NGS Coordinate Conversion and Transformation Tool (NCAT)

[shelby-h-griggs-pls]

I still don’t know what the accuracy values are based on or really mean, but I ran nice points through the NCAT software and compared to the 1995 record:

Values in Feet; Green = Average differences, Blue= minimum difference value, Magenta = maximum difference value, it seems pretty legit! Strange, highlighting shows up on my preview, but not after posting, anyway, you can see the values. The last three values are average, minimum, maximum. Summary is nine points agree on average 0.05’N and 0.00 ‘ East

Diff Y	Diff X
-0.03	0.01
0.09	0.01
0.02	-0.00
0.02	-0.03
0.11	0.03
0.08	0.01
0.10	-0.01
0.03	0.00
0.02	-0.05
0.05	-0.00
-0.03	-0.05
0.11	0.03

[loyal]

Howdy Shelby,

I don’t have much to add. I have been “playing around” with NCAT a bit lately, and as with most “models” results vary somewhat. In virtually all cases, GPS derived NAD83(2007 or whatever) convert to NAD83(2011) in such a way that the Lat/Lon (and usually the Eh) IMPROVE “somewhat,” relative to the “NEW” observations results (as they should). I have used it on somewhat larger projects than yours (10-20 mi?ý) in several western states, and can’t complain about the results.

Obviously the “accuracy” of the underlying velocity model is probably the wildcard (at least to some extent). In “stable” well constrained areas (lots of CORS & PBO sites), it works BETTER than in more dynamic areas with sparse CORS/PBO coverage (no surprise there).  

I’m more than a little fuzzy on exactly what the “quality values” really mean too, and being the old curmudgeon that I am, I am satisfied with the empirical evidence suggesting that it “works” within the statistical variances that come out of the OLD v. NEW coordinate/height comparisons. When it all fits together within a couple centimeters, and I’m comparing 5 hour static (NEW) to ??? RTK (OLD), who can complain.

Loyal  

[jhframe]

Anyone know how those values are derived and what they mean? one sigma, two sigma, etc.? Anyone extensively used this tool and have a feel for generally how good it is?

I have even less experience with NCAT than either of you, but I was curious so started nosing around.  Documentation is pretty sparse, but I note that NCAT uses the NADCON 5 engine, and the NADCON 5 technical report (NOS NGS 63) has a discussion of formal error estimates.  That section (8.2) contains the following passage:

If one then compares the transformed NAD 27 coordinate to the published NAD 27 coordinate, a
difference is generated. What is hypothesized is that this difference will be smaller than the formal error
estimate about 68% of the time. Further it is hypothesized that the difference will be smaller than twice
the formal error estimate about 95% of the time, and smaller than three times the formal error estimate
about 99.7% of the time. That is, the formal error estimates compared to the true differences will follow a
normal distribution.

This leads me to believe that the sig values spit out by NCAT are 1-sigma, which makes sense to me.

[shelby-h-griggs-pls]

Thanks guys, well after looking at this some more and pondering it for another 24 hours, I am going to deliver the CORS based survey and translated by NCAT to NAD83(91) or HARN. Best I can tell I am hitting this points within reason especially given the old points are 24+ year old survey data with potential movement (in a coastal area) plus the improvements in GNSS over that time. On the only NGS point I had with both epochs published I am within 0.03′ north and 0.01 feet east (that is my CORS based surveyed position converted vs the record. If I use the NGS published 2011 data and convert that to HARN using NCAT it hits the published historic HARN values with 0.01 feet x 0.01 foot) The DOT and other private survey from 1995 and earlier I would trust less, but still hit all the points worst case by 0.11 feet in either northing or easting. Empirically it seems this bit of software is working pretty good on this project and for 2-3 cm level control for aerial mapping control points, I don’t think I will sweat it. FYI, the shift from current epoch to our HARN is around 0.45 feet south and 0.44 feet west in this project area.

SHG

[gene-kooper]

Well, late to the show as usual.  I presume the 3D transform tool that you are using is the map datum/transform module in Leica’s upgrade to LGO.   In the past I did a lot of map datum transforms of satellite imagery using topo maps.  There were always a stinker or three among the points picked off of the topo maps so I would isolate them by turning off GCPs with excessive RMS values.  The stinker(s) were not always the one(s) with the highest RMS.  As an analogy, I’d suggest picking 6 of the 9 control points and compute the transform parameters.  Going through your control stations will likely isolate the station(s) that may have been nudged over time.

Following up on what Jim said, the NCAT transforms are done using NADCON 5.0.  From the NADCON 5.0 History web page:

More specifically, NADCON 5.0 supports transformations between all 3-dimensional geometric reference frames and nearly all traditionally-determined 2-dimensional horizontal datums of the National Spatial Reference System, dating back to the 1890’s. Unlike previous transformation tools, NADCON 5.0 provides local error estimates for each transformation grid. Transformations only exist between chronologically adjacent datums, but the transformations and their error estimates can be summed to move across several datums.

So it appears that the local error estimates for your example would be the summation of the local errors transforming between NAD83(HARN) to NAD83(FBN) to NAD83(2007) and finally NAD83(2011).  You can check and see if NCAT is doing the summation suggested above when you transform from NAD83(HARN) to NAD83(2011).  The history web page also provides a link to NOAA Technical Report NOS NGS 63   I haven’t looked at the technical report in detail, but it does discuss local error estimates (see Sec. 3.12 Produce Local Error Estimates) and also from Jim’s post, Chapter 8 contains an explanation and error table (see Sec. 8.2 NADCON 5.0 Formal Error Estimates).

My only real experience with NADCON is to provide “modeled” NAD27 coordinates for mining clients and esp. for geologists who seem incapable of working in NAD83 (at least the old-timers) because all the USGS topo maps (used for their base maps) are in NAD27!  I explicitly state that the NAD27 coordinates are modeled coordinates based upon the NGS lookup tables in the NADCON software.  At least that way there is a connection between my geodetic coordinates and what the geologists play with after I apply the lookup tables.

Oh, one other thing how does your 3D transform that isn’t quite right compare to the NCAT results you published?

[shelby-h-griggs-pls]

@gene-kooper thanks for your input, sorry for the late feedback, basically sent the coordinates off and then went camping!

The tool works, or works good enough for what I wanted, so I didn’t fiddle with Datum and Map any more. It might of been the distribution of the points known in both systems as the conversion tool gave reasonably close results on the known common points, certainly didn’t see any real”stinkers”. I have used the Datum and Map tool successfully in past, in fact the previous two projects before this one were also delivered in NAD83(HARN) and I was able to develop a decent 7-Parameter transformation using Datum and Map to get from current CORS. The project this week will also be delivered in an “old” realization of NAD83 too, seems all I am getting in the door of late 🙂 These have all been updates to an ongoing mapping program each for a City or County, they don’t like change….

There is already a canned transformation for the upcoming project, but I think I may see how the NCAT tool performs just to get some feel for a different area.

SHG