RE: [OKGIS] 1995 DOQ Accuracy?


Chronological Thread 
  • From: "Phil Stevenson" <pstevenson@sbcglobal.net>
  • To: "'Oklahoma GIS Community'" <okgis@gis.gis.ou.edu>
  • Cc: Randy Mosley <randy@nfork.com>
  • Cc: Deral Paulk <Dpaulk@cityof.lawton.ok.us>
  • Cc: Burk Cornelius <burk@osls.org>
  • Subject: RE: [OKGIS] 1995 DOQ Accuracy?
  • Date: Sun, 3 Oct 2004 11:09:05 -0700
  • Importance: Normal
  • List-archive: <http://gis.ou.edu/pipermail/okgis>
  • List-id: Oklahoma GIS Community <okgis.gis.ou.edu>

Pete and David,

If this gets too boring or too off subject for the list
server please let me know and I will make my next response
more personal. My response is late but some other science
projects had my attention for a while.

Thanks for the flower, Pete, I keep saying that
documentation of my science projects needs improvement.
There are some individual projects done by Randy Mosley and
Deral Paulk that look very impressive to me. Randy has done
extensive testing of OPUS against HARN stations. Deral has
recently been testing WAAS and GPS hardware against GIS maps
and monuments built from a framework of the HARN.

You can read more about their efforts on the OSLS forum and
the message boards at

http://www.osls.org

http://www.i-boards.com/bnp/pobtha/

http://www.i-boards.com/bnp/pobtop/

It seems that one of the main tech support topics of the
last few weeks has been about coordinate conversions in many
far flung places. This difficulty exists in most places
where GPS receivers are used to capture data that is
intended for plotting on maps that use map projections other
than WGS84(ITRF00). Every time I use the term WGS84 I
really mean ITRF00 because there is not a way for me to
determine an actual WGS84 position.

Confusion about coordinate systems and conversions abounds.
GPS makers, whether because of budget or confusion, have not
figured it out. Coordinate conversion software companies
have also dragged their feet when it comes to getting the
most exact answers possible. One such company asked if I
would pay for the engineering required to get it right.

The end result is that the best advice is not to trust any
GPS device or commercial software product to give a correct
coordinate conversion until you have tested it against the
NGS standard by using NGS survey monuments or NGS software
for the coordinate conversion. The NGS is the national
standard in the USA. They are part of the FGDC.

http://www.ngs.noaa.gov/

There is some help, and some comprehension available.

Part of the problem is NAD27 itself. It is not made for
mapping with GPS. In 1927 there was no space travel, no
computers or fancy software, and no way to connect geodetic
control across large bodies of water. The reason why the
NGS says that the use of NAD27 is a bad idea is because
mapping and surveying professionals bought electronic
measuring tools and computers for calculations. Conversions
from ITRF00 to NAD27 are always going to be approximations
of the truth with regional variations that make correct
answers difficult.

Testing standards exist. All commercial coordinate
conversion products should be checked against the HTDP,
NADCON, VERTCON, UTMS, or whatever appropriate NGS tool is
available. Read the documentation carefully and be careful
about assumptions. Somebody told me that, "All UTM
coordinates are referenced to WGS72." In his industry that
may have been a correct statement but I used official US
Government UTM coordinates before there was a WGS72.

In Oklahoma there are maps related to UTM coordinates that
are referenced to NAD27 and NAD83(1986) at least. Since
some of the maps do not specify what version of NAD83 is
used it may be NAD83(1986) or NAD83(1993). In some products
you will see NAD83(1993) referred to as NAD83(HPGN) or
NAD83(HARN). These are, or should be, equivalent terms.
Now there is a third national adjustment of NAD83 called
NAD83(CORS96) that is not a very big deal in Oklahoma but
here in Santa Clara, California we can measure the
difference with our pocket tapes. Your national news
programs may have provided some recent reminders about why
the west coast geodetic control monuments move around just a
little. Some will say that the difference between the
various versions of NAD83 does not matter. If differences
of four feet do not matter then I will agree.

The problem is not about the accuracy of the maps. The
problem is our expectations. We used to be tolerant of
utility maps that had manholes and water valves fifty feet
in the wrong place. Times have changed.

The NGS software is the standard we must use. However, you
may be shocked to discover that the NGS software is not an
exact conversion. Read the documentation for HTDP and
NADCON. It is based on models. Compare some conversions
with the NGS data sheets to gain understanding.

Some will say that WGS84 and NAD83 are equivalent. In 1984
that was close to the truth. Twenty years later the DoD has
moved WGS84 a few feet away from NAD83(all versions). The
folks at the DoD might say that the NAD83 has moved since
WGS84 really is earth centered and earth fixed while NAD83
is now approximately so. You can see the very precise
difference, within 0.050 meters, by looking at a CORS data
sheet. The NGA tells us that the current version of WGS84
is within 0.050 meters of the ITRF00. That is as close as
we can get to the truth. For practical purposes we must
view the relationship between present day WGS84 and ITRF00
as equivalent. Then we can make real comparisons between
those ITRF00 and NAD83(CORS96) coordinates on the CORS data
sheets.

http://www.ngs.noaa.gov/CORS/coordinates/

Some people rely on the NGA tools, like the GeoTrans
software, to provide accurate conversions. Read the
documentation and you will discover that GeoTrans is
intended for use with those old maps where a few meters does
not make much difference. GeoTrans type conversions have
been included in some commercial products. The NGA
documentation will tell you that it is intended for use with
old maps with accuracy estimates that vary from one meter to
more than 25 meters.

http://earth-info.nga.mil/GandG/

So how can we evaluate the coordinate conversions?

Some of the NGS survey monuments in Oklahoma have a history
that will let you test both the commercial software and the
NGS software against the surveyed position of the monuments.
Go to the NGS web page and download a data sheet for an A or
B order survey monument near you. Keep looking until you
find one that has records going all the way back to NAD27.
The best choice would be a monument that was 1st or 2nd
order before the GPS work that was done in 1992-1993. That
way you have the best quality coordinates for that monument
that will take you from NAD27 to NAD83(1993). You can get
the ITRF00 if you can draft somebody with a dual frequency
GPS receiver. By using OPUS you can determine a modern day
ITRF00 and NAD83(CORS96) position for that monument. The
data sheet will take you from NAD27 to NAD83(1986) to
NAD83(1993). You will need to spend a few hours on that
monument with a dual frequency GPS receiver to get the most
current NAD83(CORS96) and ITRF00 coordinates for that
monument. There are a few surveyors in Oklahoma who have
done exactly this sort of thing. You may find some of them
by posting a message on the OSLS forum or by simply making a
request for information to the OSLS and letting them pass
the word around.

http://www.osls.org

This effort would provide a point on planet earth where you
could make a physical comparison of a few generations of
geodetic positions.

That is a spot where you can "ground truth" your GPS
receiver. If you find a GPS receiver that will make an
exact conversion to NAD27, to NAD83(1986), to NAD83(1993),
or NAD83(CORS96) I will be very surprised. The GPS
receivers I have touched see WGS84 and NAD83 as the same.
They do not see a difference between the various versions of
NAD83 with the end result that any conversion to NAD27 is
going to be both varied and approximate.

What is the difference made by going from 2nd order on the
NAD27 datum to B order on the NAD83(1993) datum? It is a
valid question for which there may not be a valid answer.
But the truth is that the existing NAD27 geodetic control
was used to build the NAD27 maps. The maps cannot be more
accurate than the geodetic control that was used to provide
the foundation upon which those maps were built. Whatever
error you will find in a coordinate conversion versus the
actual data sheet is also going to be in any coordinate
conversion you make to put your modern measurements on an
old map.

If that does not make you go, "Hmmmmm?!?!" then this epistle
was wasted effort.

I use a few brands of commercial coordinate conversion
software on my computer here at home. None of them get all
of the conversions right. Some are closer than others. I
am not ready to throw any of them away and start over.

A few brands of GPS receivers and data collectors make it
into my hands. Almost all of them will make good and
repeatable GPS measurements within the specifications. None
of them have been capable of making an accurate conversion
between the various datums. I define accurate as a
conversion that does not negate the specifications for
accuracy. A specification that claims accuracy to three
meters fails when the conversion to NAD27 has five meters of
error in it. A specification that claims sub-meter accuracy
fails when the conversion to NAD83(1986) misses by two
meters. These are just my opinions. Some would advise that
all of our measurements should be made relative to the
ITRF00 and the conversions should be a rubber sheet on our
individual maps. This would probably provide the most
accurate results but it puts the burden for coordinate
conversion directly on the map user.

One of the reasons why I am willing to buy expensive
hardware and software is to reduce the burden of work on me.
But I remember a seminar where I asked the students why they
have purchased modern software. One participant blurted
out, "Because we want everything to take longer and cost
more!"

No matter what the GPS hardware and the mapping software
makers do, it remains the personal responsibility of the
professional user to ensure that the answers are correct.

I feel privileged to have started my surveying and mapping
career in Quang Tri Province, Vietnam where we used maps
made in the USA and maps made in France. They often showed
some differences in the topographic features but it was even
more of a challenge to deal with the coordinate differences.
It was not just about an origin of longitude in Greenwich,
England or Paris, France. It was about a different way of
looking at maps. It's a good bet that none of those map
makers walked around out there in the weeds. That part was
our job. Map or measurement errors sometimes resulted in
what was called "friendly fire". I am a living witness that
there is not one single thing "friendly" about friendly
fire.

Some people see UTM as being a "universal" mapping system.
This is going to become even more commonplace if people
begin to adopt the USNG. World travels over the last few
decades tell me that UTM coordinates are not all referenced
to the same starting point. A variety of datums have all
been used to create UTM map projections in places around the
world and in the USA. Does it make a difference? Get UTM
coordinates for the same road intersection from a USGS DRG
based on NAD27 and UTM coordinates from a DOQQ based on
NAD83(????) and tell me whether it makes a difference. Will
adoption of the NSRS and the U.S. National Grid help? It
will only help if it really is adopted and implemented.
Next question: Will the GPS, GIS, CAD, and COGO software
makers get with the program? Will there be tools to get the
old digital maps to the modern coordinate systems? Will
contracts specify the use of the most modern coordinates?
Will the FEMA and other federal agencies get on board?

How many people will look at the cost of modification to
their "legacy" data and shoot down any notion of uniformity
with the most current version of the coordinates?

http://www.ngs.noaa.gov/PUBS_LIB/develop_NSRS.html

http://www.ngs.noaa.gov/cgi-bin/showdoc.prl?Doc=usng_sys.doc

http://www.ngs.noaa.gov/NationalReadjustment/

Datums that predate space travel are not going to fit well
with GPS measurements. It is always going to be a semblance
of a rubber sheet operation to get modern GPS measurements
on an old map datum. It is better to convert the maps if
you can. The old maps are not going to be made much worse
by running them through the coordinate conversion to a
modern datum. You will make your GPS measurements less
accurate by converting your new data to the old map datum.

It is always going to be important to ground truth your GPS
tools against local geodetic control. It is the only way to
know if the specifications and your skill with the tools
will provide the right answer.

It is equally important to ground truth your map against
your GPS tools.

Never assume that a map is correct because of the source or
the metadata. I have seen official Oklahoma maps that
messed up the coordinates. Never assume that a commercial
product is correct because of a manufacturer's
specifications or a salesman's claim. I have seen steel
tapes, transits, and total stations that gave wrong answers.
What makes a GPS receiver give the right answer? A life
saving lesson learned in Vietnam: Never think that the tool
that somebody just handed to you is ready for use!

One of Oklahoma's sons has been using a MobileMapper in
Afghanistan. I am pleased to say that he is home in Tulsa
once again. As he and I started working together our first
discussions were about his maps and how they related to his
GPS receiver. He scanned some paper maps and downloaded
some shape files that provided multiple sets of coordinates
for the same spot on the ground and none of them matched the
modern digital aerial photography. He very successfully
used his GPS receiver with the aerial photography and was
able to get a good feel for the relationships of points on
the old maps to his GPS measurements. Professor Cliff
Mugnier was concerned that we would get them all killed with
what we did. He wrote an article for a national magazine
that talked about what a bad idea it was. That was
contradicted by the stories that came back from Afghanistan.
Is John alive today because he had a MobileMapper? Probably
not. My experience says that an infantry officer is most
dependent on his rifle and the soldiers with him. Did the
MobileMapper help him engage the enemy? What he and his
comrades were able to tell me says that it did.

Ground truth. It is about maps, hardware, software, and map
users too.

I hope that was not too boring?

A copy of this message is going to John, the infantry
officer, who just came back from Afghanistan. He recently
asked how he could get more involved in GIS work and
mapping. I told him to find user groups and get started
with some education on the subject. The Oklahoma GIS
Community is a good example of where he needs to look.

Phil


Phil Stevenson
N37°21'04.2" W121°56'06.4"
pstevenson@sbcglobal.net - http://www.nfork.com

http://www.thalesnavigation.com
ICQ# 20819542 - AIM PhilNF1
MSN PhilNF1@hotmail.com


-----Original Message-----
From: okgis-bounces@gis.gis.ou.edu
[mailto:okgis-bounces@gis.gis.ou.edu] On Behalf Of Peter
Earls
Sent: Thursday, September 30, 2004 9:57 AM
To: Oklahoma GIS Community
Subject: Re: [OKGIS] 1995 DOQ Accuracy?


David,

This is a delayed response to your question. I have tried
to get at that question in a couple of ways.

One way was by comparing DOQ coordinates with a GPS receiver
for numerous locations in TNC's Tallgrass Prairie Preserve
(TGPP) in northern Osage County that I could identify just
about to the pixel on the DOQs (BIL format; ~8 contiguous
DOQs involved). I collected ~200 points for each location
with a Trimble GeoExplorer II (no external antenna but all
locations in unobstructed prairie). I used Trimble's
Pathfinder software to transform the geographic coordinates
(WGS84) to UTM NAD27 and then average the points and get a
standard error. For all locations the mean of the GPS
points coordinates was 2-12m S.S.E. of the DOQ coordinates
for the same location when I used the NAD27 DOQ headers.
The standard errors for each set of ~200 points (a measure
of GPS coordinate reproducibility over a 3-4 minute period)
were in the range of 1-9m and mostly 3-4m. This is all
questionable though in regards to verifying the accuracy of
the DOQs, so I subsequently found out (more below). I also
compared GPS coordinates for the same location on different
days on a very limited basis (1 point only and just 2-3
days) and found the average of ~200 points for each day to
be pretty close (about a meter's difference as best I can
remember).

The other comparison I have done is to check locations on
adjacent DOQs where they overlap. In the TGPP these
locations were always pretty close to overlapping - maybe a
couple of meters off at most as best as I can remember. In
TNC's Nickel Preserve in the more rugged terrain of
northeastern OK, locations were generally much more
separated on adjacent DOQs although I can't remember the
numbers now. This didn't seem too surprising, given the
difficulty with orthorectification due to relief distortion
that would be in imagery taken at a relatively low flying
height compared to satellite imagery. One could do this
comparison for locations where 4 DOQs overlap at their
corners and have 4 pairs of coordinates do some statistical
analysis with for what this would be worth.

Phil: I wish I could be organized to document my testing
like you do. In the MobileMapper PDF, were you comparing
geographic coordinates for the National Spatial Reference
System survey monument (was this a HARN monument?) test? If
so, can you get 'bona fide' (National Geodetic Survey?) UTM
NAD27 coordinates for the monument? If so, compare the
MobileMapper UTM NAD27 coordinates to those for the monument
and see if they match within a couple of meters or so. All 3
manufacturers of GPS receivers I have tested (Trimble,
Garmin, and
Magellan) give consistently different UTM NAD27 coordinates
for the same geographic coordinates compared to ESRI's
software (ArcInfo and ArcView 3.x) and the UTMS program from
the NGS web site. The GPS receivers in the field and also
Trimble's software transformed to a location ~27m west of
what the ESRI and NGS software gave (these latter 2 matched
just about exactly except for possibly round-off error -
well less than a meter's difference). You once told me that
NGS sets the standard for doing coordinate transformation,
so presumably the GPS receivers are incorrect. But they are
so consistent that I wonder if I am missing some fine
detail.

I have tested this in the TGPP and on the OSU campus with
consistent results. I have wanted to test this in a
different UTM zone and in UTM NAD83 but haven't done yet.

Back to 1995 DOQ accuracy, the GPS receivers' coordinates
match the DOQ coordinates considerably closer than using
ESRI software to transform the geographic coordinates from
the GPS receivers. The ESRI transformed coordinates wind up
being 25-30m west of overlapping locations on the DOQs.

Phil is right about needing to ground truth the DOQs. I
think I have in my notes somewhere more exact numbers about
those DOQ coordinate comparison tests that I could dig out
if they would be useful.

- Pete

============================================================
======
Peter Earls
Department of Botany
Oklahoma State University

Phil Stevenson wrote:
> David,
>
> This is another example of the need for field work to
ground truth
> those
> maps.
>
> Some seem to be quite good and match accurate ground
measurements with
> excellent results.
>
> Some are not so good.
>
> They must be checked against ground measurements in the
area where you
> are
> working to be able to claim certain levels of accuracy.
>
> The first job is to test the measurement tools against the
NGS survey
> control.
>
> Then make measurements to points on the ground that are
easily matched
> with points on the photos.
>
> You can see some of my own tinkering with this in a file
named
>
> MobileMapper.pdf
>
> available from the MobileMapper folder in the training
folder on the
> ftp
> server at
>
> ftp://ftp.thalesnavigation.com
>
> Phil
>
> Phil Stevenson N37°21'04.2" W121°56'06.4"
pstevenson@sbcglobal.net -
> http://www.nfork.com
>
> http://www.thalesnavigation.com ICQ# 20819542 - AIM
PhilNF1 MSN
> PhilNF1@hotmail.com
>
>
> -----Original Message----- From:
okgis-bounces@gis.gis.ou.edu
> [mailto:okgis-bounces@gis.gis.ou.edu] On Behalf Of David
Wheelock Sent:
> Thursday, September 16, 2004 1:50 PM To: OKGIS List
Subject: [OKGIS] 1995
> DOQ Accuracy?
>
>
> What is the horizontal positional accuracy of the 1995
DOQ's?
>
> I have a user asking me for the horizontal positional
accuracy of the
> 1995
> DOQ's, and hence the NRCS 2003's that are based on the
1995 control. When
> I look in the metadata for the 1995 DOQ's it doesn't give
a clear numerical
> value for this. I've looked in the headers for some of
the quads and see
> values for RMSE (Root mean squared error) between 1 and
2.5 meters. Are
> they really that accurate? I believe I've seen a lot more
variation than
> than.
>
> David Wheelock Application Specialist
david@oda.state.ok.us
>
> Oklahoma Department of Agriculture, Food and Forestry 2800
North Lincoln
> Boulevard Oklahoma City, OK 73105-4298 Phone: (405)
522-5477 FAX:
> (405) 522-3590 http://www.oda.state.ok.us/
>
>
>
>










  • RE: [OKGIS] 1995 DOQ Accuracy?, Phil Stevenson, 10/03/2004

Archive powered by MHonArc 2.6.16.

§