by Antoine Caté, PhD, PGeo,
Senior Consultant (Structural Geology), SRK Consulting (Canada)
Throughout my evolving career journey as a geologist, I have logged my fair share of meters of core on many projects, in highly variable geological environments, and for a multitude of purposes. As times change, in the last eight years I have also been involved in developing solutions to automate, in part, core logging using tools with fancy names such as machine learning (ML), computer vision, and recently—large language models (LLMs). Looking back at this experience, I am both amazed by and concerned about the value that geological logging provides to mineral exploration.
The extraction of information from drill core is the very point of drilling in mineral exploration. Information can be sourced from geochemical analyses or the measurement of rocks’ physical properties but, arguably, the one dataset that is collected the most is geological descriptions compiled by the geologist logging the core.
These descriptions are of high value to exploration. One example is a project where the client wanted me to investigate a very high-grade intersection that did not seem to have any continuity in the neighboring drill holes. With only visual observation of the core, I could see a thin gold veinlet that was within the foliation of a shear zone at a very low angle to the core. The current drilling pattern was not optimized to intersect this shear zone. Because the core was oriented, I was able to model the orientation of the gold-bearing shear zone and advise the client to drill new holes with a different orientation that would better intersect the structure.
With less than 1 m (3.28 ft) of core from the right place, one can get incredible insight and create a lot of value for an exploration project. However, it remains common to hear modeling or resource geologists complain about the quality of geological logs. Whether it is poor quality of the geological descriptions, or confusing, never-ending lists of lithology codes, or the important information not being collected at all, shortcomings in geological logging data are far too common. These result in unnecessary drill holes, errors, uncertainties in geological models and resources, and lead to missed exploration opportunities.
In the last 15 years, new technologies such as core scanners, high-resolution geochemistry, and machine learning have been introduced in the core logging industry with the promise of making geological descriptions systematically reliable and removing all the biases introduced by the all-too-human geologist. These new technologies do indeed provide value for exploration. As an example, we realized on a project that gold mineralization is hosted in shear zones that no one had bothered to log. We then trained a machine learning model to recognize foliation intensity in core images from which we could model these shear zones and define the gold mineralization domains.
New technologies have a clear value in extracting information from drill core. However, these technologies never addressed the underlying issue with geological logs: geological logging is often left to more junior staff, with a lack of coordination to improve logging quality. The same degree of care for quality control on assays is often not taken for the quality control of geological descriptions. The most significant mistakes in exploration and resource estimation are often related to geological interpretation errors, and not to mistakes in the collection and analysis of assay data. In many projects, the uncertainty in the mineralization volume has more impact than the uncertainty in the grade within the volume.
Unless more thought is put into geological data acquisition and interpretation, the use of advanced logging technologies will not resolve the issues in logging reliability. It will only help fill servers with gigabytes of data. The most important aspect is to put thought and effort into geological logging. The task is often completed by junior geologists after minimal training on a project, with minimal support, and with little feedback from the end-user of the data. For the success of a project, logging should be closely managed by experienced geologists with the support of domain experts when and where required. Experienced geologists should regularly look at the core and the resulting data, and lead the task of improving how the data is collected, used, and interpreted. This is implemented in some projects, and the results are worth the effort.
In a recent project I worked on, the team implemented a systematic well-managed core orientation along with using two core scanning technologies. Instead of just storing the data, they used it to improve and gain confidence in their geological interpretation. The result was an outstanding understanding of the project geology considering the relatively widely spaced drill holes. One could argue that the extra effort put into logging leads to having to drill less and thus saving money in the end.
We all know what to do, and we all know what the value in it is. This is not about revolutionizing how core logging is done but about implementing best practices and incrementally improving how we do things. Advanced technologies are an excellent tool to help us get more value from rocks. But the ‘core’ of the solution is human effort.
For more information: Get in touch with Antoine on LinkedIn
