Without the actual data to work with, this is speculation BUT -
The GPS data will have an error ellipse associated with each point and is not gospel. Given that the GPS data was not collected under ideal circumstances as mentioned in your post, the GPS data should not be expected to meet the manufacturers specs for ideal conditions. Ideally you would be looking at the possibility of a few hundredths in your static position.
IF your two begin points were off in differing directions and you do not meet the specs due to foliage, you can very easily see the 46" over a 4000' line.
Given the site and the traverse line, you must decide if you would be better off with
1. Holding the GPS coordinates (both pairs) and adjusting the conventional data between these points.
2. Least squares adjusting all values (provided you can confidently weight the various observations). I don't know what value to use in placing a weight on a GPS vector under less than ideal conditions.
If you hold the GPS values and all observations were under similar conditions, there is really no reason to think that the first pair are any more accurate than the second pair.
If applying the correction to only the first angle gives you a better result than applying the error equally throughout the traverse, I would suspect that you either have a bust in the first angle and should recheck it or more likely you have an initial point pair that has more error on one point than the other.
I suspect with ONLY 400-500' back sight distances and admitted fubar'd GNSS conditions that the error may very well be in either the starting or closing azimuths you are holding as truth.
Also, on a linear traverse with no redundancy, a LSQ adjustment and a compass rule will produce very similar results.
I vote for some sort of celestial body observation on BOTH GNSS pairs to confirm what you are holding for control, you can already check the distances between those points as they are obviously intervisible, might as well figure out if what you are fixing is correct before doing so.
I have noticed for years (especially when I did a lot of contract GPS control for others in the 1990's) that often folks simply are not running near as good of conventional work as they think they are, it really could be 46" of error in your traverse, BUT you won't know that until you check the GNSS azimuths with an independent method.
Edit, Jon said a lot of the same while I was writing my post!
SHG
There is a considerable amount of cover and the GPS data had to be worked to get the acceptable 95% confidence level
By itself 95% confidence level is meaningless. What is the Standard Error of the GPS points?
You need to run this project through a least squares adjustment using valid Standard Errors for all the measurements, including the GPS points. Only then will you know if you have any problems.
I'm going to attempt to be nice, but these kinds of threads drive me nuts. I'm envisioning your basic rectangle where the 4 corners were GPS'd in and tweaked to perfection. Then along comes Mr Total Station and there is a bust by somewhere between a tenth and a foot. Ok, it's screwed up. No different that not closing a 1/2 mile level loop by 0.06'. Re-do it.
As mentioned above, you cannot adjust a blunder. Something is screwed up and it's likely field procedures or equipment issues, and not something the office or any majic software can fix.
I suppose some of the answer could include what the purpose of the job is? Building a bridge, a parking lot, a boundary survey, a cell tower. What? Sometimes that foot don't mean diddly, sometimes it will break the bank.
Good luck & have a good time with it.
Chan, formerly known as Wayne Griffin
Another suggestion. You start with 2 GPS pts, and you end with 2 GPS pts.
Why not add 3 gps to this network, and isolate the error, and then rerun the total station traverse through these. I think you could do that by yourself, in a day, or less.
N
> Why not ..... rerun the total station traverse through these.
There is no bust here. Just normal measurement errors. Nothing to be gained by re-running. I'm convinced that an LS adjustment report on this data would show residuals within predictable norms.
IF you found it piled up, that is.
N
Let's Start With "GPS data had to be worked"
GPS data had to be worked to get the acceptable 95% confidence level. With cover problems to boot.
NGS says to expect 2cm precision with OPUS under good conditions with longer observations than 1 hour. So right off the bat you are expecting too much from your GPS observations. You give no indication if any of the GPS observations included more than 1 point at a time. Expect your precision to drop further.
I will be generous and accept that your GPS points had 2cm precision. That is an allowance of 0.066' per point. Squaring that, multiplying by 2 and taking the square root offers an expected 0.09' error in horizontal alignment of the 2 start points.
0.09'/500' = an angular error of 37" before you even start your traverse.
Then you traverse 4300' and say you doubled the angles, however what do you mean by duoble? Unless you shot 2 Direct and 2 Reversed you cannot meet your instrument specs and did not meet the quality for a 4300' traverse. I note you shot distances in both directions which is required. However you do not mention if you carried elevations. Not carrying elevations is a waste of GPS data, because it is a very good check on your overall precision. Lastly you do not mention the quality of your total station, 1", 3", 5"? Assuming you had a 3" gun, 2D2R, and allowing 0.005' for setup error, your angular precision is 4" per setup, or 15" for the traverse run.
You need to evaluate your instrument, your methods and your GPS precision use least squares and allow all your GPS positions to float.
Going from a pair of fixed NGS points to a pair of fixed NGS points is acceptable because of the long number of observations of those points over times. A GPS pair to a GPS pair is inadequate, GPS triples are required to get near the quality of long standing NGS pairs. You have mixed apples and grapefruit, the sizes are different.
Observing Sol, Polaris or distant towers can substitute for that third GPS point.
If you want to be graded on your project, I give you a "D".
Paul in PA
> If applying the correction to only the first angle gives you a better result than applying the error equally throughout the traverse, I would suspect that you either have a bust in the first angle and should recheck it or more likely you have an initial point pair that has more error on one point than the other.
If I apply, only the first angle, and not the rest of the angles in the conventional work; doesn't that also, change the value of the closing angle?
:-S
> One of the central tenants of adjusting data
Brief diversion: A tenant is one who occupies an estate. A tenet is "any opinion, principle, doctrine, dogma, etc., especially one held as true by members of a profession, group, or movement." Both words share a Latin root: tenere, to hold.
Carry on.
> I suspect with ONLY 400-500' back sight distances and admitted fubar'd GNSS conditions that the error may very well be in either the starting or closing azimuths you are holding as truth.
>
> Also, on a linear traverse with no redundancy, a LSQ adjustment and a compass rule will produce very similar results.
>
> I vote for some sort of celestial body observation on BOTH GNSS pairs to confirm what you are holding for control, you can already check the distances between those points as they are obviously intervisible, might as well figure out if what you are fixing is correct before doing so.
>
> I have noticed for years (especially when I did a lot of contract GPS control for others in the 1990's) that often folks simply are not running near as good of conventional work as they think they are, it really could be 46" of error in your traverse, BUT you won't know that until you check the GNSS azimuths with an independent method.
>
> Edit, Jon said a lot of the same while I was writing my post!
>
> SHG
Very true, this traverse needs redundancy and when doing conventional traverses one has to follow the procedures of the NGCC (the yellow book). I'm surprised more of you old timers don't know this or haven't mentioned it. Here it is, if you are doing conventional control networks: http://www.ngs.noaa.gov/FGCS/tech_pub/1984-stds-specs-geodetic-control-networks.htm
I have encountered a somewhat similar situation. the "error" is most likely in your conventional traverse, where the measured angles are close to 180° as those legs are the weakest. As stated by others I would apply a Least Squares Adjustment putting the least weight on those 180°± angles.
Broken link.
When you are measuring angle and distance, there is nothing special or bad about being close to 180 degrees. You may be thinking of shallow angle-angle intersections, which indeed are weak for determining distance.
In any case, for the LS analysis you use the same standard error in seconds for any angle and let the LS work out the effects of geometry.
Link worked for me
.
> If I apply, only the first angle, and not the rest of the angles in the conventional work; doesn't that also, change the value of the closing angle?
Yes. That is the point of what the original post was discussing.
From what I read in the original post, that was the suggested procedure for 'correcting' the error. They had 46" of angular error and 'corrected' that by making the first angular measurement something other than what was actually measured. Thereby, they rotated the rest of the traverse (with no real basis for the adjustments and held the rest of the measured angles fixed) until they had an 'acceptable' closure distance of 0.08'. Which means they now have a different angular error from the original 46". If that process provided them with the closest 'adjustment' of the data, then I would be very suspect that the first angle was a bust -
EXCEPT THAT:
The same amount of angular error can easily be in place due to the coordinate values reported from the GPS observations NOT being correct by a very small amount of error in opposite directions. With 2 GPS point pairs, the scenario presented has two lines reported on two azimuths. Those two azimuths are not necessarily (and usually are not period) running directly in the reported directions - especially if observed in less than ideal conditions. 23" over the 500 feet is around 0.04' (ahhh the elusive 4 hundredths). That amount of error in GPS points is highly likely, especially with the GPS conditions being described as less than ideal.
What I get from the first post is that someone in the office is holding the post processing software as absolutely correct (even though it took a good bit of tweaking to get the results). So they want to hold the GPS coordinates as fixed and are basically throwing all of the adjustment arbitrarily into the first angle of the traverse for the sake of convenience. When it very easily could be mostly error in GPS positioning.
Well Done
>
> I will be generous and accept that your GPS points had 2cm precision. That is an allowance of 0.066' per point. Squaring that, multiplying by 2 and taking the square root offers an expected 0.09' error in horizontal alignment of the 2 start points.
>
> 0.09'/500' = an angular error of 37" before you even start your traverse.
Great summation of the situation, Paul. And the math really speaks for itself on this one!
You can not hold the GPS points as gospel - especially when they had to be worked over and were observed under less than ideal conditions. Even the best case expectation starts with a fair amount of angular error when starting and closing on GPS pairs.
...
> There is no bust here. ....
Boy that sounds like famous last words!;-)
I think Jeff is right on this one.
N
didn't Kent do a good series on this kind of thing??
I was thinking the same things, Paul. Just crunching the se numbers before you even take a measurement would be helpful.
Isn't there a routine in Star Net that lets you put in your instruments se numbers and do some simulations ?? I know Kent did a long and thorough analysis on TOB (the other board) of Star*Net.
I always wonder just what kind of project plan these kinds of things are for. You spend hours and hours collecting and manipulating your GPS obs. then go out and run a traverse as well? why not just traverse the whole thing from the get go? two little 500 foot legs between 4500 feet of project just seems weak, geometrically speaking.