A Datum Problem Solved?
I work for the U.S. Geological Survey on the National Atlas of the United States (nationalatlas.gov). We have distributed raster and vector datasets since 1997. The raster datasets are generally 1 km resolution, in Lambert Azimuthal equal-area (LAZ) projection, and cover North America in extent. Any hardcopy products we create from 1-km products are at 1:10,000,000 scale.
You’ll notice I didn’t say what datum these North America images are in. Back in the 1990’s, we created these images in ArcGrid, the subprogram of old ArcInfo Workstation. The LAZ projection was implemented only in spherical form, because 1. The equations were much simpler, and 2. Everyone (including USGS) only used the spherical form anyway. There was no reason for an ellipsoidal LAZ projection for products at 1:10,000,000. The difference would not have been noticeable anyway. We used to put a near-whimsical note in our metadata that users of these datasets should specify them as NAD83, in order to “fool” ArcGrid into letting you reproject them. At that time, ArcGrid would not reproject images that were in a spherical datum.
Times have changed, and for the better. Geographic transformations / datum shifts are much better implemented and more clearly documented than before. No more ignoring datum conversions during reprojection, as the old software let you do.
What I have, then, are circa 70 raster datasets, nearly all of them in LAZ on a sphere of radius 6370997 meters. After discussions both internal at the National Atlas, and with our partners in Canada, we realize we need to convert these images to WGS84. Thank you to Charlie Frye for the excellent Mapping Center article from 2008 on that topic (http://blogs.esri.com/Support/blogs/mappingcenter/archive/2008/07/17/gcs-wgs84-why-should-you-care-about-it.aspx).
I’ve read enough on the ESRI Web sites (especially Melita Kennedy’s excellent forum posts) to understand that there isn’t a systematic transformation from Sphere to WGS84, analogous to nadcon’s NAD27-to-NAD83. I had an idea on how to approach this problem for our new million-scale datasets, but when I ran some data through the Project command in ArcGIS 9.3, it seemed to transform correctly – meaning I think you all may have solved the problem already!
The input dataset was a raster in LAZ on a Sphere of radius 6370997 meters, and the output projection was specified as LAZ in NAD83 (I’m confident of getting from NAD83 to WGS84; that’s not an issue here). The Project command did not require specification of a datum transformation method, the function ran to completion, and when I looked at the results in ArcMap, the output image was shifted compared to the input (the coordinate system was cleared for the map layer, of course, so I could check). As a test, I overlaid some vector layers known to be in NAD83 on the output image, and everything registered correctly, certainly more than well enough for our new 1:1,000,000 scale products.
Let me be real clear: I’m not complaining! This is great. It will save me having to come up with a half-baked method of my own to convert from Sphere to NAD83, _if_ this is what is really happening. My question is: is this really what’s happening? Does the Project command in ArcGIS 9.3 convert from Sphere to NAD83? Or is it all just an illusion?
Mapping Center Answer:
The process ignores the difference between a sphere and NAD83. The raster is being unprojected to geographic on a sphere. At that point, the software is assigning NAD83 to the data and projecting it to the final coordinate system. In this case, Poject is more lax than the Desktop tools like Project Raster.
If you would like to post a comment, please login.