New method for estimating subsurface flow systems
An adaptive-tempered sequential Monte Carlo procedure estimates geological properties and flow.
A new adaptive-tempered sequential Monte Carlo (SMC) procedure was developed. It performs joint Bayesian estimation for subsurface flow systems with hierarchical uncertainty.
Hierarchical geological uncertainty involves uncertain metaparameters and uncertain cell-by-cell properties. The method uses noncentered geological parameterizations. These represent geomodel realizations with metaparameters and Gaussian latent variables.
The procedure applies adaptive likelihood tempering, importance weighting, resampling, and blockwise Markov chain Monte Carlo (MCMC) mutation. Flow surrogates are used for all function evaluations.
The method was tested on a faulted carbon storage system with three stacked aquifers and two extensive faults. It was also tested on a channelized oil-water system with three facies.
For four synthetic true models, SMC posterior metaparameter distributions and flow predictions agree closely with reference MCMC results. SMC requires fewer function evaluations.
SMC provides speedups of 2.5--7.0 for single-stage data assimilation. Even greater speedups are achieved with multiple GPUs.
Key facts
- SMC provides speedups of 2.5--7.0 for single-stage data assimilation.
- The method uses flow surrogates for all function evaluations.
- The procedure was tested on a faulted carbon storage system with three stacked aquifers and two extensive faults.