[OSGeo-Discuss] RE: [vtp] 3d geoweb - Digital Earth - reference model

Ben Discoe ben at vterrain.org
Sat May 5 01:50:26 PDT 2007


Chris, 

> -----Original Message-----
> From: chris [dragonmagi at gmail.com]
> 
> I have been thinking we should take a different tact from 
> conventional approaches so far.
> We may now  be able to create what I call the Great Cartesian 
> Model of the Earth.

Choice of coordinate system (CRS) is actually largely unrelated to the
challenge of getting the various 3dgeo platforms to interoperate.  The vast
majority of existing data in the world is in normal geographic/projected
CRS.  Any existing and future software already has to deal with this fact
and accept all these CRS.  Putting some new data in a geocentric CRS does
not make the problem any simpler; it simply adds another CRS that we must
additionally accept.

> The idea is to directly calculate high 
> precision cartesian coordinates for the earth surface from 
> raw data to a certain level of detail, such as a 3 arc second 
> grid. I think the NASA scientists, using satellite 
> measurements and astronomical reference data, should be able 
> to do this.

You can already do it, pick any Datum and use geocentric if you like.  But
exceedingly little data is stored that way, any cartographer will tell you
PCS exist for good reasons!

> I.E. no going through intermediate ellipsoidal approximations 
> which introduce error both from the approximations and from 
> the extra calculation of converting to them (e.g. to lat, 
> lon, height), then projecting from the approximation to 
> cartesian again. We end up converting back to cartesian 
> coords when displaying graphically.

Many conversions are going on regardless of data storage CRS.  These
calculations can be shifted from one place to another in the data flow, but
not avoided unless we move to a flat planet.

> So, do you think  this is just a crazy idea or can we move 
> away now from the gross approximations and the additional 
> projection calculations permanently?

It's not a crazy idea to put data in another CRS.  It just may not be the
solution you are looking for. :)

-Ben




More information about the Discuss mailing list