Re: newbie, netCDF/isocontour/SpreadSheet ?'s: packed netCDF data

• To: Steve Emmerson <steve@xxxxxxxxxxxxxxxx>
• Subject: Re: newbie, netCDF/isocontour/SpreadSheet ?'s: packed netCDF data
• From: Roland Schweitzer <rhs@xxxxxxxxxxxx>
• Date: Tue, 08 Jun 1999 14:40:47 -0500

Steve,

Thanks for your reply (and to Tom Rink for his comments).

Steve Emmerson wrote:

> I wrote the netCDF import module for VisAD.

Thanks, it's quite nice.

I thought I would reply "on list" since I have a couple of other things I'd like
to throw out for possible discussion.

The netCDF files we work with (anything in ftp://ftp.cdc.noaa.gov/Datasets for
example) do use the variable attributes "scale_factor" and "add_offset" to
define the packing/unpacking.  The files are not HDF-EOS compatible, but follow
the COARDS convention
(http://www.unidata.ucar.edu/packages/netcdf/conventions.html).

There are a couple of other details about the netCDF import module that might
come into play when I get further into importing our netCDF files.  First of all
we use two variable attributes to describe the data range for a variable,
"actual_range" and "valid_range".  The valid range is intended to describe the
range of values possible for that particular parameter from any file in any data
collection we support while actual range is the range for that particular file. 
If I'm guessing correctly, valid range is used for the slider that selects the
isocontour value.  That works perfectly for our data files.  However, I think we
may do something a little unusual in the netCDF world in that we store the valid
range and actual range as the UNpacked data type so humans can read it from an
ncdump.  (Outside users have told us that this was not what they expected for
those attributes and their software thought these attributes should contain
packed values.)  I don't know how this is normally handled, but thought it was
important to point out.

The second thing I noticed has to do with translation of the time coordinate. 
In our local software we use UDUNITS to translate the time values of "days
since..." or "hours since...".  The times represented in my test file are daily
1992-08-02 through 1992-08-04, but show up as 2001-1-1 for all three times. 
This is not critical right now, but more of a curiosity.
 
> The netCDF import module doesn't "understand" packed data at this time
> (i.e. it doesn't use the netCDF variable attributes "scale_factor" and
> "add_offset").  I intend to add that capbility, however, and your need
> will make it sooner rather than later.
>
> What's your time-frame?

I am doing some preliminary explorations of using VisAD right now.  I hope to
begin working on this in earnest in the late summer to early fall and will
continue to pursue it beyond that probably for the next year or so.

Sorry if this is too far "off-topic" for the list, but I hope other netCDF users
might find it useful.

Thanks again,
Roland

-- 
Roland Schweitzer
NOAA-CIRES Climate Diagnostics Center           325 Broadway
NOAA/ERL/CDC - (R/E/CD1)                        Boulder, CO 80303
   .... (303) 497-6249 .... (303) 497-7013 FAX .... rhs@xxxxxxxxxxxx ....

• References:
  • Re: newbie, netCDF/isocontour/SpreadSheet ?'s: packed netCDF data
    • From: Steve Emmerson