[Qgis-user] EPSG:27700 - OSGB 1936 / British National Grid shapefiles are incorrectly interpreted by QGIS 1.8.0 and later

[Andre Joost]

As long as *all* your project and data is in EPSG:27700, you will not 
see any difference. But as soon as you add Google or openstreetmap tiles 
with openlayersplugin, or data from other sources in other CRS, the 
+towgs84 datum shift parameters comes into effect. The values used up to 
QGIS 1.7.4 were quiet good, QGIS 1.8.0 sadly returned to some 
low-accuracy parameters. For best accuracy, OS has issued a ntv2 grid file.

The .prj file by ESRI does not have any information on datum shift 
values incorporated. Mainly because there are several choices to do the 
transformation with different accuracy and complexity, as stated above.

HTH,
André Joost


Am 31.01.2013 14:25, schrieb Andrew Chapman:
> Pardon my ignorance here - I'm not (nor do I wish to become) a GIS expert
> with detailed knowledge of all the ins and outs of CRSs. While I can
> understand that to some users it could be critical to know with ultimate
> accuracy the location of any specified point as a world coordinate, what
> ultimately matters to me is to be able to agree with other users that we
> consistently interpret our mapping data in the same way - point A is always
> in the same place on each of our systems, subject obviously to specifying
> the CRS.
>
> Using shapefiles, as I understand it, the .prj file specifies how everything
> should line up. For OSGB 1936 this has historically been
>
> PROJCS["OSGB_1936_British_National_Grid",GEOGCS["GCS_OSGB
> 1936",DATUM["D_OSGB_1936",SPHEROID["Airy_1830",6377563.396,299.3249646]],PRI
> MEM["Greenwich",0],UNIT["Degree",0.017453292519943295]],PROJECTION["Transver
> se_Mercator"],PARAMETER["latitude_of_origin",49],PARAMETER["central_meridian
> ",-2],PARAMETER["scale_factor",0.9996012717],PARAMETER["false_easting",40000
> 0],PARAMETER["false_northing",-100000],UNIT["Meter",1]]
>
> If I get layers from Ordnance Survey, my county council GIS department or
> other town and parish councils, I would expect my example point A to stay in
> the same place whatever GIS software I use.
>
> I'm not sure what is being suggested here as a fix. Is it going back to the
> pre 1.8.0 solution or providing an alternative CRS that the user will need
> to select for all affected layers in every project? If the latter it will
> potentially cause problems for those of us sharing projects between machines
> using different version of QGIS.
>
> Sorry - I'm just having trouble getting my head around this!
>
> Andrew Chapman
>
> -----Original Message-----
> From: Even Rouault [mailto:even.rouault at mines-paris.org]
> Sent: 31 January 2013 12:24
> To: Andrew Chapman
> Cc: qgis-user at lists.osgeo.org
> Subject: Re: [Qgis-user] EPSG:27700 - OSGB 1936 / British National Grid
> shapefiles are incorrectly interpreted by QGIS 1.8.0 and later
>
> Selon Andrew Chapman<andrew.chapman at donkagen.co.uk>:
>
>
> There's a GDAL ticket about that : http://trac.osgeo.org/gdal/ticket/4597
>
> Frank, I've attached in it a libgeotiff patch that should solve this by
> defining an override.