Bayesian fusion of satellite-based precursors for upcoming large earthquake magnitude prediction

The reliability of an earthquake early warning system critically depends on minimising uncertainty in estimation of the key earthquake parameters, including time, location and magnitude. The growing availability of LAI (Lithospheric Atmospheric Ionospheric) seismic precursors derived from satellite observations highlights the need for multi-precursor data fusion to achieve more accurate and robust earthquake parameter estimation. This study applies a Bayesian Fusion (BF) framework to integrate multiple satellite-derived precursors observed in the spatial and temporal vicinity of three recent major earthquakes: Turkey (6 February 2023), Myanmar (28 March 2025) and Russia-Kamchatka (29 July 2025). This study is conducted in three stages: 1) estimating the intensity of anomalies; 2) determining the earthquake magnitude; and 3) applying a BF algorithm to integrate various parameters derived from multiple precursors in order to estimate the final earthquake magnitude, along with its associated uncertainty. The proposed approach is evaluated against previously established method, i.e. Fuzzy Inference System (FIS). Comparative analysis demonstrates the effectiveness of the BF method in reliability of final earthquake magnitude estimation.