Cannot transform coordinate system from British National Grid to WGS_OSGB_1984 Mercat

[SamWilding]

Hi all

I will start by saying that I am very new to this, and although I have taken the introductory course to ArcGIS Desktop 10.0, it isn't really helping with my problem.

I have been gathering information from a number of stakeholders and marking it out on a map (charted raster).  In order to analyse the information I have imported the same background maps and the collected data into a new map document, however they do not line up.  It appears that the reason for this is that in the software that we are using (a specially developed questionnaire, the CS was set as British National Grid, yet the charted raster are WGS_OSGB_1984 Mercator - when I first imported the maps to the questionnaire package it informed me that there was a need to transform the coordinates which I did.

I have researched this as far as I can, and apparently there is a transformation for BNG to WGS84 available for 9.1, but I cannot seem to find anything for 10.0.  All of the transformations provided in Arc 10 are useless, and I have spent days trying to calculate a transformation myslef but it is going all over the place with no real logical pattern.

PLEASE HELP ME :`(

I would add at this point that I work for a non-profit and therefore we do not have the budget to pay a developer to sort this out, but if you can help me you would be an absolute hero!

Thanks

S Wilding
CBFA Ltd

[MelitaKennedy]

Is the raster in WGS_1984_Web_Mercator or WGS_1984_Web_Mercator_Auxiliary_Sphere? That will make a difference. The first one uses a custom geographic coordinate system (modified WGS 1984, "GCS_WGS_1984_Major_Auxiliary_Sphere) and there's no direct transformation between it and OSGB 1936. You have to use two transformations, which is...difficult...in ArcMap, but can be done by converting the data in ArcToolbox using the Project or Project Raster tools. The 2nd one uses standard WGS 1984 and there should be available several transformations.

Melita

[SamWilding]

Hi Melita,

Thank you for your input.  The Coordinate system is described as WGS_1984_Mercator, there is no mention web_mercator or web_mercator_auxillary_sphere.

I will try to use the transformations you have mentioned, but any additional info you could provide would be gratefully received.

Many Thanks



Sam

[MelitaKennedy]

Hi Sam,

Oh yes, that's a generic Mercator definition (centered at 0,0) that we added as a test case. If I add WGS84 data to ArcMap and set the data frame's coordinate system to British National Grid, I get this list of transformations with the areas of use and accuracies added:

OSGB_1936_To_WGS_1984_1  "UK - Great Britain; Isle of Man" 21.0
OSGB_1936_To_WGS_1984_2  "UK - England" 10.0
OSGB_1936_To_WGS_1984_3  "UK - England; Wales; Isle of Man" 21.0
OSGB_1936_To_WGS_1984_4  "UK - Scotland" 18.0
OSGB_1936_To_WGS_1984_5  "UK - Wales" 35.0
OSGB_1936_To_WGS_1984_NGA_7PAR  "UK - Great Britain; Isle of Man" 21.0
OSGB_1936_To_WGS_1984_Petroleum  "UK - Great Britain; Isle of Man" 2.0

When checking for possible transformations, only the geographic coordinate systems matter.

Melita

[SamWilding]

Hi Melita

I have tried the transformations listed for both transforming my stakeholder data (gathered under GCS 1936/British National Grid) to 1984 (used for the background map) and transforming map data to 1936/BNG GCS.  I have put a marker on a distinctive point to try and align them without much success, below are the coordinates for each set of transformations to show the scale of the problem.  I don't know whether this will enable you to help me any further, but its all that I can think of at the moment (logic is not getting me very far).

Transforming 1936 data to 1984 (all values in metres)

Reference point                   -475663.757, 6844471.977
Transformation 1                 -283557.103, 4080120.052
Transformation 2                 -283551.527, 4080094.476
Transformation 3                 -283602.731, 4080103.425
Transformation 4                 -283547.172, 4080068.931
Transformation 5                 -283547.172, 4080107.315
Transformation_NGA_7PAR    -283568.237, 4080103.425
Transformation_Petroleum     -283568.237, 4080068.931

Transforming 1984 data to 1936 (all values in metres)

Reference point                   245573.383, 283857.525
Transformation 1                 215822.76, 2396583.867
Transformation 2                 215825.84, 2396594.223
Transformation 3                 215825.84, 2396594.223
Transformation 4                 215825.84, 2396610.354
Transformation 5                 215809.709, 2396578.091
Transformation_NGA_7PAR    215825.84, 2396642.617
Transformation_Petroleum     215825.24, 2396642.617

I had previous tried to work out my own transformation using this reference point and could get it quite close (within a km or so), but I think there was a scaling or rotation issue (I was using a simple xy transformation).  When I have tried to develop a more complex transformation the data seems to jump all over the place without any logical pattern.

Any advice would be gratefully received.


Sam

[MelitaKennedy]

Hi Sam,

How far off is the 'best' transformation compared to your test point? All these transformations are built for different purposes and have different accuracies.

There's another possibility (not supported in ArcGIS yet, so it's a little more work). Ordnance Survey has released an NTv2 file for the UK that's built from their OSTN02 transformation data and method. You can get it here:

http://www.ordnancesurvey.co.uk/oswebsite/gps/osnetfreeservices/furtherinfo/ostn02_ntv2.html

After downloading the files, extract them. You'll need to locate the pedata folder in your ArcGIS installation. In that folder, you'll see an ntv2 folder. You can either put the .gsb file there or create a uk folder and use that. Now use the Create Custom Geographic Transformation tool in ArcToolbox to define a transformation for it. For the input GCS, use OSGB 1936. For the target, use WGS 1984.  Method will be NTv2. For the parameter value, if it's just in the ntv2 folder, use the file's name without the extension. If it's in the uk folder, use uk/filename.

Melita

[SamWilding]

Hi Melita

I have tried the NTV2 transformation as suggested and it doesn't work either.  It is close to transformation 1 in terms of how close it gets to the reference point, however transformation 1 remains the closest.  That said it isn't that close.  Transformation 1 puts the data 92086.659m to the east and 2764351.925m to the south of the reference point (which is probably somewhere around the Azores rather than the offshore waters of Wales).


Sam

[MelitaKennedy]

Hi Sam,

Now I'm wondering if there's something odd in the Mercator coordinate system definition. Offshore south Wales, if the data's using Esri's "WGS_1984_World_Mercator" should have easting values around -500k to -600k and northing values between 6.6 and 6.8 million. Is that what you're seeing? If you add the raster to ArcMap, but set the data frame's display units (data frame property page, general tab) to decimal degrees, do the values make sense?

If I convert to BNG using the 7par transformation, the easting values are around 150k to 260k and northings around 175k to 270k. The output values you're reporting do not look like BNG, which means something odd in the source coordinate system.

Melita