A geologist is equal parts scientist and artist. If geology were solely science, the discipline would require nothing more than analysis of subsurface data at a series of control points. However, as there is no predictive power to understanding the control point only, a geoscientist must, using creativity and craftsmanship, model away from the known to approximate the subsurface across space and time. This is the essence of exploration.

Handcraft geology is a well-known concept amongst professionals within the discipline. Even if the term ‘handcraft’ is not commonly used, a strong handcraft culture exists. A corollary of the handcraft culture is a deep distrust of numerical techniques. For example, a hand-contoured map is always preferred to its numerically derived counterpart. Generally speaking, this culture is good for the discipline: while quantitative techniques cannot improve the fundamental ability of an interpreter, an inappropriately applied method can most certainly degrade an interpretation. I probably don’t need to point out that the most advanced numerical tools have produced a great deal of poor interpretation. However, skillful use of numbers to interpret — what I call computational geology — will not erode the quality of a geologist’s craft and will provide

a number of opportunities not afforded to the handcraft geologist. Before getting into that, I need to explain more clearly what I mean by computational geology.

Computational geology is the deliberate use of numerical models to approximate an interpretation. The foundational interpretation remains the same as for the handcraft geologist: cores are described, stratigraphic tops picked, seismic horizons mapped, core data are analysed, and so on. The difference is the next step: instead of limiting himself to a sparsely distributed primary data source, and using heuristics to force a result, a computational geologist will look for his interpretation in the numbers. The computational geologist then builds a numerical model using data with superior spatial or temporal control that yields a result similar to the expected outcome. Here are a few examples. Rather than relying solely on pore pressure data, the computational geologist instead models a compaction trend from sonic logs or seismic data to characterize pore pressure in a basin. Similarly, rather than using only cored wells to characterize a reservoir, a computational geologist might model facies described in core using a linear algebraic equation of widely available wireline logs.

I have suggested that computational geology will not necessarily produce a superior result to the handcraft method — both techniques are ultimately limited by the interpreter’s understanding of his craft — nevertheless there are a number of important advantages enjoyed by computational geology:

**Iteration and validation.** Handcraft is slow and tedious. A handcraft geologist typically produces a single deterministic model. There is often no time for a second interpretation. This severely limits a geologist’s ability to characterize uncertainty and validate the model. Computational geology is a more efficient workflow where the human spends time on things only a human can do well (interpreting) and the computer takes care of the things the human is relatively inefficient and unskilled at (computation, repetition, precision). This leaves more time for iteration and validation with new data.

**Uncertainty.** Handcraft geology is not repeatable. When a handcraft geologist interpolates, she cannot quantify the likelihood of a contour being at a particular position and not another. Nor can she calculate the probability a facies boundary sits at a particular depth and not a metre lower. There is no rigorous way to quantify uncertainty for handcraft geology. In contrast, uncertainty quantification is intrinsic to all numerical models and is simply calculated as an attribute of the model. And while geologists love to criticize numeric interpolation, the numeric model ensures that all points in space are calculated consistently and, consequently, share a common uncertainty.

**Integration. **Handcraft geology tends to deal poorly with issues of spatial resolution. Detailed core description may not be calibrated to well logs. Core description is not commonly calibrated to seismic. This integration gap makes quantification very difficult. The work of a handcraft geologist is resolutely qualitative in nature and commonly integrates poorly with the quantitative subsurface disciplines — petrophysics, geophysics, basin modelling, and reservoir engineering. Consider that integration of the quantitative with qualitative aspects of subsurface characterization stands to benefit enormously from a numeric methodology, uniting the entire workflow.

Availability of digital data is rarely a limiting factor today. Creativity with respect to using these digits, however, is. Challenge yourself to make use of the advances in technology and make computational geology part of your workflow. You might be surprised at how well your interpretation can be teased out from the numbers.