[Qgis-user] Best practice when using UK national grid and GPS

[Jim Jackson]

And there I was thinking I was beginning to understand this stuff and the 
first reply goes Whooosh into space over my head :-) But I'll try ...


On Fri, 1 May 2026, Greg Troxel via QGIS-User wrote:

> Jim Jackson via QGIS-User <qgis-user at lists.osgeo.org> writes:
> 
> > QGIS version  3.40.6-Bratislava
> > QGIS code branch Release 3.40
> 
> That's old, but probably not hurting you in this regard.
> 
> > As a member of a community Archaeology Group, I'm using QGIS for recording 
> > archaeological features in the UK. I'm using Google Satellite, 
> > OpenStreetMap and National Library of Scotland historic basemaps, which (as 
> > I understand it) use EPSG:3857.
> 
> You'd have to ask NLS but surely those historic maps are not in 3857,
> but something else, and they've transformed them for webmapping use.  In
> the US, the USGS has done such transforms.  But sub-meter and historic
> map don't go together.

Ok. I didn't set the CRS I just used the XYZ url they advertise in XYZ 
tiles and I get the maps and the CRS shown for the layer says EPSG:3857

> > I have an RTK GPS receiver for collecting data points which are in 
> > EPSG:4326. 

> Really?  That sounds fishy.  What provider are you connecting to that is
> sending reference data (via NTRIP and RTCM3 probably)?  

Yes yes, but the output to the app is WGS84 lat lon co-ordinates.

> How did they
> establish base station coordinates in 4326?  Do they understand that
> it's an ensemble, and that they are saying they don't know what datum
> their base coordinates are really in?

Ok so I load GPX data saved by my phone app using this external GPS 
receiver into QGIS and as if by magic it has a CRS of EPSG:4326 WGS84
I didn't do anything.

> If you set up your own base and did "survey in", you're probably pretty
> far off.

Nah used a nearby "centipede-rtk.org" base station.

> > I also get UK National Grid data points from a colleague who has
> > access to a total station.
> 
> You should ask your colleague to explain where their coordinates come
> from, as total stations don't do absolute positioning.  Lots of
> possibilities here.

Total station triangulation in from known UK OS grid points, then surveying
and recording UK OS grid points.

> Read this, and note that the grid is a projection from an underlying
> datum, OSGB36:
> 
>   https://en.wikipedia.org/wiki/Ordnance_Survey_National_Grid
> 
> I think you should be aiming to use "OSTN15" as a transform between
> various modern datums and OSGB36 (well, it's for one modern datum, but
> you will go through that datum from others).
> 
> > As I think I understand it, I need to set the project CRS to UK national 
> > grid EPSG:27700 to get sensible distance measurements However I then get 
> > occasional warnings "Used a ballpark transform from EPSG:27700 to 
> > EPSG:3857" indicated less than accurate transformations, and then I get 
> > lost!!!!
> 
> The measure tool should read in meters regardless. 3857 does not have a
> single scale from coordinates to distance, but the tool should work.

Ah but it doesn't. 

> If you are trying to measure distances between your RTK measurements and
> the total station measurements, you're likely going to have errors.
> 
> However, using 27700 as a project CRS is sensible.  UKNG coordinates
> should be close enough to meters - but use the measure tool.
> 
> > Can anyone point me to a tutorial on how to set up QGIS in the UK to get 
> > accurate transformations in these circumstances (which can't be that 
> > unusual).
> 
> There is no such thing as accurate data in 4326, if you care about a
> meter.  It is an ensemble with an intrinsic uncertainty of 2m and it
> makes zero sense to use it for RTK.  You will have to resolve that
> problem.
> 
> Probably, the coordinates in your RTK system are in either some recent
> ITRF or in some recent ETRF (<humor>but maybe the UK stopped using ETRF
> due to Brexit?</humor>).
> 
> proj, and hence qgis, will use a "ballpark transformation" that is
> typically low accuracy when one of the frames in the transform has
> similarly low intrinsic accuracy.
> 
> If your RTK data meaasured in April is really "ITRF2020 epoch 2026.3",
> then if you label it that way, and make sure you let proj load grids,
> you should end up with accurate UKNG/OSGB36 coordinates.
> 
> 
> I would recommend that you work with your total station colleague to
> take your RTK setup to an OS passive control and measure its location
> and see if you can get your values and the OS published values to line
> up.
> 
> Or, establish some passive controls at the edges of your site and have
> both of you measure each of them.
> 
> In the US (MA), I've measured points with RTK in "NAD83(2011) epoch
> 2010.0", our current national datum, and lined them up with imagery that
> has been ground controlled to on the order of 10 cm and they match well.
> So I'm pretty confident in my RTK-derived coordinates.
> _______________________________________________
> QGIS-User mailing list
> QGIS-User at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-user
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-user
>