Lobe 3
PhD studentship opportunity
Fully funded 3.5 year NERC DTP studentship available for October 2018 based at the University of Leeds.
Fans of machine learning: Deep-water distributive systems as ideal environments to test AI approaches to identification of subsurface depositional architecture
Supervisors: Prof Dave Hodgson, Dr Adam Booth
Project partner: Earth Science Analytics
Contact: D.Hodgson@leeds.ac.uk
The complicated subsurface geology offshore UK and Norway is riddled with boreholes collected over the last 50 years by the oil and gas industry. For economic reasons, over time less and less core per well has been collected, meaning an increased reliance on the interpretation of depositional environments and stratigraphic elements from well log data that can be weakly calibrated to seismic reflection data. However, as the industry moves towards more challenging prospects in the mature North Sea Basin, and as the need to test the suitability of abandoned fields as potential sites for carbon sequestration increases, the necessity to accurately and efficiently assess the vast archive of existing borehole data has never been more acute. Machine learning, an application of Artificial Intelligence based around giving machines access to data to learn for themselves, offers one approach to improve interpretation of complicated 3D depositional architecture from 1D data in the subsurface. However, for machine learning approaches to be successfully applied there is a need for software to be tested against well calibrated datasets. Furthermore, many of the depositional settings in the North Sea, and other sedimentary basins, are highly complicated in their architecture, have ambiguous well log patterns, and data are under-utilized due to a lack of time. This leads to disparate and inconsistent interpretations from industry practitioners.
Submarine fan systems provide a suitable environment of deposition to test machine learning approaches because there is an established architectural hierarchy and constrained range of dimensions of depositional elements. Furthermore, there are many mature fields with a large amount of publically available borehole data (e.g. the Forties Fan). However, the first stage of this studentship will be to test machine learning approaches against a well constrained analogue system; the Skoorsteenberg Formation, Karoo Basin, South Africa. We have collected a huge dataset of calibrated core and well logs, totalling more than 3 kilometres, through this submarine fan system. The advantage is that the close-to outcrop wells are tightly correlated to lobes, channel-fills, and levees. This unique resource will be used to advance the automated identification of architectural elements in subsurface settings in the North Sea with partially or uncored wells from submarine fan successions. The student will start in areas of dense data coverage in mature fields, and then apply the machine learnt protocols to areas with more sparse control on the subsurface stratigraphy. Ultimately, the development of machine learning to identify depositional environments in the subsurface will need to be tied to seismic reflection data, and to quantify what in the depositional architecture is imaged, to improve the population of 3D geological models.
The application of machine learning approaches to the subsurface is an essential step in extending the life of mature hydrocarbon fields, and also in evaluating the suitability of abandoned fields of sites for carbon sequestration and storage. In the absence of machine learning, this approach would otherwise take many man-years of work, and lack consistency in approach. This studentship is focused on a topic of international importance and will be affiliated to the Lobe3 project, which is an industry-funded research programme. We expect you to submit manuscripts to international scientific journals during the course of your studentship, and to present your results of their research at relevant national and International conferences. The project will provide you excellent training in computer programming, fieldwork, deep-water sedimentology and stratigraphy, and data analysis. You will join a large group working on earth surface processes and sedimentary basins.
Click here for further details
The application deadline is 5pm on Friday 6th July 2018, with interviews taking place on Friday 20th July.