Data-Driven Stochastic Modeling of Dual-Frequency GNSS Measurements Using Cycle Slip Parameter Variance

This study introduces a novel data-driven stochastic modeling technique developed to enhance GNSS measurement noise modeling under varying geomagnetic conditions. The technique is applied to three consecutive days of multi-GNSS data from nine globally distributed IGS stations, encompassing calm and extreme geomagnetic states triggered by a large solar flare. Evaluation of the new model is compared against a traditional elevation-based approach, highlighting its adaptability and robustness across different GNSS systems: GPS, GLONASS, Galileo, and BeiDou. Results show that while pseudorange noise remains stable for each system and across days with and without geomagnetic activity, carrier phase measurement noise is modeled as 21.8%, 58.4%, and 78.8% higher than GPS noise for respective BeiDou, Galileo, and GLONASS systems under extreme geomagnetic disturbances, with lower increased phase noise under calm conditions. Pseudorange noise is modeled as more variable between systems, with at least 19.4% GLONASS noise increase and up to 4.2% BeiDou noise decrease compared to GPS processing, while Galileo pseudorange noise is at least 15.1% better than GPS on all days. These differences reflect the advantages of modernized signals over legacy ones and the ability to accurately model these effects using the new stochastic approach.