This is mildly interesting. One of the parts of a project I'm working on is resurveying several lots of about 200 acres average size extending in aggregate about a mile N-S and about 2.6 miles E-W. They are part of a subdivision that was made about twelve years ago of a tract about 4 miles E-W and 4 miles N-S, all evidently surveyed on the ground at the time of platting, which is as it should be in Texas.
The subdivision was apparently laid out with RTK GPS in a local grid system that was oriented to geodetic North at some unspecified point within the subdivision. From positioning the points in the list below within the 1 mi. x 2.6 mi. subarea, I determined that the North of the local grid in which the subdivision was surveyed most likely had a bearing of N1°33'05"E in the Texas Coordinate System of 1983 (South Central Zone). Considering that the convergence varies between -1°32'42" and -1°34'40" at the maximum Easterly and Westly extents of the subdivision, that means that somewhere within the subdivision the "North" of the local grid does coincide with geodetic North.
That is probably readily explained by the grid having been a local projection generated for a "HERE" position occupied by an RTK base. The grid was oriented geodetic North at the meridian of the "HERE" position.
The puzzle is how the scale of the local projection got so oddly distorted. It is approximately correct for distances reduced to the GRS80 ellipsoid. I determined the coordinates of the following 18 boundary markers in the 1 mi. x 2.6 mi. subarea via short (<2 miles) rapid static L1 GPS vectors with standard errors of about +/-0.03 ft. in N and E and used those coordinates to compute the parameters that would transform the coordinates of the same markers in a local system calculated using the survey data noted on the subdivision plat into the Texas Coordinate System of 1983.
The rotation parameter was, as I mentioned, +1°33'05" to transform the Y-axis of the local system to the orientation of grid North of the Texas Coordinate System. No surprise there, really. However, the scale parameter of the transformation was odd. The distances noted on the plat needed to be scaled by 0.999970 to reduce them to grid distances on the projection surface of the Texas Coordinate System of 1983. Ellipsoid heights of ground surface varied across the project between about 2245 ft. and 2457 ft., so obviously the height scale factors are much different than 0.999970.
0.999970 is very nearly the map projection scale factor (i.e. scale factor relating ellipsoid length to grid length) at a high point on the project that would have been a likely candidate to set up the RTK base. I assume that the early Trimble RTK controllers offered some novel ways to distort an RTK survey for posterity.
The table below shows that after correcting the distances noted on the map by a scale factor of 0.999970 derived from a best fit transformation, to transform the local system to the Texas Coordinate System of 1983, the apparent accuracy of the RTK survey was about what one would expect for RTK GPS, i.e. +/-0.10 ft. standard error in N and in E components. In computing the best fit transformation, I didn't include a number of RTK wild shots that had much larger residual errors in them, but the rejects accounted for about 20% of the positions checked.
[pre]
Residuals after Helmert
Transformation of Coords
Calc'd from Survey Data
on Subdivision Plat
Pt Residuals (ft)
No. N E
2023 -0.02 -0.14
2056 -0.03 0.12
2057 -0.02 0.03
2059 -0.06 0.03
2063 0.04 0.06
2065 0.07 0.13
2066 -0.15 0.23
2069 -0.01 -0.09
2070 0.00 -0.12
2071 -0.04 -0.04
2072 -0.04 -0.04
2073 -0.05 0.09
2077 -0.01 -0.10
2078 0.01 -0.11
2079 -0.01 -0.13
2080 0.02 -0.06
2081 0.10 0.13
2082 0.17 0.02
----- ------
Mean 0.000 0.000
s.e. 0.068 0.109
[/pre]
"... about what one would expect for RTK GPS, i.e. +/-0.10 ft. ..." Yes, or even twice that.
that would be plenty good for some folks.
how big were the outliers? those are the ones to worry about
The four that I did not include in computing the best fit transformation had absolute errors of position of :
0.59 ft.
0.87 ft.
1.27 ft. and
2.23 ft., respectively.
This site is pretty much optimal for GPS surveying. Wide open sky at all points except for one where I had to cut down a small hackberry (that may well have grown up during up since the subdivision was laid out).
Seven of the boundary markers were quite close together, all fitting in a box about 120 ft. E-W and 780 ft. N-S.
As may be seen, the random errors among that cluster of points were quite nearly the same in magnitudes as those in the entire set across the project area.
[pre]
Residuals after Helmert
Transformation of Coords
Calc'd from Survey Data
on Subdivision Plat
Pt. Residuals (ft)
No. N E
2065 0.07 0.13
2066 -0.15 0.23
2078 0.01 -0.11
2079 -0.01 -0.13
2080 0.02 -0.06
2081 0.10 0.13
2082 0.17 0.02
----- ------
Mean 0.030 0.030
s.e. 0.100 0.137
[/pre]
These large random errors in the independently surveyed positions resulted in discrepancies of more than 0.20 ft. between points less than 100 ft. apart, which is unimpressive.
Your conclusions are entirely plausible for RTK technology from 12 years ago. Current systems, however, should perform much better - especially in the landscape you describe.
When using best practices, centimeter or better for boundary evidence is the current norm.
That's not to say that there aren't plenty "coordinate cowboys" that shouldn't even be surveying, let-alone using GPS.
Within an area of four miles by four miles I would hate to see what some people would put on the ground using a total station, modern GPS actually elevates some peoples product.
> Your conclusions are entirely plausible for RTK technology from 12 years ago. Current systems, however, should perform much better - especially in the landscape you describe.
>
> When using best practices, centimeter or better for boundary evidence is the current norm.
Yes, the data sheet for a Leica Viva GNSS GS12 receiver quotes a single baseline (<30km) accuracy of +/-8mm + 1ppm (rms error).
Considering that the average baseline length from where the base station probably was set up was about 5km, that would have meant that the baselines from it with the GS12 should have had rms uncertainties of +/-1.3cm under what Leica calls "normal to favorable conditions". That would be on a Wednesday. Presumably in reality that means uncertainties of +/-2cm or more for the rest of the field week, which is still better than the original layout of the subdivision.
Naturally, the better practice would have been to have added conventional measurements to connect the groups of boundary markers set relatively close together and to have adjusted the RTK vectors in combination with those conventional measurements. That would have tidied things up nicely and would not have taken much effort. I'd think that would still be true for RTK surveys.
Pretty much exactly what I do. Except you're also right about the ppm, best practice is to set a few static points across the job.
RTK is only ever as good as the framework from which it propagates.
> Pretty much exactly what I do. Except you're also right about the ppm, best practice is to set a few static points across the job.
I think that the main problem with RTK is going to continue to be users not following best practices. Once the base is set up somewhere, there seems to be an almost irresistable force requiring that everything within radio range be surveyed from that base that can be. The other natural law seems to be that if you can't get the rover pole actually on some mark, then as close as the pole can be held will be identical with the actual position of the mark sought to be surveyed.
> Seven of the boundary markers were quite close together, all fitting in a box about 120 ft. E-W and 780 ft. N-S.
>
> As may be seen, the random errors among that cluster of points were quite nearly the same in magnitudes as those in the entire set across the project area.
>
> [pre]
>
> Residuals after Helmert
> Transformation of Coords
> Calc'd from Survey Data
> on Subdivision Plat
>
> Pt. Residuals (ft)
> No. N E
>
> 2065 0.07 0.13
> 2066 -0.15 0.23
> 2078 0.01 -0.11
> 2079 -0.01 -0.13
> 2080 0.02 -0.06
> 2081 0.10 0.13
> 2082 0.17 0.02
> ----- ------
> Mean 0.030 0.030
> s.e. 0.100 0.137
>
> [/pre]
>
> These large random errors in the independently surveyed positions resulted in discrepancies of more than 0.20 ft. between points less than 100 ft. apart, which is unimpressive.
So the precision is lower than you like, but Iam guessing the overall accuracy of those points relating to the original monuments of the larger subdivision is quite good?
(Probably greater than 1:100,000?)
> So the precision is lower than you like, but Iam guessing the overall accuracy of those points relating to the original monuments of the larger subdivision is quite good?
Actually, the relative accuracy to other nearby monuments is how survey accuracy was to be determined and was a fail. It only makes sense that you would still want to test for relative accuracy in the neighborhood of a marker because when markers are to be replaced, those nearest will provide the logical basis for replacement, not some monument on the far side of the subdivision.
The other natural law seems to be that if you can't get the rover pole actually on some mark, then as close as the pole can be held will be identical with the actual position of the mark sought to be surveyed.
Yes, of course the came can be said for a prism pole used with a total station.
At least one of the new GPS receivers will allow the user to tilt the rod away from the obstruction while having the point of the rod resting on the monument, sense the tilt and calculate the point in it's correct position, pretty cool!!!;-)
> At least one of the new GPS receivers will allow the user to tilt the rod away from the obstruction while having the point of the rod resting on the monument, sense the tilt and calculate the point in it's correct position, pretty cool!!!
I'll wager that there is yet another category of blunders that some malfunctioning tilt sensor will create.
I am somewhat skeptical of their use on critical points.
If I understand correctly the tilt compensation is completely beholden to the accuracy of the internal compass (magnetometer?), and subsequently completely vulnerable to local attraction.
You have to "calibrate" an R10 by flipping and rotation it before use, also the compass is corrected by GPS heading - but I still would have to think that it is vulnerable to local attraction.
> If I understand correctly the tilt compensation is completely beholden to the accuracy of the internal compass (magnetometer?), and subsequently completely vulnerable to local attraction.
Yes, and considering that the majority of obstructed survey markers I deal with that have to be located from offsets are along wire fences, I can see using a compass to determine the direction of tilt as the seed from which a new crop of FUBARs will grow.
> I am somewhat skeptical of their use on critical points.
I don't trust the tilt feature, either. With the Javad Triumph-LS there's a box in the display that shows, in what I take to be dimensionless units, how far off plumb the receiver is. It consists of two digits, one indicating fore-and-aft tilt and one indicating side-to-side tilt. When both are zero the display background is green and the receiver considered plumb. As you tilt the receiver and one or both digits goes above zero, the display background turns black (there may be options to set the colors, but this is the way it's set on my unit).
On a 2.000 m pole, the difference between a green/zero/"plumb" reading and a 1 or -1 is about 15 mm. So "plumb" is anything inside a 3 cm circle of the point. That may be fine for dirt shots, but I see no reason to introduce yet another error source of that magnitude for anything else.
I'm still trying to figure out how RTK fits into my business model, but as things stand right now I don't expect to use the Javad tilt-o-meter for anything other than rough work.
> > So the precision is lower than you like, but Iam guessing the overall accuracy of those points relating to the original monuments of the larger subdivision is quite good?
>
> Actually, the relative accuracy to other nearby monuments is how survey accuracy was to be determined and was a fail. It only makes sense that you would still want to test for relative accuracy in the neighborhood of a marker because when markers are to be replaced, those nearest will provide the logical basis for replacement, not some monument on the far side of the subdivision.
I understand. I wasn't making an argument, asking a real question.
This is one of the reasons that I am not a proponent of using RTK to set center-line monuments that will likely be used to re-establish corner monuments in the future. It is also a reason to not simply set property corner monuments within a plat using RTK without checks.
I know that each corner monument is as much a plat monument as the road monuments, but we have always used a higher standard to set the road monuments, simply because they are likely to be used to re-establish lines when corners go missing.
> This is one of the reasons that I am not a proponent of using RTK to set center-line monuments that will likely be used to re-establish corner monuments in the future. It is also a reason to not simply set property corner monuments within a plat using RTK without checks.
I'll confess that I can't imagine staking a small lot subdivision with RTK unless the local standards are extremely loose. It seems so much better practice to set static GPS control adjusted as a network with conventional survey date and then just set the lot and centerline monuments from that control. That takes the worry out of being close.
I'll confess that I can't imagine staking a small lot subdivision with RTK unless the local standards are extremely loose. It seems so much better practice to set static GPS control adjusted as a network with conventional survey date and then just set the lot and centerline monuments from that control. That takes the worry out of being close.
I don't understand it either, why not use your instrument to set close fitting property corners, RTK just doesn't have the same accuracy. And it really doesn't add to the costs. I've even seen buildings being staked with it:-(
> I don't understand it either, why not use your instrument to set close fitting property corners, RTK just doesn't have the same accuracy. And it really doesn't add to the costs. I've even seen buildings being staked with it
Next up: ANCHOR BOLTS! "Oh, I figured that the iron workers were going to set them where they wanted to, anyway. So my RTK stakeout was just to get them started."
> Next up: ANCHOR BOLTS! "Oh, I figured that the iron workers were going to set them where they wanted to, anyway. So my RTK stakeout was just to get them started."
Years ago, I staked out a building. (on a tangent here, nothing to do with RTK). We completed staking building that day, and had to come out to the same job site the next day. The contractor came out and complemented us on our stakeout job. (We always did as precise of work checking the hubs and the diagonals with a chain and taping level etc.) We generally left with confidence. I think most guys do this when staking out a building (?).
But what the contractor said that killed me was that usually he has to come out behing the surveyors and check their work and that this was the first time he didn't have to adjust the hubs. I wonder how many times he took a perfectly square building stakeout and adjusted it to be out of square.
