Importance of Parameter Uncertainty in the Modeling of Geological Variables

Abstract

Quantitative modeling of geological heterogeneity is critical for resource management and decision-making. However, in the early stages of a mining project, the only data available for modeling the spatial variability of the variables are from a limited number of exploration drill holes. This means that the empirical cumulative distribution function of the data, which is one of the key inputs for the geostatistical simulation, is uncertain, and ignoring this uncertainty may lead to biased resource risk assessments. The parameter uncertainty can be quantified by the multivariate spatial bootstrap procedure and propagated through geostatistical simulation workflows. This methodology is demonstrated in a case study using the data from the former lead and zinc mine at Lisheen, Ireland. The joint modeling of the lead and zinc grades is carried out by using (1) all of the available data, (2) a representative subset (approximately 10% of the available data) without parameter uncertainty, and (3) the same subset with parameter uncertainty. In all cases, the turning bands simulation approach generates realizations of lead and zinc grades. In the third case, the uncertainty in the lead and zinc grade distributions is first quantified (i.e., prior uncertainty) by the correlated bootstrap realizations. This joint prior uncertainty is then updated in simulation by the conditioning data and domain limits, which results in posterior uncertainty. The results indicate that a more realistic resource risk assessment can be achieved when parameter uncertainty is considered.