I have read that it is not advisable to perform spatial analysis on two layers with source datasets having the same datum but different projections loaded in ArcMap.

Since the second layer should supposedly line up with the first one due to lining up data on the fly in ArcMap, I was wondering what could possibly go wrong?

  • 1
    What does this mean: different projection system but identical coordinate system? Maybe it's ArcGis/ArcMap terminology. I think coordinate system depends on ellipsoid, datum and projection system. – Martin F Jan 17 '14 at 5:21
  • 1
    Can you reconsider your terminology and edit your question, please? Do you perhaps mean two layers with source datasets having the same datum but different projections? – PolyGeo Jan 17 '14 at 5:59
  • @PolyGeo that is correct. I revised the question. – Arash Jan 17 '14 at 15:25
  • 2
    @elrobis That reads like a pretty good answer :-). – whuber Jan 17 '14 at 15:59
  • 1
    @whuber, thank you. I went ahead and moved the comment into an answer field. At the time I suspected someone else might provide a higher quality answer, perhaps with a graphic, etc., so I didn't want to risk discouraging anyone from working up such an answer and posting it if a weaker, yet effective answer was already posted and accepted. – elrobis Jan 17 '14 at 20:01

As for what could go wrong--data in different projections may not share the same origin; therefore, mathematical calculations between such layers could be permitted without the software first interrogating the data for potential differences in origin, or with a view to making appropriate adjustments if necessary. Without the same origin, alignment issues are guaranteed.

The issue in Arc, I believe, is that not all routines are created equal. Some routines will defensively check for differences in projection, first, and either warn you or try to accommodate such differences if possible. However other routines do not check, do not accommodate, and blindly proceed under the assumption that the analyst already protected the routine against mismatched inputs.

Ultimately, Arc is being courteous by warning you that there might be a problem if you proceed, because at some more abstracted, outer ring of the software, it sees what may be an unintended operation about to occur. ..at the very least, it's seeing a bad practice unfolding, and as such it's nagging you. :)

@0kcats makes another good point with respect to performance sacrifices.

This being said, is it such an obstacle to translate both datasets into the same projection before running an analysis?


1) Performance issue - projection on the fly could be repeated many times during analysis and it could be slow due to horizon clipping and other things. If you bring your data into same spatial reference, you ensure better performance. Often significantly better.

2) Projection operation is not perfect. When projecting straight line segments, it projects the endpoints and reconnects those with straight lines. That is, to get good lineup you may have to density data before projection. If you project before analysis, you may observe and account for this.

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.