[Proj] How to convert a sphere to ellipsoid with correct datum?

Jan Hartmann j.l.h.hartmann at uva.nl
Sat Sep 4 07:15:32 EST 2010

On 09/03/10 18:22, Hermann Peifer wrote:
> I think this story with the Dutch church towers already circulated 
> through the proj mailing list, in 2008, as far as I can see:
> http://osgeo-org.1803224.n2.nabble.com/template/TplServlet.jtp?tpl=search-page&node=2062109&query=Dutch+church+towers 
> <http://osgeo-org.1803224.n2.nabble.com/template/TplServlet.jtp?tpl=search-page&node=2062109&query=Dutch+church+towers> 
> I see that you started it, so it seems to be an ongoing issue? I am 
> sorry, but I do not have any special expert knowledge in this specific 
> case.
Thanks Hermann, that was really great help. To explain why I keep coming 
back on the issue: I am georeferencing the oldest cadastral maps of the 
Netherland from 1830. There was no nation-wide triangulation then, but 
for the zero point of the local grids the surveyors chose in most cases 
the same church towers that were twenty years later used for the 
national triangulation. If I can put that point exactly on the modern 
map, I can also put the whole cadastral map of a particular community on 
the modern cadastral map, with accurucies up to a few meters.

That accuracy is important, since there are lots of legal issues about 
old buildings, parcels and landscapes that influence their price and 
what you are legally allowed to do with them. A difference of 50 meters 
in  a very parcellated city like Amsterdam can have severe consequences 
for prices, subsidies, "protectability" etc and could affect lots of 
real people at this actual moment.

So the 60 meters accuracy I can get by just using the conventional PROJ 
parameters in transforming from the 19th to the 20th century system just 
isn't enough. I can transfer the zero point of the 19th century 
triangulation by 60 m., but as long as I can't underpin that in a 
rigorous mathematical way, the results will always be disputable. I 
really expect to open a big can of worms when people overlay their 
historic properties over Google satellite with an apparent meter-like 
accuracy, that turns out to be disputable when they have a closer look 
at the methods I have used.

Again thanks for your help: I am now confident that converting latlon 
values on different ellipsoids give correct results when using 
+towgs84=0,0,0 to both parameter lists. I still have to hunt for the 
last 60 meters of accuracy, though. The devil is, as always, in the 
latest en finest detail.


More information about the Proj mailing list