[Proj] Re: What proj library version is best?

Clifford J Mugnier cjmce at lsu.edu
Wed Jul 7 17:41:07 EDT 2004






Gerald Evenden wrote:
> This discussion reminds me of the email interaction Frank Warmerdam and I
> had about what should be included in a projection library.  My position
was,
> and still is, that *only* the projection (conversion between geographic
and
> cartesian space) is the domain of the library.  All non-projection
elements like
> datum shifts and  x-y-z transformations are separate and unrelated
problems
> and belong in their own library systems.
>
> The reason for this is twofold: 1) projections have uses that do not
involve
> datums and their is no reason to burden these uses with elements that are
> not involved and 2) developers often only have expertise in only one of
the
> areas and should not have to burden themselves with library maintenance
> involving unfamiliar elements.
[...]

I don't know how to interpret this : Datum is an important part of
coordinate's
transformation.  The same projection on two diferent datums can lead to
errors
of 500 m (at least in France, between the NTF and WGS84 Datum).  Projection
and Datum transformation are two different steps (and hence may be done in
different functions), but you cannot forget Datum transformation.  For
exemple,
the correct way to tranform UTM/WGS84 to NTF/Lambert coordinate is :

(E,N) UTM -> lat/long WGS84 -> X,Y,Z (tridimentional) -> Datum shift,
rotation
and scale from WGS84 to NTF -> lat/long NTF -> (E,N) Lambert.

Even with this formula you have an accuracy of 5 m in some places : You
have to
use a correction matrix to improve the precision to a geodetic one (5 cm).

You can only forget datum when you work at very small scale (1:2,000,000
and under)

--
Michel Wurtz - Auzeville-Tolosane

------------------------------------

I believe the major feature that separates the philosophies is spherical
vs. ellipsoidal.  Datums are immaterial with spherical applications.
However, if ellipsoids are important, than so are classical/inertial datums
important.

As to whether ellipsoidal projections and concomitant grids should be
separated from datum shift techniques - that's like listening to cowboys
argue about the best manufacturer of pickup trucks!

The new 5th edition of the "Manual of Photogrammetry" is being introduced
at the International Society for Photogrammetry and Remote Sensing meeting
in Istanbul later this month.  I cover much of the above subject matter in
Chapter 3 under "Object Space Coordinate Systems."

Cliff Mugnier
LOUISIANA STATE UNIVERSITY




More information about the Proj mailing list