Impact of stochastic modelling on GPS height and zenith wet delay estimation

Abstract

Most stochastic modelling techniques assume the physical correlations among the raw observations to be negligible when forming the variance covariance matrix of the GPS observations. Such an assumption may, however, lead to significantly biased solutions. The Minimum Norm Quadratic Unbiased Estimation (MINQUE) method is an iterative technique that can be used to estimate spatial correlation among GPS measurements. Studies by previous authors have shown that MINQUE improves the accuracy and the reliability of the ambiguity resolution, and ultimately, the geodetic solution. However, its effect on the estimation of zenith wet delay (ZWD) is somewhat unknown. In this paper, an investigation into its impact on ZWD, as well as heighting, is carried out using simulated data. The results obtained from MINQUE for an observation window of five-days in static mode indicate an average improvement of 51% and 71% in the station height precision when compared against elevation-angle dependent and equal weighting models, respectively. This development, however, did not translate into better ZWD estimation, for which the differences between each respective stochastic model are generally at the sub-millimetre level.