Some recent discussions have made reference to traverse closure. I've always been a bit suspicious of that as a measure of quality. It is tempting to think that just because your closure was 0.04, all the points of the traverse are that good - FALSE. Accidental cancellation of errors along the way is likely. Without redundancy, you don’t really know much about your results.
Small closure error is NOT evidence there is no problem in the traverse, but rather lack of evidence of a problem.
I simulated (in Matlab) thousands of cases under these assumptions:
-Regular hexagon traverse
-No systematic errors or blunders - all gaussian normal measurement errors
-Per leg distance std err=1 unit, angle std err producing 1 unit distance sideways
-Open loop closure was noted
-Then loop was adjusted
-Actual error at midpoint of traverse after adjustment was noted
The scatter diagram shows that the adjusted error at the traverse midpoint is not tightly related to the closure error.
The error at the traverse midpoint is often larger than the closure. The possible ratio of midpoint error to closure is unbounded, with larger error being less likely, just as any statistical error distribution. I saw ratios in the hundreds in my simulation of thousands of cases.
The cumulative distribution of the ratio shows, for instance, that the midpoint error tends to be about half the closure, but will be off by more than the closure in 20% of the cases, over twice in 5% of the cases, and almost 5 times the closure in 1% of the cases.
> I simulated (in Matlab) thousands of cases under these assumptions:
> -Regular hexagon traverse
> -No systematic errors or blunders - all gaussian normal measurement errors
> -Per leg distance std err=1 unit, angle std err producing 1 unit distance sideways
> -Open loop closure was noted
> -Then loop was adjusted
> -Actual error at midpoint of traverse after adjustment was noted
It's worth pointing out that the assumption that there are no blundered measurements would be somewhat optimistic in many cases where the Compass Rule is used for the adjustment. In this case, the primary means of detecting blunders would typically be in:
- error of angular closure and
- absolute error of closure.
Depending upon how the user sets his or her rejection limits for the error of angular closure, there may easily be blundered measurements in the traverse that are tossed into the mix for adjustment.
Likewise, where distances are only measured in one direction, the only means of detecting distance blunders would be in the absolute error of closure. So depending upon the rejection limit for that, distance blunders may be present as well.
It would be sort of interesting to evaluate models with blundered measurements to see how many of the blunders would be captured under different assumptions and rejection limits.
I've always felt that random errors swinging for or against your favor played a part in your closure. However I ran a lot of closed traverses around sections achieving 1:100, 000 closures on a regular basis. We were running Wild T-2's with top mount EDM's shooting to prisms on tribrach / tripod. We turned four sets DDRR. Shot distances to fore site and back site five times each for a total of ten measurements on each leg. Would have to say procedure played a big part.
I agree. There are plenty of procedures one can take that will minimize human error. One thing that I have done that goes a long way to minimize blunders is that I have my field crews take hand notes along side their data collector notes. They complain sometimes, but I also have had them see situations where their first angle doesn't match their second angle by a large amount. Then they see that they were sighted wrong on the instrument. Same thing for shooting forward AND back on each setup of the traverse.
StarNet was right-I had two blunders
On a mixed GNSS/Conventional project...both instances were target height busts.
The first equaled 0.10' difference between foresight and backsight. We measure the height to the bottom of the orange triangle on the target plate then add a tenth. The foresight was to a pole all the way down so I'm sure that is correct, added a tenth to the other and that eliminated that asterisk flag.
The next was harder to see. I suspected the same thing but I have two sets of foresights to the same point. After pulling the data out of the DAT file (and paste into a blank notepad) so I can see the differences. I finally spotted it. The second set of foresights had a foresight to another point first, forgot to change the TH the second time. Fixed that and those flags disappeared.
I take paper notes just so I can figure out what I did 3 months ago.
What is traverse closure error?
I got to re-reading the textbooks about the open-loop closure test and relative precision ratio.
It seems to me the only fair way to rate your work using closure error is to calculate the open-loop traverse as measured and find how much it misses coming back to the first point.
But I find that the textbooks don't do that. Davis, Foote, and Kelly 1968 edition, Kissam 4th ed, and Wolf & Ghilani11th ed. all show examples where they do the angle adjustment to achieve 180(N-2) degrees, and THEN calculate the misclosure and precision ratio while tabulating the latitudes and departures that will be adjusted by compass rule. That's what I was plotting in the figures above.
The calculations using the raw angles, of course, would be pretty useless for anything else besides finding the closure error, so I guess I see why they did it back when calculation was tedious.
It seems like their method would tend to look better, but using the raw angles would be more representative. So if nothing distracts me tomorrow, the task is to add that calculation to my simulation and see how much the traditional way differs from what I think is the "right" way.
What does your software (or hand calculation) do if it finds a closure error or relative precision ratio?
What is traverse closure error?
So my simulations show that the average or rms of the raw closure isn't a lot different from that of the textbook closure. It is greater by a factor of about 1.2 or so. Even with the angle error increased to dominate 10:1 over distance error, the ratio of closures is still only 1.4 or so.
What is interesting is the scatter plot. This says that the textbook closure is only a moderately good indicator of the raw closure or vice versa. If one type of closure calculation looks good, the other may not.
So again, just because your closure looks good, you shouldn't bet on everything being good.
> I take paper notes just so I can figure out what I did 3 months ago.
I've been taking nothing BUT paper notes (and measurements), and I can't figure out what I did 3 months ago!:-S
can u do closed compass traversing with 12 side polygon in matlab their rise and ur fall, reduction level and the out put on external tabular format
can u do closed compass traversing with 12 side polygon in matlab their rise and ur fall, reduction level and the out put on external tabular format
Please use sentences, and a little more detail of what you are seeking.?ÿ
My first impression of your question tells me that yes, you could program such a computation, but I'm not at all certain I know what you want.
Of?ÿ course, scalar errors are not detected by traverse closures. Uncalibrated chain, EDM wrong barometric pressure/temperature correction, that sort of thing.?ÿ The polygon closes well but is larger or smaller than reality.?ÿ Now that I think of it, that would be true for a network type of survey too.
Say what you want, and I know all about the errors beyond our control, but when you have good calibrated equipment, and some can consistently close on a dime and others can't, it tells me something.
