Inverse Modeling & Data Assimilation
I am interested in scalable inversion methods that incoporate multi-physics model and various types of data with different information content.
For example, different types of data with coupled physics models can improve subsurface characterization as below:
 |
 |
(a) True K (left) and tracer test (right); (b) Estimated K and corresponding simulation using
only head data; (c) Estimated K and corresponding simulation using both head and tracer
data
• True log-permeability (K) field (fig. a) was used to generate steady state pressure (o in fig.b)
and transient concentration (+ in fig. c) from tracer injection at (x,y) = (10 m, 10 m).
• Principal Component Geostatistical Approach (PCGA) [Lee and Kitanidis, 2014] was used to estimate K
fields using only pressure data (fig. b) and both pressure and tracer travel-time data (fig.c).
• USGS MODFLOW (flow) and MT3DMS (transport) were linked with PCGA as black box simulators with only a few
hundred runs in total.
Large-Scale and Big Enviromental Data Inversion and Uncertainty Quantification
Sand box characterization using a huge tracer data set obtained from MRI [Lee et al., WRR 2016]
Reactive mineral imaging [Fakreddine et al., AWR 2016]
Saline Aquifer Characterization using Density-driven flow and transport modeling [Kang et al., in review]
Joint inversion using pressure and heat tracer [Lee et al., in review]
Data-worth analysis for reliable prediction of DNAPL fate
Efficient Real-Time Data Assimilation for Near-shore Bathymetry Identification
Nano-scale Carbonate Rock Reconstruction
Nano-scale FIB-SEM image processing and multi-point geostatistics
Software Developments
Below are the list of software that will be released soon..
H-matrices based randomized SVD
PCGA
Scalable geostatistics simulator