10.18453/rosdok_id00004838 Styp-Rekowski, Kevin Kevin Styp-Rekowski http://d-nb.info/gnd/1369127588 https://orcid.org/0000-0001-8267-521X Universität Rostock 2024 004 Data processing Computer sciences 500 Natural sciences 530 Physics 2024 en https://purl.uni-rostock.de/rosdok/id00004838 urn:nbn:de:gbv:28-rosdok_id00004838-8 This research explores the evolution of Earth's magnetic field, emphasizing the importance of accurate data for analysis and prediction. The dissertation introduces a novel Machine Learning-based approach to enhance the calibration of platform magnetometers on non-dedicated satellites, addressing the challenges of their rough calibration. The methodology, applied to the GOCE and GRACE-FO missions, significantly improves data accuracy, enabling scientific application. This work increases data availability for geomagnetic studies and sets the stage for future applications in satellite missions.