Activity Feed › Discussion Forums › Education & Training › Least Squares Adjustment Training
With the entry of affordable & easy to use GPS receivers & software, LS adjustments has & will become obsolete. The main reason why LS is used in survey computations is because of the propagation of errors due either to instrument or because survey measurements will always have a certain degree of error no matter how good you do your data collection.
We in the past used LS for long road alignment surveys that have a lot of closures using the intersecting roads as network paths. When we began using GPS receivers, we would then place them at every 1km intervals or at intersections if there are no obstructions. You then just adjust your traverse to the GPS coordinates as you reach each GPS point. If your traverse coordinates differ with the GPS coordinates more than what is allowable under the circumstances then you add another GPS point midway along the traverse line to catch where the error might have occurred. Maybe a rerun would be necessary just to make sure that you are not adjusting an erroneous traverse line to force it to meet the GPS position.
In a sense the GPS receiver has become the LS solution.
Ouch! Watch out….
Utter blasphemy to many on this site. There are likely even those who advocate total stations with cross ties and LS should always be used to verify GNSS solutions. And, if you don’t have a total station on the job site you definitely should be driven out of the profession.
I remember our LSA (least square adjustments) professor in college who said LSA is a form of legal data doctoring. It’s like all your loops & spur lines don’t agree with each other and you don’t know which is correct so you use a difficult math process to justify your errors and select a final coordinates for points with errors.
GPS post processing also uses LSA in its computations of point positions. There are so many vectors between base/rover observed that PP is really just selecting those vectors with the least standard deviations and eliminating those that fall outside of the standard deviation.
oh, how I wish Kent was here to see this one…..
I miss his participation overall, I know he had his detractors but he put forward quality content and engagement on the regular.
- Posted by: @party-chef
I miss his participation overall, I know he had his detractors but he put forward quality content and engagement on the regular.
I agree totally. Kent knew what he believed in and defended that belief as well as anyone I know. I have 2 sons who are licensed, but I would have been honored to have had someone with Kent’s knowledge, patience, care and skill to mentor them.
- Posted by: @jt50
With the entry of affordable & easy to use GPS receivers & software, LS adjustments has & will become obsolete.
Sitting in for Kent this evening will be Norman Oklahoma ……..
- First, affordable and easy to use GNSS receivers have been around for over 20 years and LS is gaining acceptance, not losing it.
- JT50 seems to be saying that GPS vectors are error free. Obviously that is not the case. And even if it were there are still several other elements of the measurement system that introduce errors. So…. it will always be valid to LS adjust GNSS measurements no matter how small the residuals may be. Because as residuals get smaller, expectations get larger. Once upon a time a 1 foot closure in a traverse around a section was considered cause for celebration. Nowadays it’s a bust to be rerun.
- What of situations where GNSS cannot be used? JT50 is perhaps not aware that in excess of 90% of all work in the PNW is still done with total stations. Many surveyors around here do not even own a GPS setup.
- And what of combining GNSS vectors and terrrestrial measurements? There is no better way than LS to get that done.
I have said that the main reason to run data through StarNet (or any other LS package) is to trap and correct busts. The adjustment part is secondary. But still a valid treatment. On the other hand, if JT thinks of LS as a way to paper over his busts he is sadly mistaken.
While likely a troll attempting to generate interest in himself, I will add to Norman Oklahoma??s comments.
JT50??s apocryphal ??least squares adjustment professor? is said to feel that the method of least squares to be ??data doctoring.? If so, he does not understand it. While developed by both Legendre and Gauss at the start of the nineteenth century, it was not widely adopted until computing power became widely available. It was never considered an inferior technique to simpler methods like that of Bowditch.
Tellingly the poster mentions that he ??adjusts? his observation data to GPS coordinates never explaining how. He asserts that he knows ??correct? data. I like a little more rigor.
That didn’t sound right to me either, a college professor supposedly denegrating a complicated mathematical computational procedure.
@norman-oklahoma
I use StarNet or TBC (and sometimes both just for comparison) for a couple of reasons:
1. It’s a great way to process and use all of my measurements in generating final coordinates.
2. It delivers information about the quality of my positional guesses, uh I mean estimates.
3. StarNet is especially good at flagging probable bad measurements.
I am about 4.5 miles in a 6 mile traverse around Sections 18 and 19, rough woodland. StarNet shines at this sort of thing, TBC not so much. StarNet says my current end station has a 4.5′ error ellipse because it’s just swinging out there, not closed yet. We do have one traverse station in a small opening which may yield an okay OPUS solution on an overnight session. The only other possibility is about a half mile north on the same line. We also need to traverse through the middle of 18 (east-west) so that will add some redundancy. Traversing through the woods you get what you get, the terrain breaks in inconvenient places or too quickly.
- Posted by: @geeoddmike
Tellingly the poster mentions that he ??adjusts? his observation data to GPS coordinates never explaining how. He asserts that he knows ??correct? data. I like a little more rigor.
First question is, do you know what a survey grade GPS receiver is for? If not then I no amount of explanation will get through to you.
If you understand what a survey grade GPS receiver is capable of then let me explain. When you do a linear traverse for a road survey or through a forest, you are going to have errors in both angular & linear measurements. If your line is say 10km long then you need to check if your carried azimuth or distances still meet the required precision. You get GPS readings at regular intervals along your alignment. Maybe 2km, 1km or 500m or if you want you can set 1 day to get GPS readings at every 100m if those points are under open sky. You then compare those coordinates with those from your traverse. If a segment is off beyond required precision then make a rerun. If within allowed precision then adjust using compass rule. If you doubt your GPS readings then do a double observation on different days or different hours or when your planning software tells you you have the most satellites available. But with today’s multi satellite system, you will more likely to get more satellites than what was available during the GPS only days.
That is one way, of course. It amounts to holding the GPS vectors fixed and assuming that all the error is in the traversing. Which obviously just isn’t the case. Do as you see fit, but that’s not the way I roll.
BTW, suggesting that Geeoddmike doesn’t know something about GPS or making adjustments is comically absurd. You don’t know who you are addressing.
@norman-oklahoma
1st order precision is 1/100,000 o you are allowed 1m for every 100,000m of measurements. every dual frequency gps receiver today has an accuracy of 0.007m + 1ppm for static mode. to demonstrate the accuracy of the gs receiver and simplify computation of precision, we will assume that you are able to observe end points of the 100km line. so if you follow manufacturer’s specs then it means for the 100,000m line you will get an error of 0.007m + 0.10 mor 0.107m. that equates to a precision of 1/930,000 or 9x 1st order.
If you doubt your GPS vectors then do redundant observations on different days & hours.
For situations that GPS is not possible? Tough luck then. Not really, you either live with bad traverse lines or make your route out of the canopy to high clear grounds.
@norman-oklahoma
you do understand that the accuracy of a gps receiver is way above 1st order?
but then you use what you have. if you don’t have access to GPS or not usable under current situations then that’s life I guess.
Oh, I think I do. Your use of the obsolete term “1st order”, and your belligerent tone, suggests to me that I’m having a battle of wits with an unarmed person. I think that you are a troll, and I refuse to feed you any more.
@norman-oklahoma
1st order is now obsolete? really?
I am not sure why everyone is upset about my opinion that LSA is obsolete. The purpose of LSA is to determine the most accurate position of a certain point from several observations. If you are able to determine a certain point’s position using the best method available then why not use it? GPS results offer that high accuracy output that is faster and more accurate than your traverse on any given day. Why say that LSA would give you better network precision when a network of GPS points would be that precision that you are after using LSA.
I am very much alive and I don’t believe in reincarnation. Not part of my religion anyway. Reviewed my past posts and you were the one that responded to my question on Magnet post processing. I now remember that you are the guy that did not understand why I want my vector pointing from base to rover. Such a basic concept in surveying about line direction in post processing and you could not get it and now you are here to teach me on LSA?
Two quick questions from a non-surveyor.
1. When you do a gnss measurement half-way between two points that are 1,000 meters apart as a check, aren’t you using a lower-order measurement to check a higher-order one? With a 0.007 meter error, 1000/0.007 = 1 : 142,000 while 500/0.007 = 1 : 71,000.
2. You can settle a disagreement between Kent and me. Is the 0.007 error absolute or is there variability in that number as well? Is there an absolutely guaranteed error of, say, 7 mm every time you set up the receiver or can that vary maybe from 0 mm to 50 mm? Or maybe some other range?
Log in to reply.