Re: Problems reading NetCDF data - get NaNs

We have found a solution to this problem, which I am posting to the
list for future reference.  Thanks very much to John Caron and Don
Murray for helping out with this.

The problem was that the valid_min and valid_max were expressed as
shorts (i.e. packed data) but their numerical range was correct for
unpacked data:

       :valid_min = -3; // short
       :valid_max = 40; // short
       :add_offset = 20.0; // double
       :scale_factor = 0.0010; // double

They should have been:

       :valid_min = -23000; // short
       :valid_max = 20000; // short

i.e. relative to the packed data - this way an app can tell if the
data are out of range without unpacking the data.  This would also
have been OK:

       :valid_min = -3.0; // double
       :valid_max = 40.0; // double

The problem was causing nj22 to read all data as out of range.

Jon

On 19/10/06, Jon Blower <address@hidden> wrote:
> Hi Don,
>
> Thanks for your quick reply.  I switched to 2.2.17 and the problem is
> still there.  The values of temperature from ncdump include (file is
> large!):
>
> 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, -20974,
>     -20974, -20979, -20985, 32767, 32767, 32767, 32767, -21054, -21072,
>     -21083, -21093, -21101, -21108, -21112, -21115, 32767, 32767, 32767,
>
> 32767 represents missing data (metadata has :_FillValue = 32767;) and,
> for example, -21072 represents (-21072 * 0.001 + 20=) -1.072 degrees
> C, which is within the valid range.
>
> According to the file metadata, the convention is COARDS.  Also, as I
> said in my original post, ncBrowse displays the data without any
> problems.
>
> Thanks again,
> Jon
>
> On 19/10/06, Don Murray <address@hidden> wrote:
> > Hi Jon-
> >
> > I suspect the problem is with the valid_max and valid_min
> > attributes.  The library will return NaN if the
> > values are outside the range of the min/max.  If the
> > type of the attribute is the same as the variable, they
> > are compared before scale and offset.  If they are different
> > they are compared after scaling and offset.  I recently
> > found a bug with this that was fixed in the 2.2.17 pre-release,
> > but I'm not sure if that is why you are seeing what you
> > are seeing.
> >
> > What are the values for temperature in ncdump?
> >
> > Don
> > *************************************************************
> > Don Murray                               UCAR Unidata Program
> > address@hidden                        P.O. Box 3000
> > (303) 497-8628                              Boulder, CO 80307
> > http://www.unidata.ucar.edu/staff/donm
> > *************************************************************
> >
> >
> >
> > Jon Blower wrote:
> > > Dear all,
> > >
> > > I'm having some problems reading data from a particular NetCDF file.
> > > The code i'm using is:
> > >
> > >  NetcdfDataset nc =
> > > NetcdfDataset.openDataset("C:\\data\\OA_20060830.nc", true, null);
> > >  GridDataset gd = new GridDataset(nc);
> > >  GeoGrid gg = gd.findGridByName("temperature");
> > >  Array arr = gg.readYXData(0, 0);
> > >  IndexIterator it = arr.getIndexIteratorFast();
> > >  while (it.hasNext()) {
> > >      double val = it.getDoubleNext();
> > >      System.out.println("" + val);
> > >  }
> > >  nc.close();
> > >
> > > I just get a load of NaNs, even though I know that there are valid
> > > data in the file (ncBrowse displays the data perfectly).  The only
> > > unusual thing about the data is that the data are stored as short
> > > integers, with an offset and scale factor:
> > >
> > >     short temperature(time, depth, latitude, longitude);
> > >        :long_name = "Temperature";
> > >        :missing_value = 32767; // short
> > >        :_FillValue = 32767; // short
> > >        :units = "degree_Celsius";
> > >        :valid_min = -3; // short
> > >        :valid_max = 40; // short
> > >        :add_offset = 20.0; // double
> > >        :scale_factor = 0.0010; // double
> > >        :comment = "Temperature estimate (by objective analysis)";
> > >
> > > I understood that the NetcdfDataset class automatically dealt with
> > > offsets and scale factors  in "enhanced" mode, which is what I am
> > > using (or trying to use).  Hence I was expecting to "see" the data as
> > > an array of doubles.  The same happens if I use getFloatNext() or
> > > getShortNext().
> > >
> > > Can anyone see what I'm doing wrong?
>
> --
> --------------------------------------------------------------
> Dr Jon Blower              Tel: +44 118 378 5213 (direct line)
> Technical Director         Tel: +44 118 378 8741 (ESSC)
> Reading e-Science Centre   Fax: +44 118 378 6413
> ESSC                       Email: address@hidden
> University of Reading
> 3 Earley Gate
> Reading RG6 6AL, UK
> --------------------------------------------------------------


--
--------------------------------------------------------------
Dr Jon Blower              Tel: +44 118 378 5213 (direct line)
Technical Director         Tel: +44 118 378 8741 (ESSC)
Reading e-Science Centre   Fax: +44 118 378 6413
ESSC                       Email: address@hidden
University of Reading
3 Earley Gate
Reading RG6 6AL, UK
--------------------------------------------------------------

===============================================================================
To unsubscribe netcdf-java, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
===============================================================================