I use locus' for survey control on state plane grid (US feet). The raw unajusted coords are usually within a few feet of the location, but the elevations are usually about 100 feet higher than actual. My survey elevations are around 500-600 feet NGVD 88. Does everyone else experience the same difference, or does it tend to differ in other places?
thanks,
terry strickland

Yes................I see the same sort of differences in the raw positions...at the same elevations you mentioned...but that is normal, from what I understand...and is why we should tie to established control, if you want anything beyond 'just' relative postions.

Tom Bryant PLS Snowstorm, MO

Agreed. The "Nav" position, I have seen will be way off from the particular datum you really desire.
Something else I have noticed, and I usually process and adjust with SPC, orthometric heights and NAVD 88 using published NGS control and Geoid 99, is that the "free adjustment" of the really big network I have tends to miss the vertical points by very small amounts, as in less than a tenth of a foot. There are probably about 8 or so vertical points in the network. SoI usually constrain the adjustment with them all and go with it.
Now, I am wondering, how many of you guys would just go with 1 vertical point and the "free adjustment" using Geoid 99 and call it good?

I would hold the vertical one at a time, and find out which one is causing the network to blow up by the tenth, then unfix that one.

Dave,

I wouldn't constrain only one vertical control. The entire network could be "tilted", so the values were only coincident at that point.

Regards,
Brian

Brian and Mike,

I'm guessing I didn't make myself clear on my previous post. What I meant to say was that having good NGS vertical control in the network, and constraining to just one vertical point, when I check my "control ties" with the ohers they are all within a tenth of a foot or so of the "correct, published" values.
Now this network is BIG. Probably about 50 miles or so if you went around the perimeter. In the end, I do constrain the network with all the vertical points, Chi-square passes, etc. etc. things look good.
My question was along the lines of "if I have all of these vertical points in my network, and I only miss them by less than one tenth of a foot (when holding only one point fixed), then why with using the geoid model should I suspect "bad" elevations on a small, say three vector triangle, where one of the sites has known, good elevation"? Does that make sense?
As for having a "bad" elevation that doesn't fit the rest of the control points, and fixing that one along with another, yes, it does "tilt the plane" and checking the control ties verifies this. I have intentionally induced an erroneous elevation and seen this happen.
But back to the "3 vector triangle" scenario. Given you don't have a blunder in the antenna height, and you do have redundency that verifies this, you observe the dependent and independent baselines, the adjustment looks great, loop closure is great, why should I suspect the elevations of the new points?
Modified By Dave Huff on 1/7/2002 at 6:21 PM

I've experimented in several different ways with a network much as you have...Ours is a little larger (two counties), but relatively the same....
I worked for many an hour trying to decide how best to approach the 'best' vertical resolution possible. We have about 60 vertical monuments in our network..
I found that fixing one in the center and comparing results on the outer boundarys led to very small differences and these tended to be on both side (under/over) the published elevation. Mostly at the tenth level (except for one NGS point which is over a half foot off, but we think that it has been disturbed from other evidence)...

I debated stopping at this point and publishing the results...Rethinking the whole mish-mash, I further started wondering about the relative accuracy of Geoid 99 and how good the resolution was in our area...Reading the spec's led me to believe that using a geoid only model (on a project of this size) would lead to a very good coorelation of the vertical's over the entire project, but probably distort local areas...Since our main goal is to provide a very good network for Survey/Engineering companies, I then decided to 'warp' the network to compensate for the local variances..I think that this, while actually adding error, was the best possible solution to fit all the legacy data from 100 years plus of surveying data..

We store our network data in Lat/Long/Height, so as better resolution geoid's become available, it should be simple to re-do the vertical if and when necessary..

But your main question centered around a good fit triangle...You should always suspect the geoid (at this point), as the separation does not always coincide with the true orthometric height. You have some error in your ellipsoid calculation, then some in the geoid..When you apply these together, you dont' always get the true orthometric height...(and yes, I'm using orthometric and ground as the same)..

TM Vertical...The most perplexing part of GPS in my mind. The NGS has solved the horizontal, now time to tackle the vertical..

Brian...I think the tilt of the vertical really is not all that much anymore with with the resolution of the geoid being what it is...In the old days, we depended upon passing the residuals back into the software to model the correct geoid modulation. I don't find that it changes near as much with Geoid 99...Is this another legacy issue....

Clear as mud...as usual...but today we spent the entire glorious 60 degree day in the field....Ahhh.....

Dave,
Given your clarifications it wouldn't make much difference.

TM,
The tilt I was speaking of could be a geometry-induced bias (admittedly less likely than it used to be), so I still feel it's a valid concern, and worth at least having a couple of other controls for a check. This can also happen for the reasons you mention in high-relief areas (Rockies, not OK). ;>}

Regards,
Brian