Chapter 13

Further reading

The master bibliography, annotated and ordered by what I would actually read first. Free resources are marked. Read the papers rather than the coverage of the papers — the gap between them in this field is unusually wide.

If you are starting from zero on the industry

Three books, in this order, and you will be able to hold a technical conversation with a geologist.

• Robb, Introduction to Ore-Forming Processes (Wiley, 2nd ed. 2020). The single best book for understanding how ore actually forms, and readable without a geology degree. If you buy one book from this entire page, buy this one.
• Marjoribanks, Geological Methods in Mineral Exploration and Mining (Springer, 2nd ed. 2010). How the work is physically done — mapping, logging, sampling, drilling. The book that makes Chapter 3 concrete.
• Abzalov, Applied Mining Geology (Springer, 2016). The practitioner's bridge from field data to resource estimate, and the best published treatment of QA/QC design.
• MinEx Consulting (Richard Schodde) and S&P Global World Exploration Trends — the two sources for essentially every credible number about how much exploration costs and how well it is working.

Read these four first (for the ML)

1 · Mern & Caers (2023), The Intelligent Prospector v1.0

Geoscientific Model Development 16, 289–313 · open access · doi:10.5194/gmd-16-289-2023

The best single entry point to the whole field. Formulates sequential geoscientific data acquisition as a POMDP, solves it with POMCPOW, and reports that the sequential policy reaches the belief accuracy of a 16-point grid in 7–10 measurements. Read the POMDP formulation section carefully — the mapping from geology to ⟨S, A, O, T, Z, R⟩ is the conceptual core of Chapter 10. Code: github.com/sisl/MineralExploration (MIT), archived at Zenodo.

2 · Mern, Corso, Burch, House & Caers (2024), Intelligent Prospector v2.0

arXiv:2410.10610 · free preprint

Planning under epistemic uncertainty: multiple competing geological hypotheses, plus early detection that all stated hypotheses have been falsified. Tested on a sediment-hosted copper deposit; the abstract states the algorithm aided characterisation of an ultra high-grade deposit in Zambia in 2023. Note the author list — two KoBold Metals people, including the CEO.

3 · Kochenderfer, Wheeler & Wray, Algorithms for Decision Making

MIT Press, 2022 · algorithmsbook.com · free PDF

The textbook behind the machinery in Chapter 10 — belief states, POMDPs, online planning, value of information — with clean pseudocode. Not a geoscience book at all, which is the point. Same lab (Stanford SISL) that hosts the Intelligent Prospector code, so the notation lines up.

4 · Rasmussen & Williams, Gaussian Processes for Machine Learning

MIT Press, 2006 · gaussianprocess.org/gpml · free PDF

You may already know it. Re-read chapters 2 and 4 with Chapter 8 of this primer open beside them and the geostatistics translation will be immediate.

Prospectivity mapping

• Lawley et al. (2022), "Data-driven prospectivity modelling of sediment-hosted Zn-Pb mineral systems and their critical raw materials," Ore Geology Reviews 141, 104635 — ScienceDirect · USGS record. Gradient boosting on H3 hexagonal cells across the US, Canada and Australia. The de facto benchmark; its outputs and inputs are both open (Ch. 12).
• Lawley et al. (2021), Datasets to support prospectivity modelling for sediment-hosted Zn-Pb mineral systems, NRCan Open File 8836 — publications.gc.ca. The training data behind the above.
• Daruna, Zadorozhnyy, Lukoczki & Chiu (2024), "GFM4MPM: Towards Geospatial Foundation Models for Mineral Prospectivity Mapping" — arXiv:2406.12756. Masked image modelling pretrained on unlabeled geospatial data, then used as a feature extractor for MVT and CD Pb-Zn prospectivity. The clearest current statement of the self-supervised direction.
• Elkan & Noto (2008), "Learning classifiers from only positive and unlabeled data," KDD '08. Not geoscience — the origin of the g(x) = c · f(x) correction in Chapter 9. Short, and worth reading in the original.
• A PU-learning algorithm for MPM — Computers & Geosciences; weighted PU with TPE hyperparameter search — Ore Geology Reviews; and a survey of imbalance handling — Ore Geology Reviews. Read one of these to see how the theory is applied in practice, and how much variation there is in rigour.

Geostatistics

• Chilès & Delfiner, Geostatistics: Modeling Spatial Uncertainty (Wiley, 2nd ed. 2012). The rigorous reference. Dense, complete, and the place to settle an argument.
• Goovaerts, Geostatistics for Natural Resources Evaluation (OUP, 1997). The most readable serious introduction; excellent on indicator methods and simulation.
• Rossi & Deutsch, Mineral Resource Estimation (Springer, 2014). Mining-specific: domaining, change of support, recoverable resources, classification. The bridge from theory to what a resource geologist actually does.
• Deutsch & Journel, GSLIB: Geostatistical Software Library and User's Guide (OUP, 2nd ed. 1998). Old, and still the clearest specification of the simulation algorithms — the SGS pseudocode in Chapter 8 descends from here.
• Caers, Modeling Uncertainty in the Earth Sciences (Wiley, 2011). Uncertainty quantification and decision-making from the author behind Mineral-X; the intellectual precursor to the Intelligent Prospector line of work.

Data, sampling and compositional statistics

• Gy, Sampling for Analytical Purposes (Wiley, 1998). The theory of sampling error — why a 30 g charge is the dominant variance term for coarse gold, and how to design around it.
• Abzalov, Applied Mining Geology (Springer, 2016). Practical QA/QC: control sample design, insertion rates, control charts, duplicate analysis. The closest thing to a manual for Chapter 6.
• Aitchison, The Statistical Analysis of Compositional Data (1986). Why you log-ratio transform geochemistry before doing anything multivariate.

Organisations and ongoing work

• Stanford Mineral-X — publications. The highest-signal reading queue in the field; check it periodically rather than once.
• Caers on reducing drilling by a factor of five — Canadian Mining Journal. Useful for the framing ("drill to falsify hypotheses, not to fill a grid"); treat the multiplier as an advocate's estimate.
• Mineral-X in Zambia — the Copperbelt University course, May 2025. Context for why the Zambian Copperbelt keeps appearing in this literature.
• KoBold Metals — Mingomba project page. Read it for what the company states officially, and notice what it does not state (Ch. 11).

Software

• sisl/MineralExploration — Intelligent Prospector v1.0 reference implementation.
• SimPEG — open-source geophysical forward modelling and inversion. The practical way to understand Chapter 5's ill-posedness: build a synthetic body, forward-model it, invert, and watch the answer depend on your regulariser.
• GemPy — implicit 3-D geological modelling in Python, with support for stochastic geological models.
• GeoStats.jl — the most complete open geostatistics framework; simulation support well beyond the Python ecosystem.
• awesome-open-geoscience — curated index of data and tools. Software Underground (its community) is also the best place to ask questions.