NAD83 to NAD83 HARN Florida transformation seems off

[KeithAddison1]

Depending on which transformation I use I get inconsistent results when converting points from NAD83 Florida State Plane East (SPE) (the vanilla 1986 one) to NAD83 HARN Florida State Plane East.

The below graphic was made by clearing the data frame coordinate system so I can get a visual on the projection differences.  (ie to get rid of the on the fly projection) The big blue dot is the original point in vanilla NAD83 SPE.  The green dot is the same point in NAD83 SPE HARN, projected using the NAD_1983_to_HARN_Florida transformation (which is the one Arc suggests)  The magenta dot is the same point in NAD83 SPE HARN but transformed using the WGS_84(ITRF00)_To_NAD_1983 + WGS_1984_(ITRF00)_To_NAD_1983_HARN transformation.  There's 1.5' of difference between the two, why do they differ?

I've been comparing different transformation from NAD83 SPE to other non-HARN realizations, and I'm increasingly suspect of the transformations that involve HARN when used as an intermediate, they seem to give inconsistent results when compared to transformations that don't involve HARN and use WGS84 as in intermediate instead.

[MelitaKennedy]

The transformation path using WGS_84(ITRF00)_To_NAD_1983 + WGS_1984_(ITRF00)_To_NAD_1983_HARN ends up basically a null transformaiton. The parameters are the same for each transformation. They were both added with the assumption that at some (in)accuracy level you can treat NAD83 / HARN / WGS84 as equivalent. 

Using the Florida HARN transformation is the correct way to go.

To see the green and blue points line up, you need to set the map to use the same transformation as well.

Melita

[KeithAddison1]

So when I use the WGS_84(ITRF00)_To_NAD_1983 + WGS_1984_(ITRF00)_To_NAD_1983_HARN it isn't really doing anything and that's why the magenta is directly on top of the blue even when the on the fly is cleared out by setting the data frame to no coordinate system?  Are the other similar transformation chains like WGS_84(ITRF00)_To_NAD_1983 + WGS_1984_(ITRF00)_To_NAD_1983_2011 likewise going to do essentially nothing, and where available I should use the transformations that go like: NAD_1983_to_HARN_Florida + NAD_1983_HARN to NAD_1983_2011 ?

My goal is to get a visual comparison of on the ground differences in the various realizations of NAD83 for a set of points in my area.  I guess where I was getting tripped up was not knowing some of the transformations aren't doing anything.

[MelitaKennedy]

Hi Keith, 

Yes, you're correct. You may want to look at the geographic_transformations.pdf file--it's located in the ArcGIS Desktop documentation folder. It contains tables of the transformations--extents, "accuracies" (very approximate), and parameter values.

Melita

[JoelCusick]

Hi Keith,

Bookkeeping transformations (ones with null values for parameters) are used quite often.

We did not have HARN up here, but for states with HARN, there is extra pain involved.  From doing this type of work your doing, here are some suggestions after being in the trenches so often its hard to count.

1) Ignore accuracies (the way ESRI puts a transformation on top.  Often, concatenated transforms are put on top, and I try to tell people to NOT do double transforms unless you have thoroughly solved one at a time.  keep to single transforms, and carefully jump from one to another.  Its insanely complex when combining transforms in one step.  

1a) Get comfortable with the Data Frame shifts on fly and effect with data assigned diff. projections.  That can really sting when your shifting data say in ArcCatalog, and the dataframe is shifting too.

2) Get onto a survey control (modern NAD83 (2011).  Conduct tests - if your using GPS - on a BULLSEYE, because eventually once you figure out what is "moving", your going to have the real question... Which one do I use.  

Finally, find a surveyor near you who knows all about HARN.  He/she will have sage advice.  Your down in the weeds man, where GIS is just mapping, what Surveyors would call "guessing".  

Joel