Sensitivity analysis of ionospheric tomographic imaging with Gaussian Markov random fields

Abstract

The reconstruction of electron density in the Earth's ionosphere is a challenging task due to the inherent uncertainty in indirect measurements. To overcome these limitations, we apply Gaussian Markov random fields (GMRFs) in a Bayesian framework to ionospheric tomographic imaging within the TomoScand project. By incorporating spatial correlations, GMRFs improve reconstruction stability and accuracy, addressing the ambiguity that typically characterizes ionospheric tomography.

This study also explores the sensitivity of reconstruction quality to different computational setups and voxel resolutions. Systematic optimization identifies configurations that balance precision with computational efficiency. Additionally, parallel computation is optimized to handle large-scale tomographic tasks, with benchmarking across core counts to achieve optimal performance.

These advancements contribute to developing more precise ionospheric models, supporting real-time applications like space weather forecasting and satellite communication. This thesis offers a starting point for further research in ionosphere models and atmospheric tomography.