[GRASS-user] Questions about datum management

[Paul Kelly]

On Fri, 24 Oct 2008, Corrado wrote:

> Dear friends,
>
> a few questions on some things I do not understand well about datum:
>
> 1) what is a datum tranformation parameter?

An extra calculation is needed when re-projecting maps and data between 
two different co-ordinate systems (in addition to the re-projection) if 
the two co-ordinate systems are based on different models of the 
curvature of the earth. E.g. OSGB36 uses the Airy ellipsoid and WGS84 uses 
the WGS84 ellipsoid, which have slightly different sizes and shapes. The 
error is not large - generally a few hundred metres at most on the ground. 
The datum transformation parameters describe this adjustment 
mathematically.

> 2) why there are 2 datum transformation paramenters for osgb36?

No particular reason - one of them has been in GRASS for a long time and 
one is a little bit more accurate and officially published by the OSGB on 
their website. Several countries have more sets of parameters than this - 
as the transformation between any two datums is approximate and varies by 
location, different sets of parameters are often offered to give improved 
accuracy in different regions of a country. In general there is no one 
"correct" set of transformation parameters - indeed the accuracy changes 
over time due to tectonic movements in the earth (an answer that you will 
get if you ask about this on the PROJ.4 mailing list!). OSGB is actually a 
special case (though not treated as such in GRASS) - see below.

> 3) why does the system ask for a datum transformation parameter when I create
> a location with EPSG code 27700? The code should be enough to create
> everything!

Why do you say that? In general, different datum transformation parameters 
are appropriate according to the region your data is in and the accuracy 
you require in a reprojection. However, OSGB36 is a special case as it has 
recently (in the least few years) been redefined by the Ordnance Survey of 
Great Britain in terms of WGS84 rather than the original 1936 definition:
http://www.ordnancesurvey.co.uk/oswebsite/gps/information/coordinatesystemsinfo/guidecontents/guide5.html
GRASS doesn't (yet) support the OSTN02 gridshift transformation defined 
there though. (EPSG code 27700 doesn't specify it either; as far as I'm 
aware the EPSG database has no way of specifying gridshift-based datum 
transformations.) But you can get a about 5m reprojection accuracy 
(depending on the area) using the 7-parameter OSGB36 transform provided in 
GRASS.

[Note to self: 5 metres is quite a low accuracy really compared to that 
available directly in GRASS for other European countries. We should look 
at converting OSTN02 into the NTv2 format so it can be used with GRASS and 
PROJ.4 to give improved accuracy for users in Great Britain.]

Of course if you don't want to reproject data between datums then you 
don't need to worry about the datum transformation settings when creating 
a new location.

Paul