Advanced Receiver Autonomous Integrity Monitoring (ARAIM) based on multi-frequency and multi-constellation GNSS is one of the future architectures for aviation integrity. The goal of ARAIM is to provide a worldwide LPV-200 precision approach by identifying and removing faults autonomously in the receiver. Generally, the user calculates Horizontal and Vertical Protection Levels (HPL, VPL) in order to satisfy the integrity requirement of the ARAIM system by using the standard deviation of GPS error sources. In such cases, Signal-In-Space error is the dominant factor among the error sources in the magnitude of the user confidential bounds, because the system based on multi-frequency autonomously removes ionospheric errors which are the most dominant factor in single-frequency. The availability of ARAIM could therefore be improved through the reduced conservatism of protection level by means of precisely modeling the SIS error. This paper uses a bottom-up method to compute the worst-case SIS errors using broadcast navigation messages from IGS and precise ephemerides/clocks provided by NGA during the period 2009-2012. We conduct a statistical analysis of SIS errors, an analysis of the relationship between URE and URA, and URE performance tests regarding GPS III specifications. From the results, we confirm that SIS error has a nonzero mean for most of satellites, and the full SIS error distributions fail several evaluation criteria of URA tests, although URA is conservative to bound rms URE. Also, this paper shows that SIS error data sets could be partitioned into two subsets by the atomic clock types of satellites, and we expect that error models are likely to be subdivided by them.

Keywords: signal-In-space error, statistical analysis, ARAIM