This thesis would not exist, without the support of my peers.

My sincere gratitude to my supervisor Mikael Lüthje for his continuous support, supervision and guidance through the perilous cliffs of pursuing a Ph.D. in a new and ever-evolving academic center. The open discussions and trust in my ability allowed me to thrive during this period. My appreciation extends to Colin MacBeth, who in the position as my external co-supervisor welcomed me to Edinburgh and provided further guidance and insight into the practical workings of 4D seismics. I am deeply indebted to my internal co-supervisor Anders Nymark Christensen, who provided valuable insight into the statistical working of machine learning and kept my weights and biases in check. Thank you for inspiring me to achieve more than I ever thought possible.

To the friends we had and made along the way! Kirstie Wright and Anna Clark you kept me sane from day to day and made Scotland feel home. Thank you. Robert Leckenby, Tim Albrecht, Bettina Schmidt, Matthias Schneider, Manuela Köllner, Clara Dabrock, Brian Burnham and Florian Smit, you were always there and I appreciate you for it. Marie-Daphne Mangriotis thank you for welcoming me to ETLP and the great conversations. Furthermore, I would like to thank Antony Hallam and Gustavo Corte for great discussion and peership.

My thanks go out to the Software Underground community, for keeping the spirit of sharing and collaboration. Especially, Matt Hall for the leadership and trust. To Lukas Mosser, for always encouraging and inspiring me to strive for more.

I would like to thank my colleagues at the DHRTC, particularly Tala Maria Aabø for being a fantastic co-conspirator in the early days and Florian Smit for welcoming me into this new environment. Furthermore, I’d like to thank Frédéric Amour, Charlotte Lind Laurentzius, Anne Lysgaard and Helle Baumann for being a pleasure to work with.

I want to thank part of the Open Source community and in particular the scientific Python community, without these tools this thesis would be much less substantial.