Re: Centered and uncentered pattern correlations

The pattern correlation function (pattern_cor) is located in:

%> less $NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl

This calls 'pattern_cor2' whicg contains

   if (opt.eq.0) then ; centered correlation
       sumWGT = sum(WGT)
       xAvgArea = sum(x*WGT)/sumWGT ; weighted area average
       yAvgArea = sum(y*WGT)/sumWGT

       xAnom = x - xAvgArea ; anomalies
       yAnom = y - yAvgArea

       xyCov = sum(WGT*xAnom*yAnom)
       xAnom2 = sum(WGT*xAnom^2)
       yAnom2 = sum(WGT*yAnom^2)
   else
       xyCov = sum(WGT*x*y)
       xAnom2 = sum(WGT*x^2)
       yAnom2 = sum(WGT*y^2)
   end if

So, "no" the overall areal mean is subtracted

On 7/18/12 1:30 PM, Noel Aloysius wrote:
> Hi NCL,
>
> This is perhaps an obvious question!
>
> I am computing pattern correlations of monthly precip climatologies
> (model v. observed) for each month for a region. When the centered
> pattern correlation is used, say for the month of January, is the
> spatial average for that month removed from each grid point within the
> region before the correlation is computed?
>
> Thanks,
> Noel
>
>
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Wed Jul 18 13:49:29 2012