Network with 2 ref. points in 2D and 1 ref. point in 1D

I am trying to solve a 3D network that has totally 6 points. There are 2 points with known (fixed) XY and 1 point with known Z. There are also 3 other points that are unknown in all XYZ.
What I want to do is solve the network so that the 2 points with known XY get a solved Z, the point with known Z gets solved XY, and the 3 points with unknown XYZ get solved in XYZ.
I thought that I have to import the 2 known XY points in "Reference Points 2D" and the one with known Z in "Reference Points 1D". I also imported the other 3 points in "New Points 3D". The observations are directions, Slope distances and Zeniths.
The software fails to solve the network and I get the message "Singular normal equation matrix. Error, could not invert normal equation matrix.". I suppose there is something wrong with the way I imported the known points, isn't it? If so, what it the right way to import points when some are known and fixed in 2D and some others in 1D?
By the way, if all points are imported as "Reference points 3D", the network gets solved.

Hello geozap,

Please take a look at the table of observations in the wiki, especially which types are allowed for the individual point dimensions.
https://software.applied-geodesy.org/wiki/least-squares-adjustment/observation

For your approach, this would mean:
for 1D, only height differences; for 2D, directions and horizontal distances; and for 3D, directions, slope distances, and zenith angles.

This is because JAG3D strictly distinguishes between a true 3D network and the 2D+1D approach. Regardless of this, however, such an approach is possible without any problems. You just have to prepare your observations yourself, as JAG3D deliberately does not automatically convert, for example, slope distances and zenith angles into horizontal distances.

I hope this helps.

Best regards
Pierre

Thanks for the clarification.
So, the network cannot be solved in one step as 3D. I have to calculate the height differences outside jag3d, then consider them as leveling observations and solve the network as 1D. In a analogous manner, solve the network as 2D, using calculated horizontal distances and directions.

Hello,

I have to calculate the height differences outside jag3d, then consider them as leveling observations and solve the network as 1D. In a analogous manner, solve the network as 2D, using calculated horizontal distances and directions.

That is the usual procedure, yes.

So, the network cannot be solved in one step as 3D.

It is also possible to process the network in 3D with additional pseudo-observations. In this case, you don't have to divide the spatial observations into corresponding planar and height components. To perform such a network adjustment, import all points as unknown 3D points - including your reference points. Import the height reference as 1D reference point and the 2D reference points point BUT with different names (e.g. add a suffix or prefix). Define GNSS baselines between the 3D "reference" points and the true reference points to constrain the adjustment.

Example: Let H be the reference height. Import the point H as 3D point. Thus, spatial observations from and to H are defined. Define a 1D reference point say H_REF having the known height. Import this reference point. Moreover, define a 1D-GNSS baseline. To constrain the height component, the height difference must be zero having a small uncertainty (last column 1 µm), i.e.

H  H_REF   0.000     0.000001

In almost the same manner, the XY components of the planar reference points can be fixed using 2D baselines.

/Micha

--
applied-geodesy.org - OpenSource Least-Squares Adjustment Software for Geodetic Sciences

RSS Feed of thread