Centring Errors and PPM Calculation of Measurement Standard Deviations

[KentWheeler]

Reason for Enhancement

• The propagated uncertainty of a survey measurement is 'observation Sdev' + 'centring Sdev'

• The majority of GNSS baseline processing software such as "Trimble business Centre" or "Leica Geomatics Office" determine GNSS baseline variance covariance matrices as a function of the satellite observation residual. Network least squares adjustments performed within these packages assign global centring errors that are in addition to the standard deviations of the variance covariance matrix. Therefore, exporting the processed GNSS baseline and adjusting in Dynadjust has the following problems:

  • Baseline observations are overly optimistic.
  • Scaling the variance covariance matrix does not accurately account for centring as it needs to be an additive constant rather than a scalar.
  • Variance covariance matrices determined as a function of satellite observation residuals may not always be the most suitable estimate of uncertainty.

Requirements

• Add the functionality to add the standard deviation of the centring error and/or instrument height to the observation standard deviation.

• Add functionality to allow observation standard deviations (and variance covariance matrices) to be recalculated from a 'ppm + Const' function.
  The following table shows the capabilities that would be required for each observation type.
  O - capability of current schema.
  X - requested enhancement.
  

Next Steps

• Modify DynaML schema to accommodate.

• Produce proof of concept for circulation and review. (I can work on a python script that will modify msr.xml files to allow centring errors for the current functionality of Dynadjust and circulate this with the ICSM - adjustment working group).

• Dynadjust development and testing.

• Documentation.

[KentWheeler]

A proof of concept script has been written in python to demonstrate this enhancement.
The currently only caters for G type observations and is available from this location.
https://github.com/KentWheeler/PynAdjust/blob/master/DynaStdDevStyles.py

[ccrook]

FYI I have had to deal with a similar issue in the LINZ SNAP software - in our National Geodetic Adjustment we have had some strange artefacts from unrealistic VCV matrices in SINEX files. I ended up adding options to modify the VCV matrix before it gets used in the adjustment documented here.

Apart from accounting for centring error I also found that some processing strategies used to generate the SINEX files created unrealistically precise locations for the centroid of the points. This is helped by adding plumbing uncertainties but I found it better to also add an uncertainty to the centroid.

I have also had issues of SINEX files with unrealistically poorly located centroids. This is not such an issue for calculating coordinates but can mess with the statistics, particularly if you then think the observations are better than in the SINEX file and scale the VCV down.