Getting started with Least Squares
I am looking to get started using least squares and was wondering about best practice for field procedures etc. Could somebody point me in the right direction for documentation, web links or even books? I would be hoping to use a combination of total station and gps data.
I will be using Survnet in Carlson Survey 2011 Intellicad version. Thanks.
- 50 Replies
Totalsurv, post: 339400, member: 8202 wrote: I am looking to get started using least squares and was wondering about best practice for field procedures etc. Could somebody point me in the right direction for documentation, web links or even books? I would be hoping to use a combination of total station and gps data.
I will be using Survnet in Carlson Survey 2011 Intellicad version. Thanks.
Carlson has two WEBINARS posted for SurvNET which are very helpful. I believe that they have several “lessons” posted also.
T.W.
Not entirely necessary but I would recommend keeping instrument and rod heights even if I am not really interested in elevations.
Shawn Billings, post: 339404, member: 6521 wrote: Not entirely necessary but I would recommend keeping instrument and rod heights even if I am not really interested in elevations.
I always carry elevations, I know the one day I don’t I will need them!!
Two points about LS:
Make sure you give the program realistic estimates of your standard errors for centering, angles, distances, and GPS positions, in order to achieve the best solution. Overweighting some measurements means it partially ignores the information in others.The biggest advantage of LS is the ease of combining redundant measurements, that are hard to deal with using compass rule or other methods. GPS and traverse info is an obvious combination. Don’t overlook the value of adding diagonal or cross-tie measurements to your traverse when they are easy to get. A few redundant cross-ties add a lot of strength to the figure (smaller error ellipses).
.
I agree with Bill about cross ties. Even to the point of using angle ties to distant features (building antennas, steeples or anything that is unique and can be seen from multiple stations. They help the most when the traverse has some bulk to it. And it will also give you another possible control point to use in the future or if you are having trouble finding other points. Practice has been the most effective learning tool that I have had with LS. I won’t ever go back to compass. The results are often the same but I work on enough complex networks that compass would wrack my traverses all over the place. You may also search through the old posts here a few times as its a popular topic.
—Dan MacIsaac, PLSIs it always necessary to have a closed traverse or traverse between 2 known points with LS?
If there are no cross ties available is the redundancy provided simply by backsight on each setup?
Totalsurv, post: 339411, member: 8202 wrote: I always carry elevations, I know the one day I don’t I will need them!!
You have to record heights if you are going to combine GPS vectors with your conventional measurements.
Least Squares allows you to be creative with your equipment.
A traditional closed traverse it not necessarily required but you do need redundancy. The nice part about LS is you can get your redundancy however it is most convenient and efficient to do so.
A simple example is a long route traverse. Traditionally you would have to close it by finding a more or less circular or rectangular route back. With LS you can do the traverse open then use GPS to make additional measurements so that you have redundancy.
Totalsurv, post: 339429, member: 8202 wrote: Is it always necessary to have a closed traverse or traverse between 2 known points with LS?
If there are no cross ties available is the redundancy provided simply by backsight on each setup?
A closed loop LS will be very close to a Compass Closure. You can run open traverses, point to point, line to line, add a bunch of RTK shots with specific errors associated (CSE).
If you start from a pair of points and traverse away you will find the error ellipses increase greatly, so traverse to something.
I would record a backsight on each setup, but that really just averages out the distances between the OP and BS plus the previously recorded OP and FS (being the same pair of points).
Someone posted a while ago about turning to something with every setup – like a church steeple or similar. Sounds like a good idea if you can see that something from everywhere.
The example at college was a 26 point closed loop. One crew setup inside a building to resect off a few points in the loop – to get a cross tie. We then adjusted the loop (similar to compass) then we added the extra data (a little different). That was using that hector the vector or whatever Hintz created for 2d LS Adjustment.
Another useful feature of many LS programs (Star*Net, for instance) is the ability to pre-plan the measurements you are going to need to get the required accuracy. You can put in approximate valuesfor your points and measurements and get the error ellipses. If you aren’t satisfied with those ellipses you can try adding measurements until it is good enough. This exercise teaches you how to pick the measurements that do the most good.
.
Totalsurv, post: 339400, member: 8202 wrote: I am looking to get started using least squares and was wondering about best practice for field procedures etc. Could somebody point me in the right direction for documentation, web links or even books? I would be hoping to use a combination of total station and gps data.
I will be using Survnet in Carlson Survey 2011 Intellicad version. Thanks.
It is my understanding that proper procedures should be understood prior to anyone even beginning surveying. I would not alter using any of my best practices just because I planned on using least squares or not using LS.
In general I do not plan on using least squares. If I fumble my work I would use least squares to find my error and then correct it.
There are occasional times where conditions dictate surveying from less than good control points or in less than good conditions. In that case least squares is appropriate, but not as part of an everyday, every project work flow.
The proper name is “Least Squares Adjustment”. If you have to adjust ALL of your field work, then you are doing it wrong. We learn least squares in school to emphasize the importance of doing it right first.
I can start a campfire with wood and a string or with flint and steel. Just because I can that, is no reason not to use a match.
Paul in PA
Paul in PA, post: 339444, member: 236 wrote: If you have to adjust ALL of your field work, then you are doing it wrong.
When you file a record of survey/plat, don’t you adjust it so that it mathematically closes?
There is always some (hopefully small) error, and you don’t know where the errors are, so the best estimate is obtained by adjusting, using some reasonable method. LS and compass rule are reasonable adjustment methods, and LS has the advantage that you are not constrained to use only the measurements of a closed traverse.
.Paul in PA, post: 339444, member: 236 wrote: It is my understanding that proper procedures should be understood prior to anyone even beginning surveying. I would not alter using any of my best practices just because I planned on using least squares or not using LS.
In general I do not plan on using least squares. If I fumble my work I would use least squares to find my error and then correct it.
There are occasional times where conditions dictate surveying from less than good control points or in less than good conditions. In that case least squares is appropriate, but not as part of an everyday, every project work flow.
The proper name is “Least Squares Adjustment”. If you have to adjust ALL of your field work, then you are doing it wrong. We learn least squares in school to emphasize the importance of doing it right first.
I can start a campfire with wood and a string or with flint and steel. Just because I can that, is no reason not to use a match.
Paul in PA
Do you perform any type of adjustment in your everyday project workflow?
Bill93, post: 339449, member: 87 wrote: When you file a record of survey/plat, don’t you adjust it so that it mathematically closes?
There is always some (hopefully small) error, and you don’t know where the errors are, so the best estimate is obtained by adjusting, using some reasonable method. LS and compass rule are reasonable adjustment methods, and LS has the advantage that you are not constrained to use only the measurements of a closed traverse.
No to first question.
Yes there is always some small error. But, a “small” error can stay wherever it is, no need to move it to some other position.
I do not limit myself to only closed traverse measurements, but I do not necessarily need LS to add other measurements.
Paul in PA
Paul in PA, post: 339452, member: 236 wrote: No to first question.
Yes there is always some small error. But, a “small” error can stay wherever it is, no need to move it to some other position.
I do not limit myself to only closed traverse measurements, but I do not necessarily need LS to add other measurements.
Paul in PA
Goes against everything I was thought at school. I was always told if you don’t adjust you just dump all the cumulative small errors on your last point as no measurement is ever perfect.
Totalsurv, post: 339450, member: 8202 wrote: Do you perform any type of adjustment in your everyday project workflow?
I thought I had made it clear, I do not adjust my everyday field work. If it falls within my expectation of erros, I accept it for what it is. Honest measurements with error.
Quote me on his, “Least Squares Adjustment does not make observations better, but it does make them different.
The measurements are not closer to the truth, mathematical perfection is not what surveying is about.”Least squares is much more useful in checking your product than it is in being your product.
What do I prove by adjusting my field angle by 1″ or a distance by 0.01′, when I use a 5″ gun and a bipod on a found pipe that is deformed or a rebar with a not so flat top?
Paul in PA
Totalsurv, post: 339454, member: 8202 wrote: Goes against everything I was thought at school. I was always told if you don’t adjust you just dump all the cumulative small errors on your last point as no measurement is ever perfect.
Since I do not regularly survey sections the amount of error to accumulate is small. Whatever it is, it does not get dumped on the last point, because in a proper closed traverse there is no last point.
Paul in PA
Paul in PA, post: 339444, member: 236 wrote: In general I do not plan on using least squares. If I fumble my work I would use least squares to find my error and then correct it.
Yes, blunder detection is one good reason to use least squares in the workflow, but even with no blunders, it gives you the best bang for the buck. It’s a force multiplier.
If you have to adjust ALL of your field work, then you are doing it wrong.
Well, all surveying observations are already adjusted for systematic errors on an individual basis: the total station has its internal circle calibration, the EDM has its internal offset and scale calibration, individual GNSS vectors are the result of least squares applied to a large pyramid of models and supporting infrastructure, etc. Surely these adjustments have merit when applied to observations on an individual basis. Why should it suddenly become undesirable to optimally process a network of such observations? It’s the same mindset applied to a larger system.
Cheers,
Peter
Well, I think it is incorrect to create a map that doesn’t mathematically close. For example, if I do a very simple survey of a rectangular parcel of land where all four corners are monumented, when I show the bearings and distances between those found monuments, they necessarily must define a mathematically closed figure. If they do not, then clearly there was an error in my measurements. Each monument only has one position. But if the courses that connect them into a figure don’t define a mathematically closed figure, then I’m essentially saying that one or more of these monuments have two (or more) positions, depending upon how it is drafted.
Most of the time, this is a very minor error. Imperfections in the equipment and our limitations as human observers introduce miniscule errors that cannot be avoided. The only resolution is to find a reasonable method to “adjust out” these errors. And that’s what things like the compass rule and least squares adjustment accomplish. They remove the small, unavoidable errors that creep in to my measurements.
Obviously, if I’m adjusting out a 5 degree angular error, then neither compass rule nor least squares adjustment is appropriate to use. That is a bust and needs to be reobserved in the field. But barring actual busts in observations, some sort of adjustment is necessary to reconcile the tiny errors that all survey measurements inherently have.
Paul in PA, post: 339444, member: 236 wrote: but not as part of an everyday, every project work flow.
In my operation, it is, with few exceptions, part of an everyday, every project workflow.
As Paul suggests, with decent equipment and procedure you could, in most cases, forego any adjustment and not get in trouble. But once you get the workflow set up, running a LS adjustment is so easy and the reporting statistics so useful in evaluating your results that it seems silly not to do it.
Log in to reply.