Re: cssgrid does not ignore _FillValue

Travis,

I believe the documentation is misleading here. It does indeed claim
"If missing values are detected in the input data, then those values
are ignored when calculating the interpolating function."

However, in looking at the code, it appears it is only applying to the
lat/lon values, and not the actual data values.

I'll check if this is a bug, or the intended behavior.

Meanwhile, I think you can work around this in a kludgy fashion by
setting one of your lat/lon arrays to missing values in the same
locations that your input data is missing.

For example, if "x" is your data array:

   lat@_FillValue = -9999 ; if not already set
   lat = where(ismissing(x), lat@_FillValue, lat)

--Mary

On Thu, 26 Jun 2008, Travis O'Brien wrote:

> It appears that cssgrid_Wrap (and cssgrid) does not ignore missing values
> (either @missing_value or @_FillValue) like the documentation says. I
> discovered this when trying to interpolate station data (some of which are
> missing data) using cssgrid.
>
> The bug can be replicated by modifying the example script from the cssgrid
> documentation on the NCL website:
> _________________________
> ;**************************************************
> ;************ Load necessary ncl code *************
> ;**************************************************
>
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_code.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_csm.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/shea_util.ncl"
>
> begin
>
> ndata = 100
> rlat = new(ndata,float)
> rlon = new(ndata,float)
> fval = new(ndata,float)
>
> ;
> do i=0,ndata-1
> rlat(i) = -90. + 180.*rand()/32767.
> rlon(i) = -180. + 360.*rand()/32767.
> end do
> cartesian = css2c(rlat,rlon)
> fval = cartesian(0,:)2 + cartesian(1,:)2 - 0.5*cartesian(2,:)2
> fval@_FillValue = -1.0e20
> fval(32) = -1.0e20
>
> printVarSummary(fval)
>
> ;
> ; Set up the output grid.
> ;
> nx = 73
> ny = 143
> plat = 90.*fspan(-1.,1.,nx)
> plon = 180.*fspan(-1.,1.,ny)
>
> ;
> ; Calculate the interpolated function values.
> ;
> fint = cssgrid_Wrap(rlat,rlon,fval,plat,plon)
>
> print(avg(fint))
>
> end
> _______________________________
> Running ncl on this gives the following output:
> Variable: fval
> Type: float
> Total Size: 400 bytes
> 100 values
> Number of Dimensions: 1
> Dimensions and sizes: [100]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -1e+20
> (0) -7.442205e+17
>
> Running this script and omitting the line fval(32) = -1.0e20
> gives a value for avg(fint) of about 0.2, versus the value given here of
> about -1e17.
>
> NCL version:
> 5.0.0
> Output from uname -a:
> Linux nimbus.pmc.ucsc.edu 2.6.9-55.0.2.ELsmp #1 SMP Tue Jun 26 14:14:47 EDT
> 2007 x86_64 x86_64 x86_64 GNU/Linux"
>
> and on another machine:
>
> NCL version:
> 4.3.0
> Output from uname -a:
> Linux localhost.localdomain 2.6.25.4-10.fc8 #1 SMP Thu May 22 22:58:37 EDT
> 2008 x86_64 x86_64 x86_64 GNU/Linux
>
> Thank you,
> -Travis-
>
> --
> Travis A. O'Brien
> Graduate Student Researcher
> Earth and Planetary Science Dept.
> UC Santa Cruz
>
> tobrien_at_ucsc.edu
> (831) 459-3504
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk_at_ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Thu Jun 26 2008 - 16:00:46 MDT