode 1? It was suggested to regress the original data onto the
>pattern but I am
>not certain how to do this given two variables, veg(lat,lon,time) and
>sst(lat,lon,time),
>and both homlft(0:3,lat,lon) and homrgt(0:3,lat,lon). Any suggestion
>is appreciated.
----Could you not use "eofcov_ts" http://www.ncl.ucar.edu/Document/Functions/Built-in/eofcov_ts.shtml The act of "projecting" is multiplying the data by the modes and summing over time.
Something like the (untested) following in NCL [which would likely be slow.] If there are missing values then other statements would have to be added.
do k=0,nmode-1 do nt=0,ntim-1 xts(nt,k) = 0.0 do nl=0,nlat-1 do ml=0,mlon-1 xts(nt,k) = xts(nt,k) + mode(k,nl,ml)*x(nl,ml,nt) end do end do end do end do
The homlft [homrgt] would be like the EOF patterns [neof,lat,lon]. eg:
veg_ts = eofcov_ts(veg,homlft)
D
_______________________________________________ ncl-talk mailing list ncl-talk@ucar.edu http://mailman.ucar.edu/mailman/listinfo/ncl-talk
This archive was generated by hypermail 2b29 : Wed Oct 26 2005 - 06:49:08 MDT