Quality Control and Its Impacts on TEC Modelling With GPS Radio Occultation Data and Ground-Based GPS Data

Abstract

The accuracies of TEC models derived from groundbased GPS data are affected by several factors including assumed ionospheric shell height, satellites geometry, the chosen satellite cut off angles, the quantity of the GPS measurement data and etc.. Potential outliers within the measurements can have a significant impact on the quality of the estimated TEC models. However, there have been no any discussions on the outlier detection procedures in the literature on TEC modelling. This paper for the first time extends the commonly used Truncated Singular Value Decomposition (TSVD)-based TEC modelling method with an outlier detection and exclusion procedure. Furthermore, the retrieved TEC is transferred as vertical electron density (VED) profiles using the Shape Function model for the outlier impact analysis. In the paper, the biases from the outlier and the Shape Function model error can be separated and evaluated with the use of ionosonde data. To test the performance of this proposed quality control procedure, a day-time data set was used to produce the VED profiles at 3 different locations in Australia where both GPS and ionosonde data are available. The real-data tests have shown that the proposed outlier detection and exclusion procedure can numerically evaluate the impact of outliers and the Shape Function model errors, which are time and location dependent, in terms of the relative RMS measure commonly used in the ionospheric modelling studies. For example, the above two influences are 17.95% and 45.95% at Townsville, and 13.13% and 13.31% at Canberra, respectively. Besides, the simulated data were also used in the analysis. The result has demonstrated that as the magnitude of the outlier increases, the estimated VED can be biased by over 50%. Therefore, it is necessary to include a quality control procedure in retrieving an accurate and reliable TEC distribution.