California Surveyors Forums

Far as I can tell, most GPS post-processing software is a least squares adjustment. So it may be necessary there just to get you into whatever coordinate system you are using for a particular project.

Otherwise, I don't think it is necessary. Now on the 2005 ALTA standards the Relative Positional Accuracy requirement can be easily computed within a least squares adjustment, so it is helpful when you need to certify to that.

I believe it is best used when you have a lot of redundancy in your measurements, like in a traverse with multiple angles distances being measured at each setup to backsights, foresights and other control points as cross ties. Picture a square shaped traverse with setups at each of the four corners and also shots to the diagonal corners from each setup. That would be a great time to use least squares.

A program like Star*Net, for example, which is the one I am familiar with, will allow you to adjust level runs through your control along with your horzontal traverse and measurement sets and your GPS vectors (baselines) exported from your GPS baseline software all in one adjustment.

You can use LS to adjust a traverse that doesn't close upon itself, but has known coordinates at each end. In that case you would want measurements to as many of the control points you can see from each setup to add to the redundancy, or take angle sets to the top of a local landmark that can be seen from multiple control points, like a tall church pire or something.

Necessary is difficult to define, since methods of adjustment aren't dictated in the law or standards of practice. It may be a best practice at times.

I do it whenever a traverse lends itself to adjusting - i.e. I closed the traverse, or did a good amount of redundant ties and I want to have the most probable value for my control.

Whenever, you would like to locate and distribute the error in survey measurements where the error is most likely to have occurred. Prior to the widespread use of LS various methods were employed. For example an angular error of 6" would be evenly distibuted among the angles of the closed figure, say 6 angles formed the figure, then each angle would have a second added to it. So LS give use the power to predicted which observsations have the likelyhood of having the greatest error and thus on angle may get 3" and the rest of the error is spread over the remaining angles.

One good thing about Star*net for instance is that one can employ this LS program to provide evidence that the positional tolereances are met as required for an ALTA/ASCM LAND TITLE SURVEY.

As far as I am concerned LS is necessary for control surveys and surveys that require a high degree of precision. LS is also a very good tool, to provide evidence of the precision of you survey.... error trapping...... many ways to do this of course...

A least-squares adjustment is part of my normal workflow. Even on small projects when there's not a lot to adjust, it provides both a reality check and a record of the field work quality. And it helps me sleep at night!

.

Answer: only after the errors have been found and corrected.

me: when is it necessary to perform an adjustment?
Bob Curtis: when you want to spread all errors through the whole traverse.
1982+/-
Bob said that when adjustments were arduous, but it is still true!

Find the errors first, always!

There are many great tools now to help find those errors, and I too much prefer star*net. It does easily give you the option of spreading the errors in likely locations, but it is still simply another error spreading method.

PS: there are a number of free LS software packages out there. do you want some recommendations?

"it is still simply another error spreading method"

A proper LSA distributes the errors in their most likely locations based upon knowledge of the error sources and their respective weights. This distinguishes LSA from other adjustment rubrics that rely upon unquantified assumptions about a very limited number of error sources and actually do "spread the errors" based upon that very restricted perspective.

There are adjustment rules, and then there's least-squares. They're related, but only distantly.

.

We also pretty much use LS for every job. Although it may not be necessary for every job.

Pete, please post the links, thanks....

"I am trying to figure out what i can do when we have a 5 mile stretch of topo that is linear...We have a gps smart rover but the program that we have does not have a least squares, just a grid to ground conversion."

Surveying without good LSA software is like fighting with one hand tied behind your back. However, if I were to approach a long linear job with only the tools you describe, I would use the GPS rover (I'm assuming you have access to a network RTK subscription) to establish control points every half mile or so along the route. I'd hit them at least twice -- more is better -- on different days and with at least a 3-hour time separation between observations. I'd mean (ugh!) the horizontal results and run levels over them. Then I'd run the topo with a total station. If it's only topo -- no boundary control -- I probably wouldn't adjust anything as long as the check-in at each RTK point is within acceptable limits for topo.

If delivering the job in CCS83 values wasn't acceptable due to the magnitude of the combined scale/elevation distortion, I'd have to consider how best to manage the distortion. The scale factor alone would only account for about 10ppm max across the project, but the elevation change could contribute much more than that.

.

I'll bump Dane's response. IMO, the ability to quickly identify blunders is one LS's greatest properties. Also, when mixing measurements of differing quality it really shines above the more primitive adjustment routines.

If you do use least-squares, it behooves you to be realistic when estimating the various errors in setup and measurement... so long as you intend to rely on the resultant error ellipses of your points.

Maybe I'm a nerd, but I generally take any chance I get to produce/review statistics.

A good working knowledge of staistics and what you LS software provides as analytical results is critial to understanding the qulality of your work and isolating errors.

After one of Kari Launan's (sp) LS Saturday Workshops. I was able to isolate the erors in a traverse to bad set ups caused by a bumb tribrach. In my opinion, Jim's analogy was close; I think it's like having BOTH hands tied.

A good working knowledge of staistics and what you LS software provides as analytical results is critial to understanding the qulality of your work and isolating errors.

After one of Kari Launan's (sp) LS Saturday Workshops. I was able to isolate the erors in a traverse to bad set ups caused by a bumb tribrach. In my opinion, Jim's analogy was close; I think it's like having BOTH hands tied.

interesting debate on the reliance of a man made mathmatic manipulation tool.

I assume that Least squares is similar to running out a traverse in both directions . . . . and "spotting" the location where a large discrepency is then assuming that spot is where the error most likely occured ?

The situation of a "Bumb" tribrach . . .is well . . .a bummer. Something like that cannot be truely corrected due to the fact that there are infinate ways the tribrach could've been rotated.

Does GPS least squares ajdustment, account for DOP at each reading ?

I'd be interested to see the actual adjustment numbers of such an adjustment. example three hundreths of a foot ??? or tenths ? \

Can prudent redundancy of field measurements, proper instrument care and good survey technique offer a valuable alternative to puzzling number crunching ? . . . . . or do we still believe in the magic of that Yellow box. ??

just asking , , ,

"Good" (with his Curta)

Good, if it is good enough for the french mathematician, who invented it in the late 1700's.... it is good enough for me.... At this stage of the game it is not really a black box/yellow box voodoo, it is fairly well understood and commonly used by surveyors with available computing power..... it can be done long hand.....

We need to use good practices whether or not we employ least squares. Using good practice does not eliminate the need for a robust adjustment package when the need arises..... just another tool....

goodgps wrote:interesting debate on the reliance of a man made mathmatic manipulation tool.

I assume that Least squares is similar to running out a traverse in both directions . . . . and "spotting" the location where a large discrepency is then assuming that spot is where the error most likely occured ?

The situation of a "Bumb" tribrach . . .is well . . .a bummer. Something like that cannot be truely corrected due to the fact that there are infinate ways the tribrach could've been rotated.

Does GPS least squares ajdustment, account for DOP at each reading ?

I'd be interested to see the actual adjustment numbers of such an adjustment. example three hundreths of a foot ??? or tenths ? \

Can prudent redundancy of field measurements, proper instrument care and good survey technique offer a valuable alternative to puzzling number crunching ? . . . . . or do we still believe in the magic of that Yellow box. ??



Edit: Dane said "it can be done longhand"

Yeah, I know it. Had to do it a couple times in school (well, with a calculator, not with tables and interpolation). What a PITA! Fortunately, we have things like StarNet, which is easy to use and is reliable. I might be able to do an LS adjustment with my calculator again if I had to, but it would involve much reading, writing, erasing, and cussing.

just asking , , ,

"Good" (with his Curta)

Valuable alternative? No.

Good field methods are not a substitute for adjustment, they are a prerequisite.

No matter what adjustment method you use, you need to start with good field data or it's just a waste of time and effort.

Compass rule is a non-statistical method that distributes error evenly throughout your control with the assumption that all of your individual measurements are just as good as any other.

Least sqares is a statistical method that actually takes the quality of each individual measurement into account to find (and you are correct here, Good) the most likely location of the greatest amounts of error, and distributes the corrections accordingly.

Thus, if you have a repeated angle with a spread of 10" between observations, it's going to get more correction than another that had a spread of 3".

Hopefully, you've checked your tribrachs before running your traverse and can then have confidence that your centering error is consistently within .005' or so. LS will take that uncertainty into account.

If you ran your control and then found that you had a tribrach that was wigged out, then it wouldn't matter which adjustment method you use, or none at all, you gotta redo the work anyway.

'Thus, if you have a repeated angle with a spread of 10" between observations, it's going to get more correction than another that had a spread of 3".'

I think Evan probably grabbed an off-the-cuff example to illustrate a concept, but I'm not sure he meant exactly what he wrote.

If two angles are measured with the same instrument under the same conditions (i.e., using the same standard error), then each of those angles will be weighted equally. Which one ends up with the highest residual (biggest adjustment) in a least-squares adjustment will depend upon which one most closely fits the calculated angle between the adjusted station positions. The measured angle furthest from the adjusted value will get the biggest adjustment regardless of the spread between observations. In Evan's example, it might be the one with a 10" spread, or it might be the one with the 3" spread. We don't have enough information to know.

If you limit the observations to traditional traverse angles and distances only, with no other redundant observations (e.g. cross ties, azimuth ties, GPS positions, leveling), then what Evan wrote is probably correct on its face. But the great thing about LS adjustment software is the ease with which it allows you to incorporate other measurements to tighten up a control network, and thus my comments.

.

I seem to recall evaluating the quality of each individual measurement and weighting accordingly. But it's been 15 years since I've done this by hand, so my memory may be a little fuzzy.

As I recall, with StarNet, you can set your standard errors for each instrument you use, which the program then uses for weighting as long as that instrument is specified. But I also seem to recall that you could override the standard weighting for any particular measurement.

"As I recall, with StarNet, you can set your standard errors for each instrument you use, which the program then uses for weighting as long as that instrument is specified. But I also seem to recall that you could override the standard weighting for any particular measurement."

You can, simply by adding explicit error values on the observation data line.

.

County Dude:

I thought a lot about this question a couple of years ago. I believe it is possible to use least squares or other similar methods to adjust long and linear traverses.

Here is one method I played around with:
1) Select appropriate backsights along your route that are visible from several set-ups. Water towers and the corners of tall buildings are examples.
2) Turn angles to the backsights from as many set-ups as possible.
3) Use the law-of-sines to calculate coordinates on the backsights based on your angle readings.
4) Compute a best fit coordinate for each backsight.
5) Use the best fit coordinate at each backsight to adjust the measurements in the traverse. (Compare the best fit backsight coordinate to the coordinate of each individual traverse measurement and use this comparison to make the adjustments.)

You'd need a total station with good angular accuracy for this method to work.

Of course, it is probably a lot simpler to set some control points along your route with static GPS.

I think some of these conventional surveying methods are becoming a "lost art" with the arrival of GPS.

It would be fun to write some code for this type of adjustment. I do think surveyors need to understand more about the concepts that programmers use to write adjustment software.

Contact me if you want to talk more about the method I described.

The Sunburned Surveyor (PLS 8489)

Somebody wrote that he hadnt run out an instrument traverse in a long time . . I think that is the case with most of us. GPS makes the work easier and therefore faster.

I agree With Hanrion.. . . "pointing the rod North" We attempt to orient the rover pole towards the north for shots also. This, only because it was strongly suggested to us in 1994.

"Consistency of methodology" . . .

Since errors tend to be random and therefore "inconsistant", One thing we can control is our methodology. "QUICKLY" check, recheck and don't ever trust anything . . . .(at first glance)

A great example is recently, my partner and I found a huge iron monument . . all painted orange and looking pretty . . .trouble was, it that it lined up perfectly with another Orange painted "pipe" . . . yet the whole thing just didnt seem to add up.

Miss "Prudence" enticed us to investigate further and we delightfully located a couple of very old monuments to the south.

Using our "Square head" common scense, we intersected the line from the south, with a record distance from the west . . . " If our calculations were correct, we could dig a hole and find wood" . . . . We did.
Carefully locating the center of the wood remains, (wood post called for in old notes) . . .we arrived at a position that was near perfect in alignment with other monuments and ties.

What adjustment should we make for this traverse ? . . . . Had we not found the wooden stake, would Least squares be able to make the adjustment in the office and point us to the proper location ? . . . .

The point I make Ladies and Gentlemen , , ,is that we can adjust traverses till the cows return, but if we are NOT on point the adjustments are worthless . . . . .

In California, land is expensive . . . getting the field mark correct should take precedence over the precision of computor "adjusted" measurements.
Citizens USE the field marker . . . .

Did a Frenchman invent the Shovel ? . . . . Whether or not . . .its "Good" enough for me . . . LOL

"Bon"

Well said "Monsieur Good". Great to see your regular posts again. Hope all is well in your neck of the woods!

No one makes the argument for bad proceedure or making precise measurements on meaningless evidence. Others have well stated the whys and wherefores, so I won't restate them here.But, least squares is a tool like any other tool that a surveyor can employ to get the job done.

DMI

YAY !!! the recognition that everything we use is a tool . . .like shovel, a GPS and the least squares pkg.

any tool is only as good as the user. GOOD JOB DMI !!!

One of my recent discouragements, was a conversation with LS8xxx.
This young surveyor shared with me, his rigorous adjustment of a traverse he'd run. "down to the 14th decimal place" . . . .
From his mighty calculations, he had proudly set his monument and tag at a calculated position.

I temporarily ignored his "findings" ( which weren't too great) and went about digging holes based upon occupation evidence. Lo and behold, we found two old redwood posts. measuring northward, and in cooperation with a third survey team who just happened to be there, we found the original concrete monument . . .a whopping 18 feet south of MR least squares.

If I bust a tenth on my map measurement due to poor PDOP, or a slightly off angle. . . . is that position still better that Mr Least squares ? . .

None the less . . .I must add . . . .I do use the least squares adjustment for nearly every job we do . . . . but first the "Most" Square head tries to reason things out.

Happy Mothers day to all . . .

What you describe has nothing to do with least squares. For the sake of argument, I accept the situation you desribe and in my view it has nothing to do with being a SQUARE HEAD or figurn it out in an old fashion type of way....or insisting on using newfangled yellow box voodoo magic to get a solution....THE CORE ISSUE IS THE FUNDEMENTAL UNDERSTANDING OF THE ROLE OF A BOUNDARY SURVEYOR. As far as I know the role of the Boundary surveyor is one of collecting and analyzing EVIDENCE. A part of this process is THE EXAMINATION AND CREATION OF MEASUREMENT EVIDENCE.

MEASURE EVIDENCE IS AN IMPORTANT PART OF THE PROCESS, BUT IT IS ONLY A PART OF THE PROCESS. This is something that you know well Mr Good, but it may be lost on your newly licensed cohort....

As far as I know there is nothing so important as a shovel and pipe locator...

Good,

In the case of LS 8xxx utilizing LSA to compute his positions, I say "so what?"

That job was screwed before any calculations were performed. As I stated in my 1st post, garbage in = garbage out, regardless of adjustment method used, or if none was used at all.

Adjustment by any method is simply one more step in a process of carefully executed work all coming together to report the best results of a survey.

So what does LSA have to do with your example? I'm not sure. Maybe LS 8xxx also used a total station and electronic data collector. Hmm, maybe we shouldn't use those anymore as too many techno-oriented know-nothings rely on them too heavily for those coords to the 4th decimal place. Maybe he used GPS. OK, let's get rid of that too. He dug the hole in the wrong spot, get rid of the shovels!

Oh wait, we like shovels. Uh, what kind of shovel was he using? Probably one of those fancy newer style PHDs instead of a round-nose shovel. Darn kids! See what happens when they try to go high-tech.


My point is that every step should be done with care. With poor measurements, the adjustment is meaningless. With bad setups, the measurements are meaningless. With the wrong points identified, the setups are meaningless. The adjustment is just one more step in the proper handling of data.

For what it's worth, of those I've run across who are likely to mishandle data at any step, most are likely to not adjust their data by any method. Their lack of understanding of proper field methods carries right through to a lack of knowledge of what to do with that data in the office.

A particular surveyor's lack of knowledge does not translate to the value of any particular tool as being useful to knowledgeable surveyors.