Geophysical subsurface modelling and imaging

Project was undertaken in three sub-projects, as follows: 1. Cooperative inversion of seismic and magnetotelluric data: Explore the idea that geological features of interest have potential to be revealed within seismic reflectivity imaging data with unsupervised learning based on seismic textures distribution. Once identified, these features can then feed into strategies for texture guided cooperative inversion of seismic and magnetotelluric (MT) data. 2. 3D inversion of magnetotelluric data: Non-uniqueness of inversion results is a serious problem for interpreting Magnetotelluric (MT) data. Adding known geological and geophysical information in prior model is one way to reach more reliable inversion outcome. However, adding more information in the model will increase the number of parameters to be solved in MT inversion (especially 3D inversion). Due to computational limitations, running the inversion with large number of parameters might take days or weeks. Using Pawsey Supercomputer reduced this running time to a few hours. 3. Molecular Dynamics simulation of H2S under pressurised conditions: The aim is to use predictive methods to gain insights into H2S geo-sequestration and the effects of pressure and temperature on governing parameters. The sequestration of hydrogen sulphide in geological formations offers potential benefits as an economical method of disposal.

Principal investigator

Andrew Squelch
Magnifying glass

Area of science

Geophysics, Geosciences

Systems used


Applications used

ModEM3D, LAMMPS, Python
Partner Institution: Curtin University| Project Code: pawsey0102

The Challenge

Each sub-project addressed a different challenge, as follows:
1. Extracting geological meaning from seismic reflection and electromagnetic (e.g. MT) geophysical exploration data, in particular in hard rock exploration environments, is challenging owing to the absence of consistent coherent or continuous seismic reflectors in these situations compared to the better represented oil and gas situations.
2. Understanding the effect of large-scale geo-electrical structure on MT inversion outcomes can reduce inaccuracy of the inversion results. The main challenge to include large-scale features in the model is computational limitations to run the ModEM3D code and also running several inversions at limited time.
3. The process of is H2S geo-sequestration governed by various parameters, including the water–H2S interfacial tension, H2S absorption into water and H2S adsorption onto water. However, the influences of pressure on these parameters are poorly understood. This project investigates these parameters as functions of pressure and temperature using molecular dynamics (MD) simulations

The Solution

Each sub-project employed a different solution, as follows:
1. The concept of “textural domaining” for 3D seismic reflectivity data provides significant steps towards mapping volumetric distributions of seismic reflectivity linked with macroscopic geological environments and processes. Dip-steered seismic texture attributes are combined with unsupervised learning to generate sets of volume rendered images accompanied by a seismic texture reference diagram. These methods have the potential to reveal geological and geotechnical properties that would otherwise remain hidden.
2. ModEM3D code includes an adjustable coarse-grained parallelization of forward problems and it intended to minimize communication between processors. By running parallel inversion on Magnus, we are able to run many inversions in shorter time.
3. Using these predictive methods, we were able to establish some parameters that could be useful in the big picture towards learning more about the geo-sequestration process. Molecular dynamics (MD) simulations were carried out for the isotherms 40, 70 and 120 oC and a pressure range of 0.45 to 15 MPa. A comparison was also made against N2-water systems to clarify the effects of adsorption on the interfacial tension parameter

The Outcome

Each sub-project achieved a different outcome, as follows:
1. Seismic texture guided cooperative inversion of the MT data was run on Magnus. In this process step, each seismic texture domain is statistically assigned a conductivity, this conductivity distribution feeds a final 3D MT inversion, which is run on Magnus with the ModEM3D code.
2. The forward modelling results obtained with ModEM3D run on Magnus show a TE and TM mode separation in low frequencies which may cause artefact in inversion outcome if it is not considered. Including large-scale geo-electrical structures in inversion improved subsurface resistivity image.
3. By utilising computing core hours on Magnus, we were able to run simulations using the installed LAMMPS molecular dynamics engine to answer some pertinent research questions about H2S geo-sequestration. These included:
• predicted water-H2S tension and phase densities below H2S saturation pressure matched experimental values well;
• adsorption can be quantified via the interfacial thickness, which correlated well with the H2S pressures;
• above the H2S saturation pressure, the interfacial thickness remained constant, which indicated a saturation of H2S at the water surface; and
• further increment of pressure above saturation revealed a significant absorption of H2S.

Note: Prior approval must be obtained from the relevant researchers and authors before supplied images can be re-published in any form.

List of Publications

1. Le, C.V.A., Harris, B.D. & Pethick, A.M. New perspectives on Solid Earth Geology from Seismic Texture to Cooperative Inversion. Sci Rep 9, 14737 (2019).
2. In process.
3. In process.
Note: Prior approval must be obtained from the relevant researchers and authors before supplied images can be re-published in any form

Figure 1. Volume rendered images of textural domains and conductivity distribution from seismic texture guided cooperative inversion of co-located seismic and MT data from Nevada, USA. The Figure contains (A) a 3D representation of seismic texture clusters, (B) output geo-electrical distribution from unconstrained MT inversion S1 and (C) output geo-electrical distribution from seismic texture guided cooperative inversion. Note: Prior approval must be obtained from the relevant researchers and authors before supplied images can be re-published in any form.
Figure 2. Volume rendered images showing high conductivity channel geometries resolved by seismic texture guided cooperative inversion strategy C6. The channels exhibit relatively high GCLM homogeneity and low entropy. The seismic textural domaining has provided a shallow detailed framework that is integrated into the texture guided cooperative inversion. The high conductivity channels or depressions that appear to be incised in basement become well defined by texture guided cooperative inversion strategy C6. Note: Prior approval must be obtained from the relevant researchers and authors before supplied images can be re-published in any form