Københavns Universitet
Uafhængig af ledelsen

PhD thesis defense

Mads Lorentzen, IGN, defends his thesis at Geology Section

PhD thesis defense — Mads Lorentzen 29 AUG


Date & Time:

Aud. C, Department of Geosciences and Natural Resource Management, Øster Voldgade 10, 1350 Copenhagen K

Hosted by:
Geology Section


Mads Lorentzen defends his thesis,

Applying supervised learning methods to geophysical data from the Lower Cretaceous succession in the Danish North Sea.
Uncovering the heterogeneous chalk reservoir

Professor Lars Nielsen, IGN
Professor Klaus Mosegaard, NBI
Senior Researcher Kenneth Bredesen, GEUS

Assessment Committee:
Professor Tor Arne Johansen, University of Bergen – Norway
Professor Alireza Malehmir, Uppsala University – Sweden
Associate Professor Majken C.L. Zibar (chair), IGN

The Lower Cretaceous succession in the Danish North Sea remains one of the geological units of the Danish Central Graben region that is known to be difficult to map due to its highly heterogeneous and complex geology.
The main objective of this research is to characterize the Lower Cretaceous succession in the area around the Valdemar Field located in the Danish North Sea in terms of outlining reservoir quality and structural variations. This is done by a quantitative seismic interpretation study for the Lower Cretaceous reservoir zone using a workflow that includes seismic and rock physics forward modelling and a Bayesian amplitude versus offset (AVO) inversion for litho-fluid classification. Moreover, supervised machine learning methods are applied to address two of the challenges influencing the quality of the former quantitative seismic interpretation study, hereof (1) prediction of missing shear sonic logs in wells by exploiting information from neighboring wells, and (2) fault mapping using a neural network based on learning from post-stack seismic synthetic data.
The main results show that the quantitative seismic interpretation indicates favorable reservoir properties in the northern part of the Valdermar Field. Furthermore, the applied supervised learning models show good reliability in modelling missing shear sonic log data as well as automating the fault interpretation using seismic data. The results are useful for exploration and prospect evaluation in hydrocarbon production as well as geological CO2 storage and geothermal energy applications.

A digital version of the PhD thesis can be obtained from the PhD secretary Mikala Heckscher at