Querying through the EPSG database, I can see the tables epsg_coordoperation where transforms and conversions between different CRS are stored, and epsg_coordoperationmethod which stores the mathematical methods used by the transformations. Correct me whether I'm wrong.

To make an example, looking through the details of the conversion form ED50 to WGS84 over Sicily (http://www.epsg-registry.org/, --> retrieve by code --> 1143), I can see that the transformation involves simple shifts on the 3 cartesian axis. Operation is ED50 to WGS84, operation method is Geocentric Translation.

My statements which I would like to be either confirmed or corrected are:

  • coordinate operations are always meant to reach a target CRS;
    • coordinate transformations involve a change of datum, e.g. [geocentric<-->geocentric] or [projected<-->projected], etc.
    • ..as coordinate conversions do, but regarding a [projected<-->geographic] error-free translation;
  • one could apply arbitrary operation methods to a georeferenced dataset, say a simple arbitrary shift on one or more dimensions in the most simple case, given that the requirementes are met.

My questions:

  1. Considering only the operation methods, does applying method X to CRS A always take to a different CRS? But which one? E.g. a Lambert Conic Conformal (2SP) - EPSG:9802 - projects ellipsoidal coordinated to a projected XY map, but to which CRS? Or again, a shift on an ED50-based image may take to a WGS84-based image, but only if the shifts are the ones defined in the correct operation (--> Parameter values).

  2. Bob wants to reproject a coverage from a source to a target CRS. Are there cases in which he would prefer to specify its own path of CRS transform concatenation instead of the default method offered by foo library?

Thanks in advance for any effort.


Disclosure: I'm on the subcommittee that maintains the EPSG Geodetic Parameter Dataset.

I may still not correctly answer all your questions. In the EPSG parlance, a transformation doesn't have to include a change of datum. Any change is an operation.

I work for Esri, and there I do differentiate and use transformation to mean a change in datum (geographic coordinate reference system).

Back to EPSG. I would not call a coordinate conversion, error-free. Perhaps error-free within the limits of the data accuracy.

A method is an abstraction until you identify the source/targets, plus any parameters needed by the method. I think there are methods that do not convert between two different CRS like unit changes when working with geographic crs. But almost all do require that the source/target differ.

Of coursse, someone might want not to use a 'default' operation / transformation. One reason might be to undo a previously applied operation. That operation could have been applied by mistake, or is simply outdated, and Bob now wants to convert the original data using the latest and greatest operation.

Update (to add a concrete example): Bob receives data collected by a colleague using a GPS device. The colleague, knowing that the database is still using NAD 1927, helpfully set the GPS device to use NAS-A as the first listed NAD 1927 'datum' (NAD 1927 with transformation parameters designed for the Caribbean). The EPSG well-known ID for this transformation is 1170. Bob uses it to convert the data back to WGS 84, then uses 15851 to convert to NAD 1927. 15851 uses the NADCON method and is more accurate than using NAS-A/1170 for several reasons. It's designed for the lower 48 states, not the Caribbean, and because it uses a file of offsets can model differences between NAD 1927 and NAD 1983 / WGS 84 (assuming they can be considered equivalent--for this data, let's assume they can be!) much better. NAS-A/1170 uses geocentric translations, so for the entire area, practically the same offsets are used.

| improve this answer | |
  • Good Lord.. Melita Kennedy. Thanks for clarifying, that helps understanding the EPSG database philosophy. I didn't understand your last paragraph however: do you have in mind a concrete example? – Campa Apr 24 '12 at 8:15
  • Hi Malita. Vince mentioned you over here: UTM/NAD83 to WGS84. – User1973 Feb 5 at 4:33

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.