Re: Strange behaviour of rcm2points with opt=2

Dear NCL team,

I just realized looking at the Fortran code that NCL uses for the
implementation of rcm2points.f that I might be responsible myself for
the strange behaviour. The code uses a value "k" which determined the
size of the patch (in number of gridpoints) where it searches the output
points to fall into. I suggested implementing bi-linear interpolation
(opt=2) and also suggested that for bi-linear setting k=1 should be
fine. It seems that there are pathetic cases where a value of k=1 leads
to the fact that no source patch is found to perform the interpolation.
The original value of k=2 should be restored both for opt=1/2 and then
everythings works fine again.

Kind regards,
Oli
 
________________________________________

Oliver Fuhrer
Numerical Models

Federal Departement of Home Affairs FDHA
Federal Office of Meteorology and Climatology MeteoSwiss

Kraehbuehlstrasse 58, P.O. Box 514, CH-8044 Zurich, Switzerland

Tel. +41 44 256 93 59
Fax +41 44 256 92 78
oliver.fuhrer@meteoswiss.ch
www.meteoswiss.ch - First-hand information
  
 

> -----Original Message-----
> From: ncl-talk-bounces@ucar.edu
> [mailto:ncl-talk-bounces@ucar.edu] On Behalf Of
> Oliver.Fuhrer@meteoswiss.ch
> Sent: Donnerstag, 24. Februar 2011 18:54
> To: ncl-talk@ucar.edu
> Subject: Strange behaviour of rcm2points with opt=2
>
> Hi NCL team,
>
> I'm seeing a strange behaviour of the rcm2points function inserting
> missing values in the results, where input data should be
> available. The
> values seem to be correct when using the opt=1 (inverse distance
> weighting), but with opt=2 (bi-linear interpolation) the
> results contain
> (to me) unexplainable missing values.
>
> I've attached a small script where the problem occurs. The
> scripts read
> two two-dimensional arrays (x2d,y2d) which contain the lat,lon
> coordinates of a weather model grid. It then reads a short list of
> lat/lon points. Afterwards, it tries to find the location of these
> points in array index space by using the rcm2points function.
> Since the points are not located anywhere near the data boundaries,
> there is no reason for missing values to occurr.
>
> The list of three lat/lon points is...
> (0,0) 47
> (0,1) -5.405255
> (1,0) 47
> (1,1) -5.31714
> (2,0) 47
> (2,1) -5.229024
>
> The resulting list of three index-space points is...
> (0,0) 97.98473
> (0,1) 180.314
> (1,0) -999
> (1,1) -999
> (2,0) 99.83066
> (2,1) 179.9179
>
> On can see that the first and third points are well away of the
> boundaries, as the arrays are dimensioned (338,393). To me
> there is not
> reason, why the second point is returned as missing. Changing
> opt=1 also
> gives reasonable results.
>
> Any ideas are appreciated.
> Oli
>
> PS. Please contact in case you would like the dump.nc input file. I've
> put it onto a publicly available FTP server.
>
> ________________________________________
>
> Oliver Fuhrer
> Numerical Models
>
> Federal Departement of Home Affairs FDHA
> Federal Office of Meteorology and Climatology MeteoSwiss
>
> Kraehbuehlstrasse 58, P.O. Box 514, CH-8044 Zurich, Switzerland
>
> Tel. +41 44 256 93 59
> Fax +41 44 256 92 78
> oliver.fuhrer@meteoswiss.ch
> www.meteoswiss.ch - First-hand information
>
>
>
> ________________________________________
>
> Oliver Fuhrer
> Numerical Models
>
> Federal Departement of Home Affairs FDHA
> Federal Office of Meteorology and Climatology MeteoSwiss
>
> Kraehbuehlstrasse 58, P.O. Box 514, CH-8044 Zurich, Switzerland
>
> Tel. +41 44 256 93 59
> Fax +41 44 256 92 78
> oliver.fuhrer@meteoswiss.ch
> www.meteoswiss.ch - First-hand information
>
>
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Fri Feb 25 01:41:35 2011