Quantcast

OSGB coordinates to WGS84 lat/lon problem

classic Classic list List threaded Threaded
8 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

OSGB coordinates to WGS84 lat/lon problem

Nick Hopton
A chap on QGIS at Stack Exchange is having problems with accuracy when transforming from Ordnance Survey grid coordinates to WGS84 lat/lon. He's finding discrepancies in the order of seven metres.

Using QGIS 1.7.4, OSGEO4W under Vista I was able to reproduce the problem, but using QGIS Lisboa RC1 under Ubuntu 11.10 produced transformations good to better than a metre. My first reaction was that there must problems with the QGIS 1.7.4 Proj4 parameter string for EPSG:27700, but on checking I found that 1.7.4 and Lisboa use identical strings for this CRS.

I'm now at a bit of a loss to understand how following an identical procedure on both systems can produce differing results.

If anyone would care to check this, below is the csv file I used for my tests.

Station,x,y
Inkpen,437346.0,161624.8
Hodmore,468296.3,178184.9
Hindhead,489984.6,135909.7

Under Ubuntu these coordinates transformed to:
Inkpen 437346 161624.8 -1.465102 51.352478
Hodmore 468296.3 178184.9 -1.017527 51.498487
Hindhead 489984.6 135909.7 -0.715814 51.115423

Under Vista they transformed to:
Inkpen 437346 161624.8 -1.46511 51.352424
Hodmore 468296.3 178184.9 -1.017525 51.498435
Hindhead 489984.6 135909.7 -0.715808 51.115362

In my part of the world the differences amount to about six metres on the ground. The Inkpen test point is in fact an old triangulation pillar and the point as transformed by Ubuntu sits exactly on this when viewed in Goorgle Earth.

Any ideas?

Nick.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: OSGB coordinates to WGS84 lat/lon problem

Micha Silver
I;m getting numbers identical to your Ubuntu results in three tests:
QGIS 1.7.4 on Win7, QGIS1.7.4 on Scientific Linux, and also using cs2cs. All the same, and all = to your Ubuntu values.
Maybe that will help to smoke out the problem?

On 10/05/2012 13:13, Nick Hopton wrote:
A chap on QGIS at Stack Exchange is having problems with accuracy when
transforming from Ordnance Survey grid coordinates to WGS84 lat/lon. He's
finding discrepancies in the order of seven metres.

Using QGIS 1.7.4, OSGEO4W under Vista I was able to reproduce the problem,
but using QGIS Lisboa RC1 under Ubuntu 11.10 produced transformations good
to better than a metre. My first reaction was that there must problems with
the QGIS 1.7.4 Proj4 parameter string for EPSG:27700, but on checking I
found that 1.7.4 and Lisboa use identical strings for this CRS.

I'm now at a bit of a loss to understand how following an identical
procedure on both systems can produce differing results.

If anyone would care to check this, below is the csv file I used for my
tests.

Station,x,y
Inkpen,437346.0,161624.8
Hodmore,468296.3,178184.9
Hindhead,489984.6,135909.7

Under Ubuntu these coordinates transformed to:
Inkpen	437346	161624.8	-1.465102	51.352478
Hodmore	468296.3	178184.9	-1.017527	51.498487
Hindhead	489984.6	135909.7	-0.715814	51.115423

Under Vista they transformed to:
Inkpen	437346	161624.8	-1.46511	51.352424
Hodmore	468296.3	178184.9	-1.017525	51.498435
Hindhead	489984.6	135909.7	-0.715808	51.115362

In my part of the world the differences amount to about six metres on the
ground. The Inkpen test point is in fact an old triangulation pillar and the
point as transformed by Ubuntu sits exactly on this when viewed in Goorgle
Earth. 

Any ideas?

Nick.


--
View this message in context: http://osgeo-org.1560.n6.nabble.com/OSGB-coordinates-to-WGS84-lat-lon-problem-tp4965339.html
Sent from the Quantum GIS - User mailing list archive at Nabble.com.
_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user

This mail was received via Mail-SeCure System.




-- 
Micha Silver
052-3665918

_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

RE: OSGB coordinates to WGS84 lat/lon problem

miked
In reply to this post by Nick Hopton
7m error often associated with the approx. transformation that consumer grade gps units (or many of the coordinate conversion websites) carry out compared to the full OSTN02 definitive transformation.  If possible should always use the OSTN02 transformation.

Below are a few examples of points in wgs84 and osgb format transformed with OSTN02 if you want to check some others

Point ID Latitude Longitude Ellipsoid Height Northing Easting Orth. height Time
RTCM-Ref 0000 52.0242012 -0.7717783 139.488 236908.8906 484368.9928 92.26152 31/05/2011 14:44
RTCM-Ref 0014 51.9015014 -0.1784953 158.652 224117.4848 525415.3853 112.47864 31/05/2011 14:38
RTCM-Ref 0019 52.25181 -2.19794 87.084 261532.2794 386583.8284 37.4539 14/06/2010 09:30
1 52.0652051 -2.6857082 99.02443 240980.03 353090 48.15 13/04/2012 15:16


-----Original Message-----
From: Nick Hopton [mailto:[hidden email]]
Sent: 10 May 2012 11:13
To: [hidden email]
Subject: [Qgis-user] OSGB coordinates to WGS84 lat/lon problem

A chap on QGIS at Stack Exchange is having problems with accuracy when transforming from Ordnance Survey grid coordinates to WGS84 lat/lon. He's finding discrepancies in the order of seven metres.

Using QGIS 1.7.4, OSGEO4W under Vista I was able to reproduce the problem, but using QGIS Lisboa RC1 under Ubuntu 11.10 produced transformations good to better than a metre. My first reaction was that there must problems with the QGIS 1.7.4 Proj4 parameter string for EPSG:27700, but on checking I found that 1.7.4 and Lisboa use identical strings for this CRS.

I'm now at a bit of a loss to understand how following an identical procedure on both systems can produce differing results.

If anyone would care to check this, below is the csv file I used for my tests.

Station,x,y
Inkpen,437346.0,161624.8
Hodmore,468296.3,178184.9
Hindhead,489984.6,135909.7

Under Ubuntu these coordinates transformed to:
Inkpen 437346 161624.8 -1.465102 51.352478
Hodmore 468296.3 178184.9 -1.017527 51.498487
Hindhead 489984.6 135909.7 -0.715814 51.115423

Under Vista they transformed to:
Inkpen 437346 161624.8 -1.46511 51.352424
Hodmore 468296.3 178184.9 -1.017525 51.498435
Hindhead 489984.6 135909.7 -0.715808 51.115362

In my part of the world the differences amount to about six metres on the ground. The Inkpen test point is in fact an old triangulation pillar and the point as transformed by Ubuntu sits exactly on this when viewed in Goorgle Earth.

Any ideas?

Nick.


--
View this message in context: http://osgeo-org.1560.n6.nabble.com/OSGB-coordinates-to-WGS84-lat-lon-problem-tp4965339.html
Sent from the Quantum GIS - User mailing list archive at Nabble.com.
_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user

--
The Open University is incorporated by Royal Charter (RC 000391), an exempt charity in England & Wales and a charity registered in Scotland (SC 038302).

_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: OSGB coordinates to WGS84 lat/lon problem

Nick Hopton
In reply to this post by Micha Silver
Micha Silver wrote
I;m getting numbers identical to your Ubuntu results in three tests:
    QGIS 1.7.4 on Win7, QGIS1.7.4 on Scientific Linux, and also using
    cs2cs. All the same, and all = to your Ubuntu values.
    Maybe that will help to smoke out the problem?
Thanks for this Micha, though it's left me flummoxed for want of an explanation. The person on SE and I both have the same problem, but if you can't reproduce it then it's obviously not universal.

For miked, thanks for the comment, I know about this problem which is why I used as my test points stations from the OS passive stations database whose positions are known accurately.

Nick.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: OSGB coordinates to WGS84 lat/lon problem

Andy Harfoot
Hi Nick,

Have you tried the convertor here:
http://gps.ordnancesurvey.co.uk/convert.asp
to see whether the higher accuracy coordinates obtained from QGIS match
those obtained using OSTN02?

It looks as though Proj4 can use grid based transformations like OSTN02
through the use of gsb files. Could it be that this is the case in the
higher accuracy situation? I find gsb files at the following location on
my Win7 QGIS 1.7.4 install:
C:\OSGeo4W\share\proj

I can't see any files in this directory suggesting they are the OSTN02
grid though, and I can reproduce your 'Vista' results as follows,
suggesting that my QGIS is using the lower accuracy seven parameter
transform.

The OSTN02 gsb can be downloaded here:
http://www.ordnancesurvey.co.uk/oswebsite/gps/osnetfreeservices/furtherinfo/ostn02_ntv2.html
Referencing it in Proj4 is described here:
http://trac.osgeo.org/proj/wiki/GenParms#nadgrids-GridBasedDatumAdjustments

BUT I haven't tried this!

Andy

Nick Hopton wrote:

> Micha Silver wrote
>  
>> I;m getting numbers identical to your Ubuntu results in three tests:
>>     QGIS 1.7.4 on Win7, QGIS1.7.4 on Scientific Linux, and also using
>>     cs2cs. All the same, and all = to your Ubuntu values.
>>     Maybe that will help to smoke out the problem?
>>
>>    
>
> Thanks for this Micha, though it's left me flummoxed for want of an
> explanation. The person on SE and I both have the same problem, but if you
> can't reproduce it then it's obviously not universal.
>
> For miked, thanks for the comment, I know about this problem which is why I
> used as my test points stations from the OS passive stations database whose
> positions are known accurately.
>
> Nick.
>
>
> --
> View this message in context: http://osgeo-org.1560.n6.nabble.com/OSGB-coordinates-to-WGS84-lat-lon-problem-tp4965339p4965857.html
> Sent from the Quantum GIS - User mailing list archive at Nabble.com.
> _______________________________________________
> Qgis-user mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/qgis-user
>  

_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: OSGB coordinates to WGS84 lat/lon problem

Nick Hopton
Andy Harfoot wrote
Have you tried the convertor here:
http://gps.ordnancesurvey.co.uk/convert.asp
to see whether the higher accuracy coordinates obtained from QGIS match
those obtained using OSTN02?
For the record, the Survey gives the following (ETRS89) lat/lon coordinates for the points:

Inkpen: 51.3524757 -1.4650872
Hodmore: 51.4984893 -1.0170283
Hindhead: 51.1154179 -0.7157938

I think ETRS89 has drifted about half a metre away from WGS84 at the present time, but even so the transformations done by Lisboa under Ubuntu look very good.

It looks as though Proj4 can use grid based transformations like OSTN02
through the use of gsb files. Could it be that this is the case in the
higher accuracy situation? I find gsb files at the following location on
my Win7 QGIS 1.7.4 install:
C:\OSGeo4W\share\proj
At present by default QGIS doesn't use a grid-shift file for EPSG:27700. The datum-shift is carried out by the +towgs84 part of the Proj parameter string.

I can't see any files in this directory suggesting they are the OSTN02
grid though, and I can reproduce your 'Vista' results as follows,
suggesting that my QGIS is using the lower accuracy seven parameter
transform.
It's not using a grid-shift file. But it's interesting that you get the same poor transformations that I get with 1.7.4 under Visa. Lisboa RC1 under Ubuntu produces much better transformations (for me) but I don't know why this is the case.

Nick.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: OSGB coordinates to WGS84 lat/lon problem

Etienne Tourigny-3
datum shift transformations have been fixed in qgis-1.8, which may
explain this.  This has been addressed in this list (and also
qgis-dev) in the last months

Etienne

On Thu, May 10, 2012 at 4:04 PM, Nick Hopton <[hidden email]> wrote:

>
> Andy Harfoot wrote
>>
>> Have you tried the convertor here:
>> http://gps.ordnancesurvey.co.uk/convert.asp
>> to see whether the higher accuracy coordinates obtained from QGIS match
>> those obtained using OSTN02?
>>
>
> For the record, the Survey gives the following (ETRS89) lat/lon coordinates
> for the points:
>
> Inkpen: 51.3524757 -1.4650872
> Hodmore: 51.4984893 -1.0170283
> Hindhead: 51.1154179 -0.7157938
>
> I think ETRS89 has drifted about half a metre away from WGS84 at the present
> time, but even so the transformations done by Lisboa under Ubuntu look very
> good.
>
>
>
>> It looks as though Proj4 can use grid based transformations like OSTN02
>> through the use of gsb files. Could it be that this is the case in the
>> higher accuracy situation? I find gsb files at the following location on
>> my Win7 QGIS 1.7.4 install:
>> C:\OSGeo4W\share\proj
>>
>
> At present by default QGIS doesn't use a grid-shift file for EPSG:27700. The
> datum-shift is carried out by the +towgs84 part of the Proj parameter
> string.
>
>
>
>> I can't see any files in this directory suggesting they are the OSTN02
>> grid though, and I can reproduce your 'Vista' results as follows,
>> suggesting that my QGIS is using the lower accuracy seven parameter
>> transform.
>>
>
> It's not using a grid-shift file. But it's interesting that you get the same
> poor transformations that I get with 1.7.4 under Visa. Lisboa RC1 under
> Ubuntu produces much better transformations (for me) but I don't know why
> this is the case.
>
> Nick.
>
>
> --
> View this message in context: http://osgeo-org.1560.n6.nabble.com/OSGB-coordinates-to-WGS84-lat-lon-problem-tp4965339p4966416.html
> Sent from the Quantum GIS - User mailing list archive at Nabble.com.
> _______________________________________________
> Qgis-user mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/qgis-user
_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

FW: Re: OSGB coordinates to WGS84 lat/lon problem

miked
I have tried with plain .csv files of points or shapefiles and both give the error in matching up OSGB and WGS84 coordinates for the same point. Basically the points are shown several metres apart when you allow on the fly transformation, error on all the same amount and in same direction.

Qgis says it has the correct CRS for the shapefiles.


-----Original Message-----
From: Etienne Tourigny [mailto:[hidden email]]
Sent: 13 July 2012 12:46
To: M.E.Dodd
Cc: Nick Hopton
Subject: Re: [Qgis-user] Re: OSGB coordinates to WGS84 lat/lon problem

you should probably send this to the mailing list.

You should neot be seing this problem with shapefiles in qgis-1.8.
Have you tried it?  What CRS does QGis set for the files?

cheers
Etienne

On Fri, Jul 13, 2012 at 6:53 AM, M.E.Dodd <[hidden email]> wrote:

> Has this issue been resolved as I still have difference between where points are shown using the wgs84 and osgb coordinates.  The coordinates for the points have been converted between the two systems using the ordnance survey converter with full accuracy but they appear about 4m away from each other when shown in qgis 1.9 (version from a couple of days ago) on windowsXP.  This is having two layers one in EPSG:4326 - WGS 84 and the other in EPSG:27700 - OSGB 1936 / British National Grid and allowing on the fly CRS transformation.
> I am sure in the past this was not an issue but is potentially causing significant problems now, not least when trying to argue that qgis should be taught instead of other gis systems and people say that its useless because it has problems like this (this is currently a debate in my department where we are trying to decide which system to go with).
>
> -----Original Message-----
> From: Etienne Tourigny [mailto:[hidden email]]
> Sent: 10 May 2012 23:59
> To: Nick Hopton
> Cc: [hidden email]
> Subject: Re: [Qgis-user] Re: OSGB coordinates to WGS84 lat/lon problem
>
> datum shift transformations have been fixed in qgis-1.8, which may
> explain this.  This has been addressed in this list (and also
> qgis-dev) in the last months
>
> Etienne
>
> On Thu, May 10, 2012 at 4:04 PM, Nick Hopton <[hidden email]> wrote:
>>
>> Andy Harfoot wrote
>>>
>>> Have you tried the convertor here:
>>> http://gps.ordnancesurvey.co.uk/convert.asp
>>> to see whether the higher accuracy coordinates obtained from QGIS
>>> match those obtained using OSTN02?
>>>
>>
>> For the record, the Survey gives the following (ETRS89) lat/lon
>> coordinates for the points:
>>
>> Inkpen: 51.3524757 -1.4650872
>> Hodmore: 51.4984893 -1.0170283
>> Hindhead: 51.1154179 -0.7157938
>>
>> I think ETRS89 has drifted about half a metre away from WGS84 at the
>> present time, but even so the transformations done by Lisboa under
>> Ubuntu look very good.
>>
>>
>>
>>> It looks as though Proj4 can use grid based transformations like
>>> OSTN02 through the use of gsb files. Could it be that this is the
>>> case in the higher accuracy situation? I find gsb files at the
>>> following location on my Win7 QGIS 1.7.4 install:
>>> C:\OSGeo4W\share\proj
>>>
>>
>> At present by default QGIS doesn't use a grid-shift file for
>> EPSG:27700. The datum-shift is carried out by the +towgs84 part of
>> the Proj parameter string.
>>
>>
>>
>>> I can't see any files in this directory suggesting they are the
>>> OSTN02 grid though, and I can reproduce your 'Vista' results as
>>> follows, suggesting that my QGIS is using the lower accuracy seven
>>> parameter transform.
>>>
>>
>> It's not using a grid-shift file. But it's interesting that you get
>> the same poor transformations that I get with 1.7.4 under Visa.
>> Lisboa
>> RC1 under Ubuntu produces much better transformations (for me) but I
>> don't know why this is the case.
>>
>> Nick.
>>
>>
>> --
>> View this message in context:
>> http://osgeo-org.1560.n6.nabble.com/OSGB-coordinates-to-WGS84-lat-lon
>> - problem-tp4965339p4966416.html Sent from the Quantum GIS - User
>> mailing list archive at Nabble.com.
>> _______________________________________________
>> Qgis-user mailing list
>> [hidden email]
>> http://lists.osgeo.org/mailman/listinfo/qgis-user
> _______________________________________________
> Qgis-user mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/qgis-user
>
> --
> The Open University is incorporated by Royal Charter (RC 000391), an exempt charity in England & Wales and a charity registered in Scotland (SC 038302).
>
_______________________________________________
Qgis-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/qgis-user
Loading...