Sect_9.3.5

9.3.5 Multi-Session Network Processing

SECONDARY ADJUSTMENT OF A MULTI-SESSION NETWORK

The basic steps in a multi-session adjustment are:

Assemble, adjust and analyse each session solution independently using any of the strategies described in section 9.2.3. The aim is to be assured, as far as is possible, that the GPS reductions (single baseline or multi-baseline) are correct by carrying out some quality control tests. The outcome is a set of station coordinates, and their VCV matrix, in which one station has been held fixed.
Assemble each session solution into a network (or multi-session) solution. This is generally done in a sequential manner: appending sessions one at a time, checking that each new session does not cause the expanding network to come apart. The tool for this is the secondary network adjustment program. (Some further comments on this are given below.)
Following the successful assembly of all the sessions into a single network solution, an iterative refinement process is commenced. Step (2) aimed to identify problem data (baselines that are not consistent with the overall network). The refinement would involve close analysis of the relative weighting of individual sessions within the overall adjustment, to take into account any factors that differentiate the quality of one session solution from another. The primary test that is applied to the overall multi-session adjustment is the variance factor test, or some variant of it.
The outcome is a set of coordinates, and their associated VCV matrix. The solution is minimally constrained (if only one datum station is involved), with the coordinate results referring to a geocentric datum that is "near" WGS84 (section 11.1.2).

Combining Different Session Solutions: Some Considerations

A multi-session solution can be obtained by:

Combining all previously adjusted single session solutions in one-step, or

Combining sessions sequentially and building up the network step-by-step.

The latter strategy is generally preferred because if a problem with the data is encountered, its source can be identified more easily than if the "trouble-shooting" had to be carried out on a one-step multi-session adjustment. The presence of more than one error, possibly in different session solutions, may "smear" the impact in such a way as to make unambiguous detection of the source of error(s) difficult.

No matter which strategy is adopted, the following must be considered:

Network Data Types. As discussed earlier, if baseline data is input into the secondary adjustment the output is a set of coordinates, with the resulting VCV reflecting the correlations that exist between the stations. To then input the single session solution into a multi-session adjustment requires that the adjustment software be able to handle station coordinate input. Hence the software may have to handle baselines at one adjustment level (single session), and then coordinate data at the next adjustment level (multi-session).
Datum Transfer and the Minimally Constrained Condition. This requires that the fixed station in each session solution is "unfixed", and hence variances and covariances terms inserted into the station coordinate VCV where formerly null matrices were present. This cannot be rigorous, although some estimate may be made, using the VCV information from a previous session (also containing the fixed station, but being a "free" station within that earlier session). In the case of baseline data, as with the data type problem mentioned above, the original baseline data can be input into the final multi-session adjustment, the datum may be defined in terms of the apriori coordinate variances of the fixed station and no approximation is made as session is appended to session.
Choice of GPS Datum. The precisions in the VCV matrix of a (multi- or single) session solution are dependent on the geodetic datum definition. In a minimally constrained secondary session adjustment the choice of fixed station will influence the size and orientation of each station's 2-D error ellipse (or 3-D ellipsoid), as illustrated in the figure in section 9.1.3. The relative error ellipses/ellipsoids are insensitive to the datum choice, and it is these which ultimately define the "class" of the GPS survey (section 10.2.2).
Mixing Different Receivers or Solutions. Initially, there is no "re-scaling" of session VCV matrices before combination into a multi-session solution. (If one session solution is an ambiguity-free one, and the other an ambiguity-fixed one, it is assumed that this will be reflected in the VCV matrices.) Re-scaling is generally attempted during the "refinement" stage. For example, if dual-frequency GPS receivers were used during one session, and single frequency receivers in another session, re-scaling may be required. The situation is even more difficult when combining GPS surveys from different eras. A great deal of information concerning receivers used, field and processing procedures, etc., must be available to make an accurate assessment of the manner in which individual VCV matrices are to be scaled, or modified in some way -- SINEX is a development to overcome this.
Variance Factor Tests. Any geodetic adjustment requires some action to be taken when the variance factor test fails. A GPS secondary adjustment is no exception.

Back to Chapter 9 Contents / Next Topic / Previous Topic