OPUS Projects
User Instructions
and Technical Guide
NOAA | National Geodetic Survey
P a g e
|
87
Final coordinates from steps 1-4 above
Your final horizontal coordinates will come from the fully constrained geometric (horizontal)
adjustment (Adjustment Step 2).
Your final vertical heights will come from the fully constrained orthometric height (vertical)
adjustment (Adjustment Step 4).
Step 4. Fully constrained orthometric height (vertical) adjustment
Constrain one CORS (2D) LL only for alignment
- Provides horizontal orientation
Constrain to all "valid" NAVD 88 bench marks (Vert-only) and
then Select GEOID12A (for current NSRS results)
Enter the published NAVD 88 leveled orthometric heights from
the IDB mark datasheets for each bench mark.
- Compare each computed orthometric height to the published orthometric heights for all bench
marks constrained in the project. The difference or shift of the
computed orthometric heights
for unconstrained stations from the minimum and full vertical constrained adjustment should be
less than what your vertical tolerance is.
This adjustment produces the final set of adjusted NAVD 88 orthometric
heights for the project marks.
*
This is normally a trial and error process and you may want to not constrain all
of the NAVD 88 heights on every bench mark but only select those that seem to
provide the best solution. Ask yourself these questions. Which bench marks
seem to be the most stable? What are their order classifications? Are some in
bedrock, or rods driven to refusal etc? Do some bench marks in the same level
line appear to have subsided or have been uplifted? Do some bench marks have
a status of "posted" meaning that there were issues with either the original field
work or with adjustment processing?
User Instructions and Technical Guide
OPUS Projects
88 |
P a g e
NOAA |
National Geodetic Survey
3. 5.0.1 Considering Constraint Weighting in Network Adjustments
A. First, remember that the results are from a least-squares solution. By its nature, least-squares will
"push" errors anywhere it can to minimize residuals. The partials used in creating the observations
equations (and, in turn, the matrix to be solved) try to direct where those errors should go, but they'll go
anywhere they can. A least-squares solution is not a collection of averages - at least in the way we think
of averages. If the models are good,
the errors are random, the unknowns fit the problem and the
adjustments are small, it can act like a collection of averages, but it is not.
B. The individual session solutions probably "fit" very well. In other words, each individual sessions'
residuals are small. So, again, we have an inconsistency. Each session solution is saying that the mark is
exactly where it should be with high confidence - by the way, so high that we're fixing ambiguities =
fixing distances to satellites whose positions we do not allow to shift at all - and we're
rewarded with
small residuals from the sessions solution. Except that we, in our position to know all thanks to the plot,
know that a mark appears to be in two different places. The session solutions don't know this and don't
care. They only know their data fits their adjustments very well.
C. But then we put the session solutions together in a network adjustment. Now think about what
we're telling the network adjustment to do. We're saying that even though the session results clearly
put the offending mark in two different vertical
locations, the network adjustment is given no means for
that to happen. So where can that inconsistency go? Not into the ambiguities, those were fixed by
session. That really only leaves the coordinates of the thing(s) that are defining the frame in which the
measurements were made, i.e. the CORS, and the tropos which strongly alias into the heights.
D. The constraints are not barriers. The NORMAL constraints instruct the program to limit the
adjustment to less than about 1 cm, but if it "wants" to be greater than 1 cm, it will although the
constraint will increasingly hinder taking on larger and larger values. Think of it as a rubber band. For
adjustments less
than a couple cm, the rubber band is not stretched tightly and the adjustment can
"easily" reach (take on) any of those values if the solution demands. Beyond that, the rubber band is
stretched tighter and tighter. The adjustment can reach (take on) larger values, but its
got to work
harder to do so. But be aware that the rubber band is
never
slack. It is always "pulling" towards the
specified constraint value to some extent.