Scalable subsurface inverse modeling of huge data sets with an application to tracer concentration breakthrough data from magnetic resonance imaging
- Stanford Univ., Stanford, CA (United States)
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Univ. of Texas, Austin, TX (United States)
- Univ. of Illinois, Urbana-Champaign, IL (United States)
When characterizing subsurface properties is crucial for reliable and cost-effective groundwater supply management and contaminant remediation. With recent advances in sensor technology, large volumes of hydro-geophysical and geochemical data can be obtained to achieve high-resolution images of subsurface properties. However, characterization with such a large amount of information requires prohibitive computational costs associated with “big data” processing and numerous large-scale numerical simulations. To tackle such difficulties, the Principal Component Geostatistical Approach (PCGA) has been proposed as a “Jacobian-free” inversion method that requires much smaller forward simulation runs for each iteration than the number of unknown parameters and measurements needed in the traditional inversion methods. PCGA can be conveniently linked to any multi-physics simulation software with independent parallel executions. In our paper, we extend PCGA to handle a large number of measurements (e.g. 106 or more) by constructing a fast preconditioner whose computational cost scales linearly with the data size. For illustration, we characterize the heterogeneous hydraulic conductivity (K) distribution in a laboratory-scale 3-D sand box using about 6 million transient tracer concentration measurements obtained using magnetic resonance imaging. Since each individual observation has little information on the K distribution, the data was compressed by the zero-th temporal moment of breakthrough curves, which is equivalent to the mean travel time under the experimental setting. Moreover, only about 2,000 forward simulations in total were required to obtain the best estimate with corresponding estimation uncertainty, and the estimated K field captured key patterns of the original packing design, showing the efficiency and effectiveness of the proposed method. This article is protected by copyright. All rights reserved.
- Research Organization:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States); Energy Frontier Research Centers (EFRC) (United States). Center for Frontiers of Subsurface Energy Security (CFSES)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA); USDOE Office of Science (SC), Basic Energy Sciences (BES)
- Grant/Contract Number:
- AC04-94AL85000; SC0001114
- OSTI ID:
- 1257809
- Report Number(s):
- SAND2016-1277J; 619422
- Journal Information:
- Water Resources Research, Journal Name: Water Resources Research; ISSN 0043-1397
- Publisher:
- American Geophysical Union (AGU)Copyright Statement
- Country of Publication:
- United States
- Language:
- English
Web of Science
Randomized Truncated SVD Levenberg‐Marquardt Approach to Geothermal Natural State and History Matching
|
journal | March 2018 |
A review of geochemical–mechanical impacts in geological carbon storage reservoirs
|
journal | May 2019 |
Randomized Truncated SVD Levenberg-Marquardt Approach to Geothermal Natural State and History Matching | text | January 2017 |
Similar Records
Large-scale inverse model analyses employing fast randomized data reduction: INVERSE MODELING WITH DATA REDUCTION
A fast algorithm for parabolic PDE-based inverse problems based on Laplace transforms and flexible Krylov solvers