9.4.2 Modifying the Stochastic Model

AN EMPIRICAL METHOD OF DETERMINING REALISTIC VARIANCES FOR GPS BASELINE SOLUTIONS

The simplest method is to assign "reasonable" values for a and b, and to construct the variance quantities using eqn (9.4-1). These values are then used to modify the VCV matrix (using any of the methods described below), the adjustment is then re-run and the Variance Factor Test is re-applied. If the test still fails, then the values for a and/or b are changed, and the process repeated until the Variance Factor Test is finally passed. There are, however, several difficulties with this approach:

• Which quantity to vary? a or b?
• What if there were a variety of baseline lengths? Varying the quantities could have unexpected effects.
• What if a mixed set of GPS hardware was used (single or dual-frequency)?
• What if several types of field procedures were used (conventional static, "stop & go", "rapid static" techniques)?
• What if there was a mixture of ambiguity-free and ambiguity-fixed solutions?

In effect the values of a and b are functions of all of the above (and more)! Given these difficulties, a common method has been to estimate the values of s² directly (rather than its a and b components). An alternative method is to estimate a scale factor for particular subsets of baselines sharing similar solution characteristics that may cause systematic differences in GPS accuracy, for example, according to whether they are produced using the same software, the baseline lengths are similar, etc. It must be emphasised that this is a more sophisticated method of estimating scale factors than simply using the aposteriori variance factor in an iterative process. This is discussed in the next section.

Back to Chapter 9 Contents / Next Topic / Previous Topic

© Chris Rizos, SNAP-UNSW, 1999