CORS - HARN transformations - what is the most accurate method

Question

This may not be an issue for many GIS-ers but has anyone else had experiences transforming from modern NAD83 (CORS) to HARN coordinates and vice versa? I realize the discrepancy might be something like 10 cm but for sources like low-level aerial, mobile lidar or ground surveys, this is still a big deal. What is the best practice? Most states still require HARNs as a deliverable coordinate system, and many companies survey tied to the latest realization of the CORS network.

An ArcGIS implementation would be appreciated.

Also this might lead to an interesting discussion for those who think the NAD83 -> HARN transformation in ArcGIS is the right thing to do with modern GPS-derived survey data (it's not!).

Answer 1

From the NGS which established the HARN's, the CORS and the NSRS2007 (which is the current nationwide re-adjustment of the coordinates of traditional geodetic monumentation using the CORS as control):

ANNOUNCEMENT

The Products and Services Committee of NGS voted to rescind the 1998 NGS policy which stated that a transformation would be provided between older versions of NAD 83 and NAD 83(NSRS2007). Initial tests indicated that the shifts were too small, the accuracy of the shifts too large, and sparseness of the data too great to produce a useful transformation.

Users who wish to transform existing surveys from an older version of NAD 83 into NAD 83(NSRS2007) are urged to recompute coordinates using original observations and newly published NAD 83(NSRS2007) coordinates.

Retrieved from: http://www.ngs.noaa.gov/NationalReadjustment/PSCannounce.html

Answer 2

I don't think you'll get an authoritative answer here, this requires someone with a deep knowledge of geodesy to answer precisely. From what I've read, CORS to HARN can be done down to about a 6cm accuracy, beyond which I'm guessing you'll need a professional surveyor to get an authoritative response. I would NOT use ArcGIS for such a transformation, if you're talking about accuracy to this level, but instead try the NGS tools directly.