Coupling processes and travelling ionospheric disturbances: detection, characterization and forecasting with GNSS observations

The ionosphere is a dynamic region where energy and momentum from the Sun, magnetosphere, and lower atmosphere converge, generating complex plasma variability that affects global communication and navigation systems. Among the most significant manifestations of this variability are Traveling Ionospheric Disturbances (TIDs), large-scale wave-like perturbations in electron density capable of propagating thousands of kilometers. Despite decades of research, the mechanisms governing their generation and evolution remain only partially understood due to the multiplicity of coupling processes and observational constraints. This thesis develops a reproducible, interpretable, and data-driven framework for detecting and forecasting Large-Scale TIDs (LSTIDs) using Global Navigation Satellite System (GNSS) observations. The work is structured as a progressive methodological pipeline: the quantification of measurement and processing uncertainties, the development of an autonomous LSTID detection algorithm based on GNSS keograms and image-processing techniques, and the creation of probabilistic forecasting models linking ionospheric disturbances to solar and geomagnetic drivers. These models are optimized via ensemble learning, calibrated for reliability, and interpreted through explainable AI techniques to trace predictions back to their physical causes. Through the integration of empirical analysis, physical interpretation, and machine learning, this thesis advances the capability to monitor and predict ionospheric disturbances in near real time. As a result, it bridges observational ionospheric research with operational spaceweather forecasting, contributing to both fundamental understanding and applied geospace monitoring.