On estimating the phase scintillation index using TEC provided by ISM and IGS professional GNSS receivers and machine learning

Amplitude and phase scintillation indexes (S4 and σϕ) provided by Ionospheric Scintillation Monitoring (ISM) receivers are the most used GNSS-based indicators of the signal fluctuations induced by the presence of ionospheric irregularities. These indexes are available only from ISM receivers which are not as abundant as other types of professional GNSS receivers, resulting in limited geographic distribution. This makes the scintillation indexes measurements rare and sparse compared to other types of ionospheric measurements available from GNSS receivers. Total Electron Content (TEC), on the other hand, is an ionospheric parameter available from a wide range of multi-frequency GNSS receivers. Many efforts have worked on establishing scintillation indicators based on TEC, and geodetic receivers in general, introducing various metrics, including the Rate of TEC change (ROT) and ROT Index (ROTI). However, a possible relationship between TEC and its variation, and the corresponding scintillation index that an Ionospheric Scintillation Monitor (ISM) receiver would estimate is not trivial. In principle, TEC can be retrieved from carrier phase measurements of the GNSS receiver, as σϕ. We investigate how to estimate σϕ from time series of TEC and ROT measurements from an ISM in Ny-Ålesund (Svalbard) using Machine Learning (ML). To evaluate its usability to estimate σϕ from geodetic receivers, the model is tested using TEC data provided by a quasi-co-located geodetic receiver belonging to the International GNSS Service (IGS) network. It is shown that the model performance when TEC from the IGS receiver is used gives comparable results to the model performance when TEC from the ISM receiver is utilised. The model's ability to infer the exact value of the scintillation index is bound to Mean Square Error (MSE) = 0.1 radians2 when σϕ≤0.8 radians. For σϕ>0.8 the MSE reaches 0.18 and 0.45 radians2 in operative testing using ISM and IGS measurements, respectively. However, the model’s ability to detect phase scintillation from IGS TEC measurements is comparable to expert visual inspection. Such a model has potential in alerting against phase fluctuations resulting in enhanced σϕ, especially in locations where ISM receivers are not available, but other types of dual-frequency GNSS receivers are present.