Reproduction of a Complex Tracer Test through Explicit Simulation of a Heterogeneous Aquifer using Bayesian Markov Chain Monte Carlo

摘要

Simulation of tracer response curves in waste rock systems can be difficult using conventional numerical techniques, due to the presence of local heterogeneities. Instead, alternative methods are required that can explicitly handle the inhomogeneous nature of flow. Such an approach is presented herein that couples sequential Gaussian simulation with Bayesian inference using a Markov Chain. Results show that complex tracer behaviour can be reproduced using simple, spatially conditioned, Markov Chains. Additionally, external conditioning routines are found to be more efficient in controlling Markov state transition compared to random resampling or collocated cokriging.