Drillhole uncertainty propagation for threedimensional geological modeling using Monte Carlo
Abstract
Monte Carlo Uncertainty Estimation (MCUE) is an emerging heuristic uncertainty propagation method designed to provide reliable and time/cost efficient estimates of geometrical uncertainties in 3D geological modeling. MCUE is a subtype of Bayesian Monte Carlo method similar to geostatistical simulation. The methods described here rely on disturbance probability distributions that are parameterized to best represent individual input uncertainty. Essentially, disturbance distributions quantify the error about the location (x, y, z) and orientation (dip and azimuth) of observed geological structures. The disturbance distributions are sampled either independently or via a MarkovChain to produce many plausible alternative datasets. These plausible datasets are then input to a 3D geological modeling engine to build a series of plausible alternative model realizations. Further processing may be applied to the series of plausible models to provide valuable decision aids such as probabilistic models, reliability models, or uncertainty reduction hotspot maps.
In this paper, a complete and comprehensive MCUE procedure for common drillhole path and log uncertainty propagation is proposed. Basic concepts of drillhole uncertainty are introduced and are applied to a Markov Chain scheme. Appropriate disturbance distributions for the different parts of the problem and their respective parameterization are discussed. The method proposed is demonstrated on three separate proof of concept case studies of increasing complexity. Results demonstrate that the method is able to propagate path and log uncertainty appropriately. First order interpretation indicates that both path and log uncertainty increase with depth and angle of attack to the geological interfaces. Ignoring drillhole uncertainty was found to be detrimental to the understanding of a modeled area which is most likely due to the overconstraining effect brought by "perfect" drillholes. The third case study (Mansfield) hints that uncertainty is better reduced when drillholes intersect the "triple line" that partitions three distinct lithologies. In crosssections, triples lines appear as triple points.
 Publication:

Tectonophysics
 Pub Date:
 November 2018
 DOI:
 10.1016/j.tecto.2018.09.005
 Bibcode:
 2018Tectp.747...16P
 Keywords:

 Uncertainty propagation;
 3D geological modeling;
 Monte Carlo;
 MarkovChain;
 Drillhole;
 Disturbance distribution