[Proj] Proj4 Datum Grids

[Martin Desruisseaux]

Hello Dan

Le 19/03/2018 à 23:15, Dan Crosby a écrit :

> Am I correct in thinking that, provided the relevant ellipse is
> defined in Proj, I can perform any datum shift with the appropriate
> +towgs84 entry?
>
It depends on the desired accuracy; transformations done with +towgs84
are only approximations. For example the accuracy is about 20 metres for
NAD27 to WGS84 datum shift in Canada. For determining the operation
accuracy (if the library does not provide this information by itself),
one can consult http://epsg-registry.org/ (e.g. enter "NAD27 to WGS 84"
in the name field).

For better accuracy, datum shift grids are required. With those grids,
an accuracy closer to 1 metre can be achieved. Note however that many
datum shift grids are defined between two national datum (for example
from NAD27 to NAD83), not necessarily to WGS 84. The national datum are
often related to a continental plate while WGS 84 is related to Earth
center; those two kinds of datum do not move in the same way. But a
transformation to WGS 84 may nevertheless be performed on the assumption
that one of the national datum is close enough to WGS 84. One may need
to chain two grids however (e.g. NAD27 to NAD83 followed by NAD83 to
NAD83(HARN) on the assumption that the later is equivalent to WGS 84
within the accuracy of the transformation).

Note that not all datum shifts can be relative to WGS 84. For example in
Martinique (for transformation from 1938 system), any datum shifts going
through WGS 84 will put errors; this transformation can not be described
accurately by +towgs84. Note also that there is no longer a single WGS
84 system, but currently 6 versions of WGS 84 with a difference up to
about 1~2 metres between them, which make the use of +towgs84 ambiguous.

Whether you need to care about the above or not depend on the desired
accuracy. An issue is that current Proj.4 version gives no hint about
the accuracy of its operation, but I believe that providing this kind of
information is on the roadmap for future versions.

    Martin


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20180320/a1f79bea/attachment.html>