Re: [netcdf-java] [CF-metadata] ESRI implementation of ellipsoidal-earth

  • To: David Blodgett <dblodgett@xxxxxxxx>
  • Subject: Re: [netcdf-java] [CF-metadata] ESRI implementation of ellipsoidal-earth
  • From: Heiko Klein <Heiko.Klein@xxxxxx>
  • Date: Thu, 03 Feb 2011 09:54:44 +0100
I can confirm Davids observation. The European atmospheric models I'm working with use also unchanged WGS84 (geodetic) coordinates with a spherical earth.

The proj4 projection library (starting with 4.5, I think) tries to solve this in the following way (see http://osgeo-org.1803224.n2.nabble.com/How-to-convert-a-sphere-to-ellipsoid-with-correct-datum-td5486117.html ):

* if input and output projection both have a DATUM set (besides the ellipsoid) a geodetic/geocentric conversion is done. * if the datum is missing on any projection (input or output) no geodetic/geocentric conversion is made

As far as I've seen, the netcdf-java library 4.2 has no code for geodetic/geocentric conversion, so it follows the models. I don't even know if there is a way in newer CF-versions to set the Datum, besides the ellipsoid?

Heiko

On 2011-02-02 14:19, David Blodgett wrote:
All, I meant to send the following to the list yesterday but I don't think it 
came through... Hopefully this one makes it through.

New to the list, Rich Signell asked me if I'd respond to John's question below. 
For what its worth:

I've done a fair amount of thinking and asked a lot of questions re:

what to use for the default ellipsoid when
these are not specified. It seems to me the obvious candidates are 1)
spherical earth with a standard radius, or 2) WGS84 ellipsoid.

I have come to the conclusion that WGS84 is probably the best default. Although 
really, without a datum definition a dataset is incomplete (according to 
geographic information science) and an error should be thrown.

My reasoning for this is that I have not been able to find an atmospheric 
scientist who has converted ground based model forcing data locations into the 
spherical datum used by their model grid. This basically means that the grid's 
geometry calculations are out of sync with the actual location of the forcings. 
So the eventual model output or reanalysis or whatever, is actually most 
similar, location wise, to a geodetic datum (like WGS84)

However,

Things get complicated with projected coordinate systems based on spherical 
datums. The issue there is that to properly unproject, you need to use the math 
associated with the spherical datum. You then have lat/lon positions with 
spherical datum indicated in its metadata. But the original data was likely not 
converted to the spherical datum but rather reassigned, so you have to change 
the metadata (hack!) without transforming to the new, correct datum. We deal 
with this by using 2D axis definitions and indicating that the lat/lon is on 
whatever datum we think it came from.

Hopefully that is helpful,

p.s. Welcome back John!

Dave

On Feb 2, 2011, at 7:01 AM, Rich Signell wrote:

John&  Co,

Also, they are wondering what to use for the default ellipsoid when these
are not specified. It seems to me the obvious candidates are 1) spherical
earth with a standard radius, or 2) WGS84 ellipsoid.

I imagine the vast majority of netcdf datasets fall into this camp.  I
had an interesting talk yesterday with Dave Blodgett, who pointed out
that even though most models are running on a spherical earth, they
mostly likely do not transform the initial bathymetry/topography,
coastline, and other model inputs into spherical earth coordinates,
but just leave them in their original datum, which is usually WGS84.
So it seems that your 2) would be more likely to be right.

And just to make sure that CFers are on the same page, if the datum is
actually known to be different, it can be specified using the CF datum
parameters ("inverse_flattening" and "earth_radius", or
"semi_major_axis" and "semi_minor_axis"  in CF 1.5
(http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.5/apf.html) where
the doc now states: "The attributes which describe the ellipsoid and
prime meridian may be included, when applicable, with any grid
mapping."

-Rich

On Tue, Feb 1, 2011 at 2:40 PM, John Caron<caron@xxxxxxxxxxxxxxxx>  wrote:
Apparently ESRI is willing to add support for CF 1.5 grid_mapping attributes
for ellipsoidal-earth/geodetic-datum definitions in the Grid Mappings and
Projections specification.

They are looking for sample data, especially using ellipsoidal parameters
(semi_major_axis/semi_minor_axis/inverse_flattening). Does anyone have any?

Also, they are wondering what to use for the default ellipsoid when these
are not specified. It seems to me the obvious candidates are 1) spherical
earth with a standard radius, or 2) WGS84 ellipsoid.

Opinions?

John

_______________________________________________
CF-metadata mailing list
CF-metadata@xxxxxxxxxxxx
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata




--
Dr. Richard P. Signell   (508) 457-2229
USGS, 384 Woods Hole Rd.
Woods Hole, MA 02543-1598

_______________________________________________
CF-metadata mailing list
CF-metadata@xxxxxxxxxxxx
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata

--
Dr. Heiko Klein                              Tel. + 47 22 96 32 58
Development Section / IT Department          Fax. + 47 22 69 63 55
Norwegian Meteorological Institute           http://www.met.no
P.O. Box 43 Blindern  0313 Oslo NORWAY



  • 2011 messages navigation, sorted by:
    1. Thread
    2. Subject
    3. Author
    4. Date
    5. ↑ Table Of Contents
  • Search the netcdf-java archives: