Uncertainty in geological interpretations: Effectiveness of expert elicitations

Uncertainty in geological interpretations creates an often unquantified risk for sub-surface industries. The challenge of quantifying interpretation uncertainty has been addressed using various methods. For interpretation of borehole data, empirical quantification of uncertainties can be derived from comparison of interpretations with a withheld set of borehole data not used in the interpretation. This approach requires dense, high-quality borehole data sets. A proposed alternative is to use expert elicitation to extract expert geologists' mental models of uncertainty. We investigated whether expert elicitations are a viable alternative to the direct quantification of uncertainty in three different geological settings by comparing elicited distributions to empirically derived uncertainty distributions. We show that uncertainty distributions derived from expert elicitations are different from those observed in empirical uncertainty quantification. This means that expert elicitations are not as appropriate for estimating uncertainty as these empirical approaches. Expert elicitations, however, offer other benefits to an interpretation workflow, such as providing insight into and challenging different conceptual models of the geology.