Integration of geophysical and remote sensing data with AI for mineral exploration targeting

Year
2026
Primary Supervisor
Chris Green <[email protected]>
Institution
University of Leeds (School of Earth & Environment)
Possibility of becoming a CASE project
No
Collaborative Project
No
Academic Supervisors
Arash Rabbani <[email protected]> (University of Leeds, School of Computer Science), Taija Torvela <[email protected]> (University of Leeds, School of Earth and Environment)
Research Themes
Project Partners
None
Research Keywords
, , , , , , ,
Relevant Degree Courses
, , , , , , , ,

Summary

  • Apply a range of map-based geophysics/geoscience/geographical data sets to target exploration for new mineral resources.
  • Integrate gravity, magnetic, radiometric and remote sensing data sets.
  • Apply data enhancement techniques to develop structural geology models.
  • Integrate data sets numerically and compare with existing geological/mineral deposit information to identify effective approaches.  Apply Convolutional Neural Networks (CNNs) with attention mechanisms for hyperspectral feature extraction.
  • Develop robust approaches to apply to new data sets and new areas.
  • Field surveying/mapping for validation and assessment.
Schematic showing a series of maps of different geoscience data sets combining with AI to generate enhanced mineral prospectivity
Schematic of potential workflow for the project.

 

 

Background

A range of materials including metals are needed to support the energy transition and a substantial amount of that is predicted to come from new mineral deposits.  The exploration and production of these resources in a sustainable manner is critical to this process.  This project addresses this requirement by developing approaches to highlight prospective areas based on pre-existing or remote sensing data prior to more local investigations.

Gravity and magnetic data have long been mainstays of geophysical mineral exploration – especially magnetic methods as high quality data can be rapidly acquired at good resolution and relatively low cost from airborne surveys.  Airborne gravity gradient data have also been found to be valuable in this respect.  A range of data enhancement and semi-automated interpretation techniques have been applied to highlight structures and their associated depths from magnetic and gravity data.  Whereas some mineral deposits can be identified directly from these data sets, most depend on building a structural and geological model based on magnetic, gravity and other data together with ground truthing.

Radiometric and electro-magnetic data sets are sometimes acquired over large areas and have also been important for geological mapping and mineral exploration.  At a later stage, magnetic, electric and other geophysical methods, together with geological mapping and investigation, geochemistry, etc., have been applied at a local scale to provide more detail of a prospect.

Satellite based remote sensing systems have also been found useful for highlighting particular minerals in the near-surface and hence for geological mapping.  These satellite data sets have multiple channels which have historically been combined in particular ways to give best correlation with observed geology/mineralogy.  Recent and upcoming satellites are hyperspectral with large numbers (200+) of channels producing many layers of data to be analysed.    Recent ML advances in hyperspectral mineral mapping achieve >85% lithological classification accuracy. The recent development of Attention-based layers in ML was found to help the models focus on the most relevant parts of the data and they have shown promising results in identifying subtle geophysical signatures. However, these models have not been previously used for mineral exploration targeting.

 

Aim and objectives

The main aim of the project is to develop one or more effective approach to highlight mineral prospectivity based on gravity, magnetic, radiometric and remote sensing data.

The direction of the project will be largely driven by the Postgraduate Researcher, but specific objectives could include:

  • Apply data enhancements to gravity, magnetic and radiometric data to identify features and structures of interest. Integrate with geological data to construct a geological/tectonic model.
  • Investigate satellite image data processing and combination to assess correlation with geophysical results and known mineral deposits.
  • Explore attention-based CNNs for automated data fusion and identification of significant geophysical features for prospectivity mapping.
  • Ground truth results indicating high prospectivity.
  • Develop best practice approaches.

 

Project approach

The idea of the project is to work in areas of the UK where both known mineral deposits and high resolution aeromagnetic and radiometric data are available (SW England or Northern Ireland). Electromagnetic  data are also available for Northern Ireland.  Gravity data are available for both areas.  Remote sensing (multispectral or hyperspectral) data are also available (e.g. Sentinel-2).    In SW England, there are multiple known tin, lithium and tungsten deposits, and in Northern Ireland, there are three gold deposits or prospects, one of which is mined. These provide case studies against which geophysical signatures can be compared, in order to establish what geophysical signatures or parameters may be associated with metallic mineralisation.  If regional analysis indicates prospective locations, limited field investigation (e.g. geophysical survey or geological mapping) could be used to assess the results.

 

Potential for high-impact outcome

Any improvement in targeting mineral resources has the potential to improve the speed of the process and reduce costs.  Working with pre-existing and/or remote sensing data to focus detailed work will also limit local environmental impact.

Developing effective ways to integrate a variable stack of data layers with different resolution, especially employing machine learning, has potential applicability in a range of geographical mapping and modelling problems.  The project is suitable for either a career in industry or in academia. We envisage that the student will be able to publish up to three scientific papers arising from this project.

Researcher profile

Applicants should have at least a 2:1 (or overseas equivalent) from a BSc degree (or equivalent) in geology, geophysics, earth sciences or a similar discipline. An appropriate Masters degree would be beneficial. The project would suit a student with good numerical skills who has an aptitude for analysing map-based data sets and is comfortable with learning, using and possibly developing new software.  Some specific skills (particularly in geophysics, programming and machine learning) can be developed by sitting in on modules from our Masters programmes.

 

Recommended Reading

  • KUMAR, C., CHATTERJEE, S., OOMMEN, T. & GUHA, A. 2020. Automated lithological mapping by integrating spectral enhancement techniques and machine learning algorithms using AVIRIS-NG hyperspectral data in Gold-bearing granite-greenstone rocks in Hutti, India. International Journal of Applied Earth Observation and Geoinformation, 86, 102006.
  • SHIRMARD, H., FARAHBAKHSH, E., MÜLLER, R. D. & CHANDRA, R. 2022. A review of machine learning in processing remote sensing data for mineral exploration. Remote Sensing of Environment, 268, 112750.
  • YENNE, E. Y., GREEN, C. & TORVELA, T. 2024. Implications to basin evolution from the interpretation of superficial and buried geological features from remote sensing and magnetic data sets, Lower and Middle Benue Trough, Nigeria. Results in Earth Sciences, 2, 100029.