Ok, the post below brings up the following discussion. Dont worry, I have on my asbestos pants.
Situation: Standard PLSS section, platting the NE 1/4 of SW 1/4 as a residential subdivision. All corners were found as originally set. Some relief and some foliage, but every point is directly observable via GPS.
Now, we run a traverse through each original corner, and balance it using least squares software.
Then we come off that traverse, and run control through the subdivision for use in setting corners and centerline monuments.
Or...we set every single one of those corners using RTK and best practices.
I maintain the following:
If you care only about the accuracy relating to the original controlling corners, and the set corners, the RTK should be the most accurate and precise option.
If you care most about the relationship of neighboring set monuments, the traverse method will produce the most accurate results.
My point is that accuracy is dependent upon the standard used to define it.
I like the traverse method for many reasons, but recognize the validity of RTK for bounday purposes.
Certain people find it irritating, perhaps, when two found corners 100' apart do not match record by 0.10'
I submit that irritation is the result of a deep suspicion that many may not be using RTK correctly. I too share that suspicion, BTW.
Please feel free to correct my many mistakes!
> Please feel free to correct my many mistakes!
That's not my place. I do agree with you there are many that are suspicious of the many different ways that GPS can be used to provide accurate locations.
I might also explain what I feel is the difference in accuracy and precision. Accurately hitting a point means you are close to the center of the calculated position, even if all of the shots are on different sides of the calculated point. Precisely hitting a point means all the shots are closely spaced, even if they are very far from the actual center of the point.
Discussions of the reliability of positional mathematical consistency will quickly cause my eyes to glaze over. But as users of technology that exceed our understanding, we still maintain a responsibility to qualify our work within a specific range of predictable error.
I use RTK daily. I also use conventional line-of-sight measurement daily. And while I do not look upon RTK locations as a source of flagrant and unpredictable error; I believe performing a survey with only RTK locationing is very similar to running a bench loop with merely side shots and no checks. You simply have no way of verifying or qualifying your results. The numbers may very well be good, but how can you tell unless you can independently verify a location you have established?
:good:
> Certain people find it irritating, perhaps, when two found corners 100' apart do not match record by 0.10'
0.10' is really not a big deal in many situations, but in an Urban location it is sloppy.
However when I find that much floating around in 100' in Any location I jump to the suspicion that all of the work is slipshod crud and begin to look deeper for gross errors in boundary determination.
measuring is the easy part. If the surveyor can't handle that with ease then I assume the rest of the work is highly suspect.
"the rest of the work" = research, reconnaissance, analysis, documentation, Honesty
First off, I do believe RTK is good and can be checked. How do you know your total station is shooting good distances and turning good angles? By checking things. Same with RTK.
Second, in the example above, RTK in the best case will approach what you can do with a total station in terms of the true position of the points. I think you have to get into a much larger area before RTK GPS starts to give better results than a total station. I would use GPS to break down a quarter section but I wouldn't use it to layout a subdivision.
Yesterdays RTK: have a job splitting off a small 300Ac piece, it is in four different sections so a bit of a pain there getting permission and all.
It has a Highway running just a 1/4 mile south of the parcel and I did a DOT metric job almost twenty years ago along it.
They supplied control and we checked it; the control was probably done in 95 the job# is 98.
Most of the control is gone but it would be nice to find a couple of them, so PC loaded up the original projection, sat on a high hill on site and did a "here" to get started, found two remaining control points and this is what he located:
record-AZ 70-34-26, 4108.84'
measured-AZ 70-34-28, 4108.82'
This is mixing 1995 static (DOT) with 2015 RTK:
shifted over the base point and loaded it into TGO last nite and the job is good to go;-)
> > Please feel free to correct my many mistakes!
>
> That's not my place. I do agree with you there are many that are suspicious of the many different ways that GPS can be used to provide accurate locations.
>
> I might also explain what I feel is the difference in accuracy and precision. Accurately hitting a point means you are close to the center of the calculated position, even if all of the shots are on different sides of the calculated point. Precisely hitting a point means all the shots are closely spaced, even if they are very far from the actual center of the point.
I understand the difference between precision and accuracy. Withing 300', a good total station is both more precise and more accurate than any GPS setup you care to show up with.
But, over a few miles, the GPS is both more precise and accurate.
The LOCAL precision is what trips up the discussion. If you pick any two visible points on the ground to define your precision (which is what the Texan seems to do), than the total station is always more precise.
But, if you were to run 10 traverses in from 5 miles away to stake a single point, vs 10 shots with RTK, the RTK would be more precise (in my experience).
>
> I use RTK daily. I also use conventional line-of-sight measurement daily. And while I do not look upon RTK locations as a source of flagrant and unpredictable error; I believe performing a survey with only RTK locationing is very similar to running a bench loop with merely side shots and no checks. You simply have no way of verifying or qualifying your results. The numbers may very well be good, but how can you tell unless you can independently verify a location you have established?
This is my point exactly. (And BTW, using best practices does provide checks.) And the need for verification stands whether using a total station or RTK, of course.
> If you care most about the relationship of neighboring set monuments, the traverse method will produce the most accurate results.
>
> My point is that accuracy is dependent upon the standard used to define it.
>
> I like the traverse method for many reasons, but recognize the validity of RTK for bounday purposes.
I don't see the use of GPS and conventional measurements as being an either/or choice at all. The way that I use GPS vectors (including PPK which is comparable in accuracy to RTK vectors of the same length) in my practice is to adjust them WITH the conventional measurements. It gives a result that is better than either individually, both high quality network accuracy and high quality local accuracy.
Naturally, to do this means that a surveyor needs to have realistic estimates of the uncertainties in the GPS vectors in a form that can be handled in a least squares adjustment. Just a whole bunch of GPS-derived coordinates is fairly worthless for that purpose.
That 0.10' seems fair for 1990s RTK, but today in open terrain you can certainly expect much better, more like 0.02'.
1
I would point out that using Trimble gear if you enable both the QC1 and QC2 records the covariance information is stored and the RTK vectors can be used in an LSA. There is no difference whatsoever between a PPK vector and an RTK vector with the same number of epochs observed. Trimble (and I would assume Leica) uses the same algorithms for RTK that they use in TBC for PPK.
> That 0.10' seems fair for 1990s RTK, but today in open terrain you can certainly expect much better, more like 0.02'.
In the context of the original post, though, 0.10 ft. referred to the relative positional accuracy of two points independently positioned by RTK.
To get a relative positional uncertainty of 0.02 ft. (6mm) about 68% of the time, you'd have to expect to position each of the two points with an uncertainty of less than 0.01 ft. (3mm) in both N and E components of the RTK position more than 68% of the time. Does any manufacturer claim to deliver RTK results like that over baseline distances of a half mile or more?
> I would point out that using Trimble gear if you enable both the QC1 and QC2 records the covariance information is stored and the RTK vectors can be used in an LSA.
Yes, I'm unclear why any RTK user wouldn't want to store survey results in a form other than vectors with the vector uncertainites (either variances and covariances or and equivalent form such as standard errors and correlations).
> 0.10' is really not a big deal in many situations, but in an Urban location it is sloppy.
Absolutely!
>
> However when I find that much floating around in 100' in Any location I jump to the suspicion that all of the work is slipshod crud and begin to look deeper for gross errors in boundary determination.
That is my first thought, as well. But I remember when I first started, and we had an issue like that, and I called the oldest PLS back in the office. I asked him about missing a mon by a couple of tenths or something like that, and he congratulated me. I thought I had a problem, he thought I had a perfect check.
Perhaps, we need to simply revert to following in the footsteps? If a plat is set using RTK, we follow in the footsteps, much like a plat staked using a tape and 5' transit. Do we get bent out of shape when that plat has a couple of tenths here and there?
Instead of doubting the work, perhaps we first understand what is there?
Maybe it goes back to the notes, as well. Look at any modern plat or survey, and they usually list multiple types of instruments used, total stations and GPS. But which one set this corner?
On that point I agree 100%.
> > That 0.10' seems fair for 1990s RTK, but today in open terrain you can certainly expect much better, more like 0.02'.
>
> In the context of the original post, though, 0.10 ft. referred to the relative positional accuracy of two points independently positioned by RTK.
>
> To get a relative positional uncertainty of 0.02 ft. (6mm) about 68% of the time, you'd have to expect to position each of the two points with an uncertainty of less than 0.01 ft. (3mm) in both N and E components of the RTK position more than 68% of the time. Does any manufacturer claim to deliver RTK results like that over baseline distances of a half mile or more?
I agree with Kent on this, and this is where the precision vs accuracy comes in.
RTK positional accuracy is randomly distributed. Meaning that this rebar is off 0.03' North, and that next one I set on this north-south line is off 0.03' South. That makes for 0.06' total between the bars before I even add in my rod, my current willingness to expend effort to get the bar close, etc.
All of our total stations we use will better that 0.06' with ease. If one didn't, it would be headed for the shop.
And, no GNSS manufacture I am aware of claims 0.01'. I am simply making the argument that in most cases 0.10' between set rebars (or pin or iron or whatever we call it now) is completely adequate.
Is 0.1' acceptable to me in a professional sense when I am out there setting the things with a total station? Not really, it kind of offends my aesthetic sense I suppose.
And, for those of us that follow the subdividing surveyor...whatever he or she used, we need to follow in the footsteps.
Now that Trimble has helped us establish that PPK is equivalent to RTK*, we could ask you to provide some experience on the issue. How well do the inverse distances between PPK pairs and total station observation between those pairs generally agree in your experience?
*Kudos to you for the breakthrough moment yesterday. I could see the light come on over Austin all the way from here in Kilgore. The notion that RTK positioning is no worse than the positioning methodology you've been using for years must be giving you much fodder for contemplation.
> Kudos to you for the breakthrough moment yesterday. I could see the light come on over Austin all the way from here in Kilgore. The notion that RTK positioning is no worse than the positioning methodology you've been using for years must be giving you much fodder for contemplation.
I have no idea whose posts you think you've been reading. I have always recognized that RTK accuracy was at best similar to PPK over similar baseline lenghts and occupations. That is exactly why I know how problematic RTK positioning is in terms of relative accuracy and why I know certain claims of RTK accuracy are most likely more optimism than reality.
The novel wrinkle to RTK, of course, is that one is much more likely to see much worse results from RTK than from PPK because of the real world ways in which RTK is used.
For the record, make that:
"Yes, I'm unclear why any RTK user would want to store survey results in a form other than vectors with the vector uncertainites (either variances and covariances or and equivalent form such as standard errors and correlations)."
No answer for the question then?
