[CF-metadata] What do models assume for the shape of the Earth?

From: John Caron <caron>
Date: Wed, 06 Apr 2005 13:00:51 -0600

toyoda at gfd-dennou.org wrote:

>Hello John and CF-metadata folks,
>
>--- John Caron <caron at unidata.ucar.edu>:
>
>
>>Normally I dont see information in the netCDF file
>>on the assumed shape
>>of the Earth, eg ellipsoidal flattening or even
>>spherical radius. Is
>>that because it doesnt matter that much? Or is there
>>an assumed standard
>>that everyone just knows about?
>>
>>
>
>I guess you would like to supply 'assumed shape of the
>Earth' information when converting netCDF datasets into
>GIS data format. If I am right, the most important thing
>should be how the coordinate information is used by GIS.
>
>GIS converts coordinate system when it overlays various
>kinds of data with different coordinate systems.
>
>For example, if we say that some meteorological data is
>based on coordinates on sphere, GIS will convert it into
>ellipsoid-based coordinates (in which GIS operation is
>going to be performed). Then the meteorological data is
>shifted equatorward.
>
>The amount of shift should be equivalent to the difference
>between geodetic and geocentric latitudes. According to
>Snyder's book, it is 11'40" (21km) at 45 degree latitude
>(in Clarke 1866 ellipsoid but not so much different in
>others for two or three-digit precision).
>
>
which Snyder book is this? which page ?

>In my knowledge, the reversed conversion (ellipsoid to
>sphere) has not been performed in data assimilation of
>numerical weather simulation. Meteorological community
>naively references location using latitude/longitude
>coordinates, ignoring ellipsoid by which the coordinates
>are defined (in short, people get observation and lat/lon,
>then put them into model at the same lat/lon). Thus the
>model data basically reflects observations on coordinates
>on ellipsoid, regardless the coordinate system the model
>use.
>
>
so observations are always using geodetic latitudes ?

>I believe such data should not be shifted so much as 20km.
>
>
>
>>Is there a different answer for global vs regional
>>models? At what
>>scale does this start to matter?
>>
>>
>
>Recent operational global models (even excluding the Earth
>Simulator) often have resolution near 20km. Thus the
>problem is not only for regional models.
>
>
>
thank you very much for this information.

for my purposes for now, i will just try to represent the earth shape as
accurately as possible. but i will be looking to clarify these issues
with the data that Unidata handles.
Received on Wed Apr 06 2005 - 13:00:51 BST