Over the past ten years, the National Geoscience Data Centre (NGDC) (based at the British Geological Survey in Keyworth, Nottingham) have colour photographed and catalogued over 300km of core creating a database of over 150,000 images. International researchers from the Colorado School of Mines have used a sample of this data, from a submarine fan system to the west of the Shetland Isles, to develop and implement a machine-learning workflow to label rock characteristics. In some cases, access to such photographs can provide a cost-effective substitute to in-person observations from slabbed core material.

International researchers from the Colorado School of Mines were able to quickly and accurately label 659 metres of core from 12 wells using five basic rock characteristics labels, as well as 287 metres of core from 5 wells using a more detailed labelling schema. Whilst the accuracy produced by the study means that the results can’t fully replace core descriptions made by trained geoscientists, it is general enough and scalable to support large-scale subsurface investigations. These require consistent lithological identification e.g. for mining, hydrogeology, geothermal carbon sequestration, and geotechnical studies. It allows users to describe thousands of metres of core in hours rather than weeks, saving time and resource in visiting the analogue core material and in conducting repetitive core descriptions.  

Over the past ten years, EDS staff based at the NGDC have colour photographed and catalogued over 300km of core creating a database of over 150,000 images (off which circa 138,000 are images of offshore core). The photography was conducted using a standardised, high-quality acquisition specification (i.e. consistent lighting, camera, pixel dimensions, and metadata), ensuring consistency in image capture and providing researchers with a rich, open access data set for further analysis. High resolution digital photographs of core provide an objective, raw (uninterpreted) dataset which preserves the fine-scale heterogeneity of rock from the core sample. For some researchers, access to such photographs can provide a cost-effective substitute to observations made directly from slabbed core material. 

We specifically chose the British Geological Survey data because of this consistency, as many open-source datasets (e.g., Norway, New Zealand, United States Geological Survey, and various United States state repositories) have quite variable conditions of the core photographs, or no photographs at all (Martin et al, 2021). 

The study highlights the importance of having open and accessible, high quality, and standardised digital photographs of geological core. It also highlights how such data can be used to save time and resources to obtain specific geological interpretations and insight using standard hardware and open-source software. For more information about the core photography collection, please visit the borehole materials collection pages on the BGS website.

References

Martin T, Meyer R and Jobe Z (2021) Centimeter-Scale Lithology and Facies Prediction in Cored Wells Using Machine Learning. Front. Earth Sci. 9:659611. https://doi.org/10.3389/feart.2021.659611