[aprssig] Re: TIGER maps what-s the problem?

[KC2MMI (Jared)]

Chris-

"If a data source like TIGER is used, how will a programmer know
whether there are errors " Someone needs to decide whether to question data,
accept it without question, or validate it. That could be asked of any
programming task.

"or how big those errors are?"
 One answer to both of these questions, would be to contact the source of the
data, in this case the USCB, and ask them directly "What is the accuracy of
these data? What uses are they suitable for?" And I don't individually or
particularly fault James here, I suspect most of us wouldn't have thought to
question the USCB this way either. A real sharp experienced lab scientist or
statistician might do this kind of questioning normally though. "Is this data
suitable for this purpose?" The author says not.

 Once we know there are problems in the data set, it becomes possible to at
least approximate the size of the errors by the cartographic process called
"ground conformation". Loosely, that means let's go out and see how reality
matches the map, and make corrections accordingly. So, if a user in the midwest
reports that his street is 1/4 mile off the position it shows on the TIGER
maps...we can look into why (usually a GPS datum error, a lat/lon conversion
error, etc.) and once we've looked at everything else, we can see if the error
is in the TIGER data set.
 We can also look at other data sets (USGS topo, state survey, etc.) which are
often available online, and see if the positions displayed "jump" when the data
set is changed. On APRSWorld, if I switch to topo or satellite maps, that
actually happens. Typically a 200-250 yard jump (200 meters) in my location,
indicating that either the programs, or the maps, have that much error in them
since they are all being fed with the same co-ordinates (same lat/lon
information all given in DDD.MM.mmm format, or any other consistent format).

"Can a blanket statement be made that they are inaccurate
+/- X-number of feet/meters/miles?"
 Yes, with regard to the accuracy limits of the systems. That is, we know the
typical errors in GPS and WAAS GPS, which are the most likely forms of GPS used
by APRS users. From that, we can say an error of perhaps +- 3 meters (6 meters
total) for WAAS GPS under good circumstances, out to an error more like +-10
meters (20 meters total) for "plain" civilian GPS.( Folks differ on what typical
accuracies are, I stick to 10 meters because it is a pessimistic figure and
that's how I've been taught to navigate. That's not the point here.)
 And No, we can't make a blanket statement unless we've actually got full ground
conformation (damned expensive, USGS or the military probably comes closest).
 And Yes, I would suggest we can come reasonably close, if we collect known
errors and from time to time revise our blanket statement to reflect the reality
we have seen. So, if we get reports of 100-meter errors, and nothing worse, we
can say that's a worse-case error and perhaps use it. If we get reports of
200-meter errors, etcetera.
 That will also involve a judgement call (or more complex programming) since the
errors will also change locally. If some part of "flatland" was well surveyed in
modern times, the data from it are likely to be good. If some other area has a
legacy datum that dates back 400 years (East Coast or CA, etc.) and that legacy
has been *kept*, we need to know that.

 Case in point, Maptech, who were government partners and arguably good
mapmakers since they work from the government source data from the USGS and
NIMA, etc. Maptech publishes the source map datum on every map they make,
digital or paper format. But if you use their online map server, it seamlessly
tiles multiple maps from multiple datums--and gives you no online indication of
which one is in use. Result? You can't tell which datum you are on, so you can't
really be sure of your position. (But that's their free product, not their
commercial ones.)

 Ships do run aground and planes do crash due to "simple" datum errors, they are
made even in the professional world. One of the private maps (Adirondack
Mountain Club, if I recall) for hikers in the Northeast used to be copied from
the USGS topos, which dated back to logging surveys in the late 1800's for that
area (again, iirc). The private map for many years literally said "NOTE TO ENEMY
BOMBERS:" in the border, noting that a certain mountain was 6/10's of a mile
away from where it was shown on the map. Well, for logging maps from the 1800's,
that was good enough. For a forested preserve area pre-GPS, that might also be
good enough. Today? That could be a problem, as a SAR team goes "here" instead
of "there". (I assume that quad has long since been corrected.)

" how that issue morphed into what appears
to be some gripe about the use of TIGER data."
 I've complained about the problem of using TIGER maps for navigation since I
found out about it, about two years ago. I found out about it by doing my
homework--ground conformation--and asking questions until I got to the source
and got the explanation directly from the USCB. I investigated with the aid of
several people, some academic, some state survey, and examined the data chain
from my GPS itself to the map source itself--TIGER. And we confirmed that the
source of error was the TIGER maps. Despite the wide acceptance of these great
free maps in ham circles, the folks who generously let the public use them also
quietly say that we are MISusing them, and buried in some 250? pages of
documentation for the TIGER maps, they do note that. It took some work (multiple
calls) to get someone at the USCB who understood the issue and who eventually
said "Oh, that?..." with full awareness of the problem. The people who answer
the phones at the "front desk" aren't cartographers.<G>

"One of the affirmative defenses against nits about the accuracy
of the map database would be:  that's all that is available for
free usage,  you want better data you have to peel off some
real money.  Is this not true?"

Absolutely! And my point has been that that would be sufficient, as long as we
are at least warned "This data is ..." with some warning about the fact that the
*displayed* accuracy is indeed false and misleading, regardless of the reason
why that is.

Of course, if we can easily make the display more accurate, or, once we know it
is inaccurate we continue to deceive the user with false accuracy, shouldn't we
try to do better? Some say no, I say yes.

How would you like to see a highway sign that said "Speed limit, vaguely
50-70mph". We can do better.

Why settle for mundane, when "really good" isn't that much harder to get?

Isn't that part of what ham radio is about? Advancing the state of the art?