QC on Network RTK
> Not really. The above data shows perfectly well that that method would likely not meet many accuracy specs for *relative* positioning, such as the ALTA spec and the Texas minimum professional standards.
>
Really? Surely you're only referring to the points under moderate canopy-right?
Also, I've had similar results with shorter observation times and the second set of observations separated by as little as 25-30 minutes. It's a matter of knowing your equipment and what it takes to get the desired results.
QC on Network RTK
It says..
>....Relative Positional Precision is estimated by the results of a correctly weighted least squares adjustment of the survey.
and not
>....Relative Positional Precision must be determined by the results of a correctly weighted least squares adjustment of the survey.
Lots of people are doing ALTA's without the benefit of LS. I do not believe that the specs preclude that. Although I certainly agree that LS is the best way.
I agree with you more than I disagree with you on this, Kent. There is a place for the prudent surveyor to use RTN, but I would never be happy with just occupying a point a single time - by any method - and calling it good*. But I agree with Gavin that if you occupy a point multiple times and keep getting the same answer -within acceptable tolerances- then you have achieved the quality control a prudent surveyor craves.
*Topo points excepted. Here we have an example of points whose quality we accept without benefit of the LS report. We just know that there is a certain quality we can expect if certain procedures are followed. And we apply different flavors of quality control to topo points and the maps and surfaces produced from them.
Besides, IMO, it's largely a moot argument. I wouldn't attempt to use RTN to tie a volume of points because the cell modem connection isn't robust enough. It keeps dropping connection when you are moving - unless you are in an urban area with strong cell service (which isn't currently overtaxed by other users). And if you are in an urban area you likely have lousy sky for RTK anyway. So I'd usually skip the RTN for productivity reasons alone when I have a lot of points to tie.
QC on Network RTK
> It says..
> >....Relative Positional Precision is estimated by the results of a correctly weighted least squares adjustment of the survey.
> and not
> >....Relative Positional Precision must be determined by the results of a correctly weighted least squares adjustment of the survey.
> Lots of people are doing ALTA's without the benefit of LS. I do not believe that the specs preclude that. Although I certainly agree that LS is the best way.
Well, as I read the ALTA spec it pretty much requires the surveyor to be able to estimate the length of the semi-major axis of the 95% confidence relative error ellipses of all pairs of boundary monuments positioned by the survey. If you aren't doing that by running all of the measurements by which boundary monuments are positioned through a weighted least squares adjustment, then what options are open?
If you're attempting to estimate the 95% confidence relative error ellipses of pairs of points positioned via black box methods, then it seems to me that the main option is the fairly inefficient one of doing statistical testing on independent repeat measurements to either (a) estimate their uncertainties or (b) show that some guesstimate of the uncertainties isn't obviously wrong.
What does a careful surveyor do when you have repeats with excessively large residuals? I think he or she is obliged to verify that it isn't just that one point that is the problem, so that means more repeats on more points. I'm not seeing that as highly efficient.
NTRIP mountpoints
"Calgon, take me away!" 😉
Just a reply to the whole thread.
RTK probably won't put you under a dime every time. There are places and measurements where it won't meet the spec. For most cadastral work I think it is plenty good, maybe not in places like New York City. Most clients probably don't care about a couple of tenths (more like a foot where I work). If it came right down to stopping all cadatral work using GPS because it is not under a dime spec, I think that the specs would just be changed to make it OK, say a couple of tenths. Very few are going to pay 4 to 5 times for a land survey to get it under a LS dime.
I've gone back to RTK surveys we done over ten years ago. Usually it puts us right back on the cap but there seems to always be a couple cm floating around, always has been since I been doing it. Get in some unfavorable conditions and a fix can vary up to a half foot or so. A bad fix will kick it up to about 5 feet.
For long distances I don't think conventional can spec with GPS. Be hard to traverse for 50 miles and be within a couple cm, but I've never tried it. Pretty easy with GPS too get a couple cm in 50 miles but not with RTK.
Ya'll is fighting within the choir. Most of the congregation don't give a crap!
gshrock
> Yes, I was warned.
Gavin,
Your generosity with your expertise and your time has always impressed me. Myself, and my company are grateful. While it appears that some are unable to learn from others, I appreciate your efforts in standing up for a professional and scientific approach to this issue.
Truth is, you and the others that work on the WSRN have probably made more money for WA surveyors than anyone knows, and you and the efforts of you colleagues are increasing the value and stability of the cadastre in this state.
Good on ya!
QC on Network RTK
> Well, I really don't know why it should be like pulling teeth to have a discussion of survey accuracy. Some of us care about it and think it actually matters.
Kent,
You really need to stop this. This statement demonstrates your ignorance of the issues and the technology. My real world experience has demonstrated to me what we already know from the science:
RTK/RTN is far more accurate than any other method of measurement available to a land surveyor over most distances.
RTK/RTN are not as precise over small distances as a total station with EDM.
EDM's are not as precise as a chain over even smaller distances.
Get where I am going here? By your logic we should all still be pulling chain.
You appear to assume that people are idiots and are misusing their tools. It is just as bad to use RTK to stake a grid on a concrete pad as it is to run 5 miles of traverse with a Total Station instead of using RTK.
The right tool for the right job. If you listen to Gavin, you have a chance to learn from a real life International expert. Do you think that you are telling him something he doesn't know or hasn't heard? Unlikely. But it is highly likely he is saying something you haven't heard, and certain he is saying something you do not understand.
-David
Good summary LR
RTK for boundary survey is standard practice and RTN is becoming so. It would be well if the users clearly understood how the system is obtaining positions. In our RTN if you set up for 24 hours and take 1 second observations you will get a spread of about 4-5cm horiz and 8-10 cm vert. From what I have been able to learn this is not that uncommon in any RTN truth be told. The 95% certainty of that data set is about 1 cm horiz and 2cm vert. If you happen to be on a published NAVD83(2011) FBN or CBN station you will be within a cm of published horiz and 0 to 10 cm ortho height on first or second order marks using Geoid 09 to compute it. On GPS observed marks you will be at about 1cm on the published ellips height. Now before someone comes back with "so you have to observe each point for 24 hours to get those results." take, the fact is you don't. But you do have to take at least two observations and the optimal spread is 12 hours. (the constellation is 180 different then - a GPS double center). You could throw a few more in at the 6 hour marks as well for the warm fuzzy. Ngs suggests 180 epoch observations. Our results have shown there is nothing much to be gained in an RTN by 180 or 10 epoch observations. The 180 came from the days of fast static. Times have changed. The results of your RTN may vary but not that much.
I should also mention that when you observe a point for 24 hours you will notice spikes in the data that result in the higher spread. These spikes are essentially outliers in the data set that are thrown out in the 95% criteria. The cause of these spikes in the RTN in undetermined and unpredictable. Could be cell, could be absolute calibration on reference stations, could be multipath conditions somewhere in the system that come and go. could be change in constellation, could be weather, could be a combination. Our practice is to take enough observations on control to be able to catch a spike in the data. For topo live with the spikes or put the RTN away.
No QC on VRS positions?
Nope, saw a crew doing EFB (4-man crew no less) just a few months ago.
1-sits in a chair next to instrument and runs DC
2-turns angles
3-runs the rod
4-chief paints the road for shots.
hated that software and threatened to quit when I was asked to join the traveling EFB crew. mind numbing non-surveying.
Kent
> > The RTN (that I don't use because it's not available here) isn't all that different from RTK. So, there are MANY different methods that one can use to prove or disprove the RTN point, similar to how one would prove or disprove the RTK shot. Proving ANY vector is as simple as remeasuring it (which you already alluded to above), so I don't see any problem with connecting to the grid, via RTN, which essentially is doing the same thing you do, only you use OPUS, the the bases are already connected.
>
> Well, the real question isn't whether RTK positions are error free, but what the errors in the RTK-derived positions are. Professional-quality measurements require that a surveyor have some reliable estimate of uncertainties in measurements, something better than "no one has proven it wrong yet".
>
> > Sure, you've run up on those quicky dicky surveyors using RTN and RTK under the canopy, or in a manner not congruent with the manufacture specifications. That in and of itself, doesn't mean the gear is faulty, but operator error.
>
> Well, a professional surveyor should be as interested in measurement processes as measurement results. Black box precesses strike me as inherently problematic.
>
>
> > We have used RTK, with HIGH success for 17 years since 1995. Do we get bad answers from time to time, hell yes.
>
> Isn't the proper statement "We FIND bad answers from time to time"? You can't find what you don't check, right?
>
> > We did this on a subdivision recently. We tied the boundary in with RTK (mostly with a total station from pairs set in the open). All points were shot twice for a positional tolerance and averaged. Some were shot three times under different constellations. The answers were (mostly) within the specs, except a few, which is where the third shot came in.
>
> So, is this an advertisement for RTK? I hope not. No, the missing piece in your method is being able to reliably characterize the uncertainties in the monuments you've placed and to verify that their relative positions meet the minimum technical standards. From what you described, I don't think you got there.
Sorry bud, but read it again. Multiple shots that agree can and do generate a positional tolerance, as required by the board.
Good try though. FWIW, the specs say 95% confidence at the centimeter level.
Kent
> Sorry bud, but read it again. Multiple shots that agree can and do generate a positional tolerance, as required by the board.
Sure, if you do enough repeats, you can measure angles with a 1-minute transit with an uncertainty of just a couple of seconds. No question about it. The problem that you still need to deal with is evaluating and documenting the quality of the results. Repeat RTK positions at different times certainly would give a basis for testing any assumptions you've made about the quality of the process.
>FWIW, the specs say 95% confidence at the centimeter level.
Which RTK system do you have in mind? An RMS error is not a 2-sigma uncertainty. You know that, right?