February 2017

Deep earth imaging – a CSIRO Future Science Platform

  • By Professor Michael McWilliams, Deep Earth Imaging Platform Leader, CSIRO

A new field of inquiry has the potential to increase understanding of subsurface rock properties, helping to unlock the resource potential of Australia

Australia’s future minerals, energy and water resources will come from far greater depths in the onshore regions and from deep offshore plays. Our ability to find, define and exploit these resources is limited by the deep and complex cover of sediments and weathered material that covers approximately 80 per cent of Australia’s land mass.

Deep earth imaging science will help us more precisely image and understand the significance of subsurface rock properties, which in turn will unlock the resource potential of our vast and relatively underexplored continent, transforming it into a ‘Glass Earth’.

Image by Carl-st205. Used under CC BY-SA 4.0.

Future Science Platforms

Six new CSIRO Future Science Platforms, which began in 2016, will underpin innovation in natural resources, health and biology, agriculture and manufacturing. Each platform will support the reinvention and creation of new industries, wealth and jobs for Australia.

Future Science Platforms fuel deeper collaboration across disciplines to tackle some of the things that haven’t been done before and will help Australia stay ahead of accelerating global disruption.

The platforms are a strategic investment in frontier research and the delivery of solutions. Some will draw on big data to make strides forward for health and environment, some will use CSIRO’s precision science to transform biological systems, while others will focus on our deep knowledge of natural resources and manufacturing to create more sustainable industries to support the jobs of tomorrow.

CSIRO will make an initial $17 million investment in 2016-17 in six new areas of frontier science, and then grow that investment to more than $52 million per year by 2020, helping to turn Australia’s challenges into opportunities and invent a better future. The platforms will facilitate an increase in the amount of staff working on frontier science as well as increased collaboration between universities and CSIRO.

The teams working on these platforms will have the chance to see their ideas fast-tracked through technology preaccelerators that will create connections between research, science and business, enabling research teams to validate their research outcomes and translate them into real-world applications.


The Deep Earth Imaging Future Science Platform

For more than 120 years, we have successfully mapped and analysed Australia’s land surface, but our resource-rich continent has yielded most of its wealth from an area comprising only about 20 per cent of the continent’s exposed or shallow crust.

Geologically, there is no reason to think that the relatively underexplored area is not at least as well-endowed in natural resources. The principal challenge is that this likely endowment is concealed by a complex cover of sediments and weathered regolith. This vast area presents an enormous opportunity to discover buried natural resources: minerals, energy and water.

Join minerals professionals on LinkedInLI_logo

Stay up-to-date with all the latest news and analysis by joining minerals industry professionals in the AusIMM LinkedIn community.

A decline in mineral exploration success across Australia is in large part due to the difficulty of exploring beneath and within the regolith and sedimentary basins that cover most of the continent. Future success requires diversification of the search and discovery space through this cover and to greater depths. Industry has identified and prioritised the development of a new suite of tools and workflows to explore through Australia’s cover as articulated in the UNCOVER vision and the AMIRA Roadmap for Exploration Under Cover.

Similarly, for energy resources, there is considerable potential for new oil provinces to be discovered and developed in frontier areas such as the Great Australian Bight. Onshore basins, particularly in the Northern Territory and central/western regions of Australia, have enormous unconventional natural gas potential. The recent COP21 meeting in Paris highlighted Australia’s leading role in geosequestration to explore and develop new subsurface carbon storage reservoirs.

Building hydrogeological understanding from deep earth imaging will also build public, government and investor confidence that resource development impacts on water resources can be minimised. Nationwide detailed water resource assessments are vital and can be supported by innovation in deep aquifer imaging and characterisation to help provide an integrated evaluation of the feasibility, economic viability and sustainability of Australia’s water resource development opportunities and potential.

Deep earth imaging will use smart analytics and algorithms to simulate geological models and properties that enable subtle patterns to be identified and interpreted, thus more precisely imaging subsurface rock properties.

We will build new analytical software tools founded in digital rock physics, drawing on predictive technology, machine learning, geological uncertainty analysis and geoscience modelling. These tools will manage real-time or near real-time data streams and assemble multiple inputs from geology, (hydro)geochemistry and geophysics.

Our work will draw on expertise from several overlapping domains, including geophysical modelling and simulation, geological/geophysical/geochemical knowledge integration and geological uncertainty reduction, with its impact delivered through a cloud-based platform to provide new tools as a service to industry and the nation.

The Pawsey Supercomputing Centre, part of Perth’s National Resource Sciences Precinct, where the CSIRO Deep Earth Imaging Platform and its staff are based. Image courtesy Pawsey Supercomputing Centre.

Theme 1 – modelling, inversion and simulation

We will build a geophysical modelling software platform using forward simulations fused via Bayesian inversion, an approach that has the dual advantage of robustly computing uncertainty in the final output and allowing joint inversion of data from disparate sensor platforms. For example, magnetic and electromagnetic responses of contrasting rock units and effects such as remanence and superparamagnetism need careful consideration given our knowledge of geological relationships.

We will devise consistent methods for combining this information. Industry clients and researchers will use available data to build better models to quantify the risk associated with acting on predictions. Three critical avenues of research and development are required to make this toolbox a reality:

  • a parameterised geological model to define possible inversion outputs
  • estimates of process noise in each of the various forward models to properly weight its contribution to the final output
  • inversions computed economically via reduced-order modelling and empirical interpolation for smart sampling.

Theme 2 – knowledge integration

We will pursue non-parametric approaches to directly infer cover characteristics and depth from an understanding of a geophysical, geochemical, surface geology and satellite data suite. Rather than employing geophysical and geological simulations to compute expected sensor measurements from candidate models, we will train probabilistic machine learning algorithms with example targets and features derived from available raw data sets. We will simulate geochemical processes and forward model geophysical responses. Coupled fluid flow and chemical reaction (reactive transport) models can be employed to simulate ore formation and deposition processes and produce a mineralogical model at various time steps showing how chemical alteration varies in space and time.

Theme 3 – uncertainty reduction

Blending probabilistic approaches with model reduction, data compression and empirical interpolation in forecasting will allow us to develop workflows and plan future observations to maximise expected information gain. Thus, rather than drilling or sensing in a grid, adaptive algorithms could optimally sample selective regions where the model is most uncertain, increasing the value of each measurement while reducing the overall cost of sample collection. This active sampling tool, in conjunction with parametric and non-parametric models, will determine the location and sensing modality of future measurements to maximise value. Greater recognition of all the contributing sources of uncertainty is required in future studies, whether hydrogeological or geological, particularly as we work across scales of measurement. Additional design constraints (eg distance from roads/infrastructure/water resources or total cost per kilometre) will be added to the cost function. The key challenge in building an active sampling tool is to maintain computational tractability, especially when integrated with Bayesian inversions, justifying an emphasis on model reduction and
data compression.

Theme 4 – large-scale computation

The tools we will build are computationally expensive to deploy. In an era of cloud computing, we do not expect our industrial customers to build and maintain new cyber-infrastructure. Rather, we will build a cloud-based platform to provide these decision-making tools to industry, seeking to lower the barrier to entry, thereby accelerating technology uptake.

Figure 2. TOP100 supercomputer Magnus at the Pawsey Supercomputing Centre, the most powerful in the southern hemisphere. Image courtesy Pawsey Supercomputing Centre.


Our ultimate aim is for the overlying regolith to become ‘transparent’ through more precise imaging and profiling of subsurface rock properties, transforming the underexplored parts of Australia into a ‘Glass Earth’. Developing a toolkit to better navigate through the subsurface will add significant value to the wealth of geological and geophysical data that already exist. The full potential of these data remains untapped, mainly because there is a lack of appropriate interpretation technologies. By linking these methods to digital rock physics and by discovering, understanding and modelling relationships between remotely acquired geophysical observables and in situ rock properties, the uncertainty associated with exploration through this cover can be reduced. This means that further drilling and collection of geophysical data will be more efficient and achieved with lower risk. In addition, outputs from the development of the platform will deliver generic software products that will boost Australia’s world-leading expertise in this domain.

The Deep Earth Imaging Future Science Platform will:

  • create a research hub of expertise in deep earth imaging
  • develop a product suite for data interpretation and decision-making to solve deep cover challenges in
    resource exploration across water, minerals and energy
  • deliver a range of commercialisation opportunities for small and medium-sized enterprise partners to take up globally
  • build a cloud-based platform to support rapid transfer and utilisation of the developed technology
  • establish a CSIRO-led geoscience network and a deep earth imaging annual conference.

For the team, the platform will be a nucleus for collaboration, the first ever exploration and environmental geoscience capability hub in deep earth imaging that crosses the boundaries of minerals, energy and water. CSIRO will play a central role in serving as the innovation catalyst for the earth science community in Australia.

Continuous innovation and technological development are critical for improving the global competitiveness and productivity of the Australian resources sector and, as such, are critical to the nation. Australia has a clear competitive advantage in its ability to conduct such research, with the mining sector accounting for 22 per cent of all business research and development (R&D) investment. However, one of the major challenges is the commercialisation of these research outcomes. The platform will connect with METS Ignited and the Oil, Gas and Energy Resources Growth Centre, which were both built to promote innovation by accelerating commercialisation and encouraging better collaboration between R&D and industry needs. Early collaboration will provide for rapid technology transfer mechanisms for our product suite into both sectors.

Likewise, there is a need for better understanding of deep aquifers in remote regions that could be vital for supporting natural environments and developing a sustainable agricultural and industrial base. Meeting the demand for new water users and uses is an important driver for improved information and analysis of deep water resources that can be addressed through deep earth imaging.

Share This Article