I have files projected in British National Grid.
After I have transformed them into WGS_1984.
I noticed that if I were to transform them back to BNG the original would not match the new.
Is there a way to make sure this does not happen especially when moving from BNG to WGS 1984 multiple times?

  • 1
    How large is the difference between original and back transformed file? – Jens Jan 31 '14 at 15:05
  • 1
    Also which transformation are you using? (and is the same one being used for both steps--aka if you are checking in ArcMap, you need to set the tfm there too) Definitely should not be seeing offsets unless the data's wildly outside of BNG's area of usage. – mkennedy Jan 31 '14 at 17:09
  • Hi, I am using the standard petroleum transformation between BNG and WGS_1984. – user26425 Jan 31 '14 at 17:22
  • 1
    You might want to try unprojecting part of the BNG data to OSGB 1936, then round-trip that. That should isolate whether it's the transverse Mercator or the datum transformation that's causing problems. – mkennedy Jan 31 '14 at 17:53
  • Please edit the question to add more detail, instead of putting key information in comments. – BradHards Feb 11 '14 at 3:39

the datum of BNG is GCS_OSGB_1936. When you project the data (or move to lat/long coordinate), you should also make sure that you set the correct datum transformation. Usually, any transform that you would choose in ArcGIS is precise enough for common mapping needs, but ArcGIS is not transforming datums by default in case of projection.


Transformation parameters are the source of discrepancy. Make sure you use the same parameters while transforming back and forth.

enter image description here

  • Hi, thanks for the info but yes the transformation is the same – user26425 Feb 1 '14 at 12:38
  • Well, I'm not exactly sure how the internal transformation logic in ArcGIS works, but in theory, you don't use exactly the same transformation parameters when transforming back and forth. It's analogous to: if you decrease 100 by 5% you get 95, but when you increase 95 for 5 percent, you don't get exactly 100, but rather 99.75. But I think, ArcGIS should take care of that. – Matej Feb 1 '14 at 21:58
  • I just checked a roundtrip for someone else, state plane data (lambert conic) through two transformations, to another state plane zone, then returning. The results differed at 10e-05 feet. The majority of was due to dropping the ellipsoidal height in the one of the transformations. – mkennedy Feb 2 '14 at 0:00
  • @mkennedy: Note that transformation parameters apply only when transforming between different datums. If you reprojet data from one projection to another while both coordinate systems base on the same datum (like changing the state plane zones) you are not applying a transformation at all, but only projection recalculation. – Matej Feb 2 '14 at 20:59
  • @Matej, I did misspeak. The output was WGS84, input State Plane was on NAD27. But I think you should check my profile page! When I ran the test on State Plane data, I used NAD27 and NAD83, so 1 datum transformation, not two and the ellipsoidal height wasn't an issue. – mkennedy Feb 3 '14 at 0:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.