Advances in Subsurface Data Analytics
I hoped that Advances in Subsurface Data Analytics would provide a broad introduction to machine learning and automation techniques. With a background in geology and marine geophysics, I approached the book thinking that it would be advantageous to better understand these trending innovations that are increasingly adopted within my industry. Many geophysical processes and tasks are progressively being automated and the benefits of this are tangible in terms of time, costs spared, and reproducibility of results.
Although analytics seemed a daunting topic, I was encouraged by the textbook’s devotion to subsurface themes and how it clarified that automation can be applied to some common and familiar geoscience problems. Techniques covered included traditional machine learning approaches, such as artificial neural networks, random forest classifiers, principal component analysis, self-organising maps, and multi-attribute analysis. Deep learning techniques were also given consideration, particularly convolutional neural networks and recurrent neural networks. These methods were assessed with routine end objectives in mind, such as the classification of reservoirs, identification of lithologies in core samples, detection of faults, prediction of acoustic velocity in substrata, full-waveform inversion, and computational fluid dynamics. Overall, the practical usefulness of different techniques was well highlighted.
Advances in Subsurface Data Analytics is comprised of multiple, thematically linked papers that were brought together coherently, with appropriate context and structure. I found the content to be well-written, scientifically detailed, and easily intelligible. Having said that, the book will likely provide a steep learning curve for anyone not well initiated with machine learning. Though it did help to lightly familiarise me with these methods, the more technical content went over my head. I would recommend the book more for those who are actively involved in modelling or automation projects and hoping to further streamline their performance. The book assumes a strong background in the geosciences, such as geophysics, geotechnics, geology, or geochemistry. As declared by the editors, it is aimed at professional geoscientists working on sub-surface related research problems, or else those using it in an academic capacity. For those looking for a gentle entry into machine learning, this is probably not the tool, nor does it claim to be. This is a specialist’s reference book in all its detailed glory.
Reviewed by Leslee Salzmann
BY: S. Bhattacharya & H. Di (eds.) (2022). Elsevier. 376 pp. (ebook).
PRICE: £80.50 www.elsevier.com