The effect of different tropospheric models on precise point positioning in kinematic mode

Abstract

Precise Point Positioning, based on undifferenced dual frequency GPS carrier phase observations, is a relatively new data processing technique for high accuracy kinematic GPS applications. No reference station data is necessary making the technique applicable e.g. for airborne high accuracy GPS positioning in remote areas where the distance to the nearest GPS reference station otherwise would be hundreds of km. High accuracy kinematic GPS positioning is used, for instance, for airborne remote sensing with gravimetry, In SAR, or lidar equipment, where data is collected for various geophysical applications. In the remote areas of Greenland, Northern Canada and the Northern parts of Scandinavia airborne remote sensing is carried out by various professional groups every summer, and most of these groups would like improved and more reliable high accuracy GPS positioning algorithms.The atmospheric effects on GPS satellite signals are significant. The dispersive effect of the ionosphere can be sufficiently addressed by the use of dual frequency observations. To mitigate the non-dispersive atmospheric effects, tropospheric a priori models are normally used. When observing for extended periods of time, additional tropospheric parameters can also be estimated as part of the adjustment process. In a differential mode and operating not too far from the reference receiver, most of the residual errors are eliminated in the differencing process. For Precise Point Positioning, however, and especially when processing shorter time spans of data in a kinematic mode, the accuracy of estimated positions heavily depends on the a priori models.This paper investigates the use of the Saastamoinen and the UNB3 global tropospheric delay models as well as the use of tropospheric delay estimates derived from numerical weather predictions. The various tropospheric correction approaches are tested for precise point positioning, and the positioning results are evaluated by comparison with known station coordinates. With the data and test scenario used the evaluation show similar standard deviations for all three approaches, but the Saastamoinen model performs with the smallest bias when the position results are compared to the known position.The tests are based on GPS data collected at 14 different sites in Denmark and Southern Sweden, and the numerical weather predictions available are from the HIRLAM system implemented at the Danish Meteorological Institute. The data processing is carried out using the ABSPOS software developed at the Norwegian University of Life Sciences.