A Bayesian Inference-Based Empirical Model for Scintillation Indices for High-Latitude

Solar wind parameters, the solar radio flux index (F10.7), the Sun's declination and the SuperMAG Electrojet index are used to construct a Bayesian inference-based empirical model for scintillation indices (S₄ and σ_Φ) at high latitudes. For the present study, measurements from three Global Positioning System (GPS) L1 receivers located in the auroral zone, the cusp and in the polar cap are selected, respectively. The solar wind characteristics include the solar wind speed (V_SW) and ram pressure (ρ_SW) as well as the Geocentric Solar Magnetospheric (GSM) B_y and the B_z components of the interplanetary magnetic field (IMF). Following a brief assessment on the independence of the variables (predictors), prior probabilities of occurrence in the case of a multinomial classification are constructed. Posterior-probabilities are then deduced for any arbitrary set of predictors. We show that the model captures most variations seen in the measured indices whether they are associated or not with transient interplanetary events. Although the model tends to underestimate the actual phase index measurements, 95% of the validated events are predicted with an error less than 0.034 rad in σ_Φ. For the amplitude scintillation index, 5% of validated events have an error larger than 0.019.