Improving GNSS meteorology by fusing measurements of multi-receiver sites on the observation level

Abstract

In GNSS analysis, tropospheric modelling is done in the form of Zenith Hydrostatic Delay (ZHD), which can be empirically computed from surface pressure and temperature, and Zenith Wet Delay (ZWD), which is estimated together with the other unknown parameters.  Analysis methods based on undifferenced GNSS code- and carrier-phase observations like Precise Point Positioning (PPP), which can achieve millimeter-accurate positioning results, provide therefore also time-series of ZWD, which can be used for meteorologic applications. However, due to receiver noise and system characteristics like cycle-slips the accuracy as well as the precision of such ZWD estimates is limited. Thus, we propose a novel approach for sites, which have several receivers connected to a single antenna or which are separated horizontally by only a few meters. For such sites, one can simultaneously process multi-frequency GNSS data by fusing observations from several receivers, while estimating a common ZWD parameter.For this purpose, we have implemented a PPP algorithm based on an Extended Kalman Filter (EKF) approach, which has the advantage that ZWD estimates are available in real-time for meteorologic applications. We demonstrate that those combined ZWD estimates are superior to single receiver estimates in term of precision and accuracy. For the latter measure, we make use of a GNSS hardware simulator and show that the RMS between the simulated and estimated ZWD significantly decreases when having two or more receivers at that site. Based on real-data we show that this concept provides less noisy ZWD estimates which agree better with physical properties of the local wet refractivity field.Moreover, we demonstrate that fusing data from several receivers by estimating a common ZWD parameter improves also positioning accuracy and precision, in particular in the up-component. In order to properly combine observations from geodetic-grade and low-cost GNSS receivers, we present our adaptive Kalman filter approach, which adjusts the observation noise covariance matrix automatically during processing. The presentation concludes with an outlook on the usage of this approach for larger networks and answers the question how arrays of low-cost GNSS receivers can compete against geodetic-grade GNSS hardware in term of providing ZWD estimates for meteorology.