Another leveling question

> To me, it is incomprehensible that you would not adjust a level line, but then again I specialize in control surveys.

I've seen and heard arguments on both sides. Both have merit.

But just to play Devil's Advocate I would like to say that adjusting a level loop really doesn't improve the quality of the data. There is a chance some of those derived elevations are good as gold. All the error may rest on one shot. Adjusting the loop only smears a probable error throughout the entire loop.

30 miles and 0.18'...I could easily live with that. Some folks really need perfect closures. If that's the case, adjust.

In 30 miles I would suspect that orthometric (Helmert) corrections would be non-trivial (at least they certainly would be in my neck of the woods).

And...YEAH! (there should have been a formal adjustment of some kind)

Loyal

I have always adjusted level loops. Can't really tell you why or give you a good reason other than 40+ years ago the boss told me to do it. And I've kept on doing it.

Loyal-not nearly as high up as you are. I don't actually know what direction this project is (I haven't see any of it, just helping out a client), so I doubt the correction would be worth computing.

As for adjustment...even a simple mean of the F-B of each section would be easy to do. Since they ran each section F-B sequentially, then moved on to the next, it would have been easy in the field to put the mean elevation of the loop in the book as the starting point for the next loop.

>adjusting a level loop really doesn't improve the quality of the data

If you do forward and backward and get different answers, which one is better? You know they aren't both dead on, and you could be picking the worse one.

The way to minimize your error on average is to take the mean of the two (adjustment is simple in this case). It's probably closer to the truth than picking the wrong one of the two.

And you don't know if the orthometric correction is negligible until you compute it. Even on flat territory with no sudden gravity changes, I'd think it could be significant if the 30 miles had a large north-south component.

> All the error may rest on one shot. Adjusting the loop only smears a probable error throughout the entire loop.

This is where equipment and procedures come into play. If you use the appropriate equipment and follow the prescribed procedures, you effectively eliminate unmodeled errors and an adjustment becomes a valid method of removing the accumulated error.

Saying that "there's a chance" that all the error is one shot is a poor excuse not to adjust. Even if the crew used a First Order level and First Order rods but observed Third Order procedures, I'd still adjust, I'd just call the results Third Order.

> My client did GPS and he saw discrepancies of a tenth or two accumulating as he progressed through the project.

What geoid model were they using? any?

I'd be closing and adjusting each loop and trusting the level more than the GPS myself...

> I'd be ... trusting the level more than the GPS myself...
We all would. But at 0.1' to 0.2' you can detect trends in the GPS elevation.

Just looking at it from a construction surveyor's point of view, it's not going to matter if the control will also be used for whatever improvements come out of it. I' m lucky to get a set of plans where the elevations of the control points in a mile check within .18'.

Adjustment theory....

It is obvious that a closing error should be distributed through the observations such that it should produce a distortion as minimally intrusive as possible.

However, balancing out a closing error by no means improves the observations, it simply makes for a mathematically correct result.

The mark set represents the true value, which in fact, is unknown.

Observing represents an attempt to obtain the true values, which of course, is never actually achievable. It may approach truth, but error in all its forms distorts the results.

Adjustment produces a new set of values, distinct from the true and observed values, which may be mathematically correct and yet represents neither the true nor the observed values.

I've had few Johnnie Greens.
Shame they don't make it anymore....
:-/

Adjustment theory....

> It is obvious that a closing error should be distributed through the observations such that it should produce a distortion as minimally intrusive as possible.
>
> However, balancing out a closing error by no means improves the observations, it simply makes for a mathematically correct result.

That wouldn't really be correct, though, in the sense that the adjustment of a level loop loop with sixty turning points in common on the forward and back runs is simply deriving estimates of the height differences between successive pairs of turning points from two mostly independent measurements rather than one.

The mean of two measurements should always give an improved estimate if both contain only random errors of the same distribution, i.e. same standard error.

That is something we discussed. They set 60 TBM's, and I don't know how many targets for the mapping. So the mapping will match the TBM's left behind. And, since there are no NSRS BM's anywhere near the far end, it is unlikely this will ever cause a problem. But I still disagree with them on their "no adjust" philosophy. To me that is a fundamental concept when you have more than one observation of an unknown. If you only have two (forward and back), and they are the same length (which they would be), then a simple mean would do the trick

I am told they will totally rip up this road and reconstruct it. Apparently they wanted the mobile lidar because there are no existing plans at all for this road, no r-o-w, no as built, nothing.

I was just rereading this and was wondering what it would look like if you plotted these up with the "X" axis being distance and the "Y" axis being height. Plot all the forward heights in one color and all the back heights in another color.

If you end up with 2 more or less parallel lines would this be an indication of systematic errors?

And if the lines started crossing (or at least converging and diverging) could this be an indication of blunders?

Nothing really scientific or rigorous here, but simply a graphic illustration of the data.

Dave: I have not actually seen any of the data, at this point I am just "advising".

But, the surveyor at my client who has looked at the data says that all of the loop closures are positive, meaning there indeed may be a systematic error. I have not seen a map either, so i don't know how much the route goes up and down. I do know it is in a very rural county. He thinks they were using an invar rod (he said he saw them and it was yellow), but I find that hard to believe, as not many companies have invar rods and this particular company is not known for control work. Our invar rods are yellow, but I would not say for sure that all invar rods are yellow and non invar are white (although that has been my experience, at least with bar coded rods). In the late 80's I did a 130 mile bluebook run with a standard (i.e. non digital) level with micrometer, and those invar rods were white. I believe the yellow color is a more recent bar code phenomenon.

He is the one who did the GPS. And, like someone said, you cannot rely on a single GPS to prove or disprove a level line, but he sees a systematic creep. Of course, that could be a bias in the geoid model, but the geoid is typically pretty accurate in PA.

The first digital level I used was a Leica, in the early 90's. We had a wooden rod, I don't remember the length of it. But, the joints would wear, and as time went on we would get systematic errors going up long hills (which we have many of in PA), caused by reading high rod/low rod alternating. The error was only about 1 or 2 thousandths of a foot per shot, but it added up. Of course, when you level back down the error would cancel, so all would appear to be OK. We would see 0.2 to 0.25 error at the top of a big hill versus GPS. Buying a new rod eliminated that problem. After that, we bought a new rod every year or two, and we would check the rod periodically.

The next level I bought was a Zeiss (now Trimble). I have a 4 m folding rod (wooden), a 3 m invar rod, 0.5 m invar strip, and a 1 m rod (fiberglass), as well as two 2 m invar rods on loan from a client. The 4 m wooden rod (4 pieces) is the original rod I bought with the Zeiss Dini, which I think I purchased around 2003 or so, maybe earlier. The level was destroyed in an accident, I then bought a new Trimble Dini. But, that original 4 m rod is still quite accurate, it always compares well with the invar.

All the bar coded rods are made by the same manufacturer (or so I have heard), but each design is supplied by the instrument manufacturer. So, apparently the Zeiss design was much better than the Leica. I would think that by now Leica would have improved their design, that one I used was first generation digital level equipment.

With invar staves you would expect a systematic creep unless two staves are used and observations carried out in the correct manner for precise levelling - ie. the even set-ups read the forward staff first and the back staff last, odd set-ups read back staff first, forward last.

This should more or less eliminate any instrument settlement between the two readings (only likely to be significant when taking precise readings when it is quite possible for the instrument to settle into the surface by an amount of the same order as the reading precision). A road survey is a very likely candidate for this source of error as the macadam will give marginally under the points of the legs.

Where the readings are always taken as back sight first then the instrument settlement will tend to give a forward reading which is too small and over many set-ups the loop misclosure is nearly always positive as a result. Reading alternately cancels out this source of error. Easy way not to get confused in the field is to label staves A and B and always read stave A first.

For normal lower order accuracy levelling this problem doesn't really arise as the normal reading errors are greater than the instrument settlement.

Chris: my digital level does not have an option for BF..FB on alternating setups. However, it does have an option for BFFB on the same setup, which should accomplish the same thing AND give a second set of readings at the same time. Somewhat akin to running each section forward and back although the advantage of the forward and back method is different environmental conditions (i.e. AM/PM).

Although we have staves which are dual graduated (so you can read two different figures on each pointing) we would always run out and back. For most monitoring jobs we would run the loops both clockwise and anticlockwise, so as to create a time gap between the two sets of readings.

On one site we monitor, which is a large shopping complex built on an old power station ash lagoon area, the ground moves around a little during the day, so the monitor runs are carefully timed to use the same part of the day each time in an effort to minimise external influences and keep the observation conditions the same for each visit. The whole job gets spread over 3 days and involves about 20km. of precise levelling. Even with ground movements we normally manage to tie it down to about 1mm.

As an indication of site "stability", in the early days (we have been monitoring this site now for over 25 years) one of the big superstores shifted the internal layout of the groceries around. Moving cans of beans/peas etc. to where the toilet rolls used to be tilted the whole floor by nearly 5mm.

I've seen similar things happen elsewhere - an oil refinery where they demolished half a dozen tanks and replaced them at the other end of the refinery. Over 12 months the whole area tilted by about 2 inches as the underlying soft stuff migrated along the top of a hard rock band.

We just last week completed our third annual subsidence survey of an area above the land portion of a deep salt mine. The mine access is on land, but the mine itself is 2000 feet (600 meters) deep below one of the US Great Lakes. One particular area is subsiding at a rate of about 5 to 10 mm per year. The existing marks (8) in the area were deep rod marks set by NGS, but the salt company believed they were not deep enough, and that it was surface subsidence. They installed ten new marks in 2012, and drilled them in to what they say is bedrock, about 30 m deep. Our second and third leveling show that these marks move at about the same rate as the NGS marks nearby.

So that rate may not seem like much, and the area subsiding is mostly a park (and the mine offices) but there is a long running lake level gauge in the area. A CORS was recently installed at the gage to try and monitor any movement there.

The stable reference marks are a few miles away. Of course the entire area is undergoing post glacial rebound, but we cannot detect that in a small area without long term GPS.

Most of what we monitor are long linear structures (dams), so we don't have a lot of flexibility as to how we run it.

A look at our loop closures (mostly F-B, but one loop around three stations show about even + and - distribution. Average one way distance is 0.5 km per section.

One good thing is that is was cold, about 1° to 2° C, so the pavement was quite hard.

-0.0002
0.0001
0.0003
0.0003
0.0000
-0.0002
-0.0011
0.0003
-0.0003
-0.0001
-0.0005
0.0000
0.0001
0.0000

I've had similar experiences with deep drilled marks, going down 30 metres to bedrock. These consisted of extended reinforcement rods, inside a steel tube sleeve and spaced off that casing by plastic rings. In theory any movement of the sleeve shouldn't affect the rod.

In practice over a period of time all sorts of silt and other material manages to get into the sleeve (mainly from contractors working nearby not taking care to protect the cover) and at that stage you can never be sure that the surrounding ground is not influencing the datum point on the rod end. You can also not be sure that the rod really is anchored into the bedrock.

On the shopping centre I mentioned before we have 6 of these still in existence and they all move slightly, but a lot less than the surrounding ground. There is a pattern to the movement which suggests that the rock layer on which they found is moving slightly (rocking with a little lift, typically changes of 0.3 to 1.0mm a year, but with occasional jerks).

The reason for the extended monitoring of this centre is that in the early days there was substantial movement, (up to 80mm, including roof glass and shop front windows falling out, not to mention a water feature which broke and emptied itself into the shopping mall - fortunately overnight) which was eventually dealt with by installing a permanent network of grouting pipes some 8 metres underground. These have been used twice, each time pumping something like 600,000 litres of grout. Yhe place is now much more stable, but not perfect.