1

I have a point feature class in a geodatabase that uses WGS84 as its SRS. A colleage of mine created points in this feature class using an external database containing lat and long values. However, the lat and long values in the external database were using NAD83 (my fault -- I knew they were, I just forgot). I thought it wouldn't be a problem -- I just created a copy of the geodatabase and changed the coordinate system used by the feature class to NAD83. Should be good, right?

I added this feature class into ArcMap. I saw that the points in the copied feature class (now using NAD83) exactly line up with the original feature class (using WGS84). That doesn't seem right. If two points, having the same lat / long values, but referenced in different coordinate systems are overlaid on the same map, one of them should be projected on the fly to the coordinate system of the data frame, no?

Now, here's the weird part -- if I project the original point feature class (using WGS84) to NAD83, and add it to the map, then the points don't line up. See, this isn't what i expected, either. I expected that they would be projected to the other coordinate system (i.e., their lat / long values would be different by virtue of being projected, but they would be aligned with the original point because they were simply projected to the new coordinate system (then projected on the fly back to the original coordinate system).

Can anyone explain the behaviour that I'm seeing? And, how do I properly ensure that the NAD83 lat and long values collected by my colleague are projected to WGS84 in my target geodatabase properly?

PolyGeo
  • 65,136
  • 29
  • 109
  • 338
Jonathan Bailey
  • 263
  • 2
  • 14
  • How did you change the coordinate system from WGS84 to NAD83? What tool/procedure did you use? Also, when you say "exactly lined up", do you mean the coordinates are identical, or that visually you can't see a difference? Also, be aware that there are multiple transformations between WGS84 and NAD83, but when ArcMap projects on the fly, it just chooses the most common one. – Baltok Feb 26 '15 at 22:04
  • I just changed the feature dataset properties in ArcCatalog.

    When I say "exactly lined up", I means coordinates are identical, and I can't see a difference, even when zoomed into a scale of "1:0".

     – Jonathan Bailey Feb 26 '15 at 22:08
  • So, your colleague's geodatabase was incorrectly defined as WGS84, but you knew it was actually NAD83? Also, how were these NAD83 points added to your WGS84 points layer? Also, what tool did you use to "project" the original feature class from WGS84 to NAD83? – Baltok Feb 26 '15 at 22:20
  • Those two CRS also use different datums. You should have gotten a warning when adding the second one that said you may need a transformation. Did you select one and if so which? If you did not specify a transformation, then yes, I can see the points lining up even though they're in different CRS. @Baltok from the comment he didn't use a tool - he just changed the properties using ArcCatalog. That's the equivalent of using the Define Projection tool. Also, ArcMap does NOT choose a transformation by default - it throws a warning or uses no transformation at all. – Chris W Feb 26 '15 at 22:27
  • @Chris W. - good point. My first comment assumed that the user didn't specify a transformation, in which case, it will default to the most common. It is good to know you can choose the transformation if you want. Hm, I seem to recall reading an ESRI white paper in which it said ArcMap will default. Not sure where that was... – Baltok Feb 26 '15 at 22:32
  • @Baltok The only reference I can find is at http://resources.arcgis.com/en/help/main/10.1/index.html#//00660000000v000000 where it says "ArcMap doesn't automatically choose a transformation for you... [omitted] There is one exception to this..." The exception is if it's NAD27 and NAD83, it will automatically use NADCON. – Chris W Feb 26 '15 at 22:44
  • Ha ha! That's the exact thing I just found! I stand corrected. Other than NAD27 to NAD83, ArcMap does not automatically assign a transformation if none is selected. So, in this case, SpatialBridge would need to choose a transformation for NAD83 to WGS84. – Baltok Feb 26 '15 at 22:50
  • so just to make sure. the two data sets did not ever live in the same feature data set did they? – Brad Nesom Feb 26 '15 at 23:12
  • OK. Thanks for the feedback, all. The problem was that I had not chosen a transformation (you're right Chris -- should have read the warning dialog). Choosing a transformation seems to result in the correct behaviour (the coordinates are slightly shifted because they are defined in relation to a datum that's slightly different. Now, how to I choose the correct transformation? ArcMap indicates that it ranks the transformations in order of appropriateness.

    Also, I read the other question -- I don't think that it's addressing the same issue.

     – Jonathan Bailey Feb 27 '15 at 15:56
  • For clarification, my colleague's database is a vanilla database -- it has text fields for latitude and longitude values. These values are in reference to NAD 83. He knows it. I know it. But, I wasn't thinking of it when I gave him a geodatabase with a feature class using WGS84. He created new points, set their x and y according to the lat and long values in his database -- that's how I ended up with the features reference to the wrong coordinate system -- I've been able to fix this by changing the feature class properties in ArcCatalog -- you can also use the Define Projection tool. – Jonathan Bailey Feb 27 '15 at 16:05
  • Unless you know you need a specific transformation, you can choose the first/top one in the drop down box. What it comes down to is the transformation method (grid vs formulas vs specific formulas) may vary a bit to be optimized for a specific area of the world. The only quick example I can find is here Esri has a more complete which transformation to select list. – Chris W Feb 28 '15 at 01:41

0 Answers0