Robin's Blog

How to choose a co-ordinate transformation in ArcGIS

When you try and reproject a dataset in ArcGIS (for example, by using the Project Raster tool) you will see a dialog a bit like the one below:

The highlighted field wants you to specific a Geographic Tranformation. Although it says that it is optional, it often isn’t (I think the optionality depends on the type of transformation you’re trying to do). I’ve often found that there are a number of items available in the dropdown box and I have absolutely no idea which one to choose!

For example, when converting from OSGB to WGS84 there are the following options:

  • OSGB_1936_To_WGS_1984_1
  • OSGB_1936_To_WGS_1984_2
  • OSGB_1936_To_WGS_1984_3
  • OSGB_1936_To_WGS_1984_4
  • OSGB_1936_To_WGS_1984_5
  • OSGB_1936_To_WGS_1984_Petroleum

How on earth should you choose one of these? Until now I had been choosing semi-randomly – picking one, seeing if the result looks good, if not then trying another. However, the other day I found out about a list of these transformations in the ArcGIS documentation – available to download from the ArcGIS documentation page. This document (available for each different version of ArcGIS) lists all of the transformations available in ArcGIS and their theoretical accuracy. So, for example, we can find out that OSGB_1936_To_WGS_1984_2 is meant to cover England only, and OSGB_1936_To_WGS_1984_4 is for Scotland. The accuracies seem to be around 20m for each transform, although OSGB_1936_To_WGS_1984_5 (for Wales) has an accuracy of 35m.

I can’t believe I’d never come across this resource before – it allows me to actually make intelligent decisions about which transformation to use. I’d strongly suggest you get familiar with this document.

(I’d like to thank the GIS.SE users who helped me with a question I asked related to this problem)


Categorised as: Academic, GIS, How To, Remote Sensing


4 Comments

  1. Yes, most people, when they come to map projections for the first time, assume every place on the Earth has a defined latitude and longitude value. It comes as quite a shock to them when they realize this is not exactly the case. Every spot on Earth has a defined latitude and longitude value in the arbitrary coordinate system you choose to use to define the spot! This coordinate system is called the datum, and there are literally hundreds if not thousands of datums to choose from!

    When you convert from one datum coordinate system (OSGB_1936) to another (say, WGS84) you have to do some kind of datum transformation to make sure the same spot on the Earth is located in both coordinate systems. In other words, the latitude and longitude values in the new coordinate system will be different from the latitude and longitude values in the old coordinate system. Most mapping software, even IDL, can do this kind of datum coordinate transformation.

    But, what other systems do, and ArcGIS is a good example of this, is also allow you to translate and rotate the old coordinate system, so that the old values you are transforming are located very close to the new coordinate values in some small, localized area (say, Scotland). It’s as though you took the ellipsoid (OSGB datum) and were able to move it from its normal anchor point in the center of the Earth and nudge it next to the new ellipsoid (WGS datum), so they were touching at Scotland. (This would make things wildly wrong on the other side of the Earth, but thankfully, you are not looking at the other side of the Earth!) Some software allows a 3-point datum shift (essentially, a translation and rotation), while some of the best software allows an 8-point datum shift (a very accurate 3D shift). Depending upon how accurately you need to know your location, these kinds of datum movements are often unnecessary. (If your data is accurate to only 250 meters, improving the results by another 5 meters or so seems rather pointless.)

    I’m working on a new book on map projections in IDL, which I hope will bring some of this information to IDL users.

  2. Robin Wilson says:

    David, thanks for the detailed comment – it explains a lot more about how this all works, and taught me some stuff I never knew. I’ll look forward to the book on map projections in IDL 🙂

  3. Mark says:

    I would also suggest that you look at the EPSG Geodetic Parameter Dataset, which you can find at http://www.epsg.org. The database gives you a useful way of searching what transformation parameters are valid in any given area.

  4. Matt says:

    Well we always use OSGB_1936_To_WGS_1984_Petroleum because someone once said it was the best! Ultimately none of the transformations are perfect, and you need to use other software to use the published OS transformations.

Leave a Reply

Your email address will not be published.