OZ Minerals uses open data and a $1 million incentive to unearth hidden treasure

deCODES
The three month long journey to data-driven discovery on the Explorer Challenge concluded on Friday 31 May 2019, with over one thousand global participants from sixty-two countries digging through more than six terabytes of public and private exploration data from OZ Minerals’ Mount Woods tenement in northern South Australia.

In addition to a A$1 million prize pool, the winning models on the Explorer Challenge will be tested in real life, with the top targets scheduled to be drilled by the end of 2019 with the prospect of unearthing the next big Australian mineral deposit.

Economic mineral deposits have become increasingly difficult to find. Explorers seek new approaches to solve this problem and develop innovative processes and ways of working that can drive up the discovery rate, and in doing so, speed up the exploration lifecycle resulting in a more sustainable and efficient future for mineral exploration.

Modern mining company OZ Minerals partnered with energy and resources open innovation platform Unearthed to deliver this unique, online crowdsourcing competition, which involved the challenge and accompanying data being made available digitally to geologists, geoscientists and data scientists from around the world, who then competed to deliver the best solution.

Congratulations to the Explorer Challenge winners:
Team Guru

1st Prize (A$500,000): Team Guru - Michael Rodda, Jesse Ober, and Glen Willis.

Background: Michael was the data scientist for the team, Jesse has a background in environmental science and GIS, and Glen has a background in process engineering in the field of oil and gas and has a keen interest in data science.

Approach: Interpretable machine learning models for mineral exploration using geochemistry, geophysics and surface geology.

Winning first prize gives our team the confidence in our data science abilities and cements our passion to shake up mineral exploration. We definitely plan to keep chipping away at this problem - we’ve barely scratched the surface of the value we could potentially get out of the data."

2nd Prize (A$200,000): DeepSightX - Dong Gong, Javen Qinfeng Shi, Zifeng Wu, Hao Zhang, Ehsan Abbasnejad, Lingqiao Liu, Anton van den Hengel, Karl Hornlund, and John Alexander Anderson.

DeepSightX

Background: The DeepSightX team exploited multi-disciplinary skills at the intersection of artificial intelligence and geoscience. Researchers from the Australian Institute for Machine Learning (AIML) and the Institute of Minerals and Energy Resources (IMER) - both hosted by the University of Adelaide – collaborated with industry experts in minerals exploration (Austrike Resources) and geoscientific modelling (Gondwana Geoscience).

Approach: DeepSightX used a multidisciplinary approach to generate an AI model, DeepSight, which provides promising exploration targets in the Prominent Hill Region (PHR) supported by best practice geoscience.

The next step is to seek supports from all levels of the university, build capability, and start commercialisation process. We have a few ideas. We will consider all options and develop a holistic strategic plan for the future. The competition is a prelude for DeepSightX, and we look forward to the exciting journey ahead.
Team Cyency

3rd Prize (A$100,000): Cyency - Hugh Sanderson, Derek Carter, and Chris Green.

Background: Cyency has a strong data science and geoscience background. Hugh has been practising deep learning for several years, Derek has been involved with the technical and software side of mining for over 10 years, and Chris is an experienced geologist who is always looking for innovative ways of doing things.

Approach: "With so much data, it was difficult to know where to start, so we started with what we knew - the results from the Data Science Stream. We had a set of models that we knew were pretty good at predicting mineralisation across Australia, so we ran them over the tenement...we applied several data science techniques to estimate a set of candidate points, and then selected the 10 best of these."

Prize money will be put straight towards developing a business in the mining/geology/artificial intelligence/machine learning space. There are so many interesting problems and the technology has opened up a greenfield, and Australia looks well placed to take advantage of it."

Student Team Prize (A$50,000): deCODES - Christopher Leslie, Matthew Cracknell, Angela Escolme, Shawn Hood, and Ayesha Ahmed.

deCODES

Background: deCODES is a team of early career researchers from CODES, University of Tasmania, including three PhD students, and two postdoctoral research fellows who gained their PhD degrees at CODES. The team represents a varied skillset, spanning economic geology and mineral exploration, geophysics, data science, geochemistry and geometallurgy.

Approach: Our approach was driven by considering an Iron Oxide Copper Gold (IOCG) metallogenic model, and then striving to produce digital proxies for all aspects of that model. Our prospectivity layers were created using a mix of manual and traditional data handling methods as well as basic machine learning approaches.

We are hungry to tackle exploration problems like this full time. We look forward to refining and applying components of the workflow we have developed as part of the Explorer Challenge in our respective resource exploration endeavours.
Team OreFox

Genius Prize (A$25,000): Team OreFox - Warwick Anderson, Sheree Burdinat, Kudzai Dube, Amy Leask, Alan Ryou Pearse, Ashleigh Smyth, and Nick Josephs.

Background: OreFox is the brainchild of two exploration geologists, Warwick Anderson and Sheree Burdinat. They built up a team of experts with backgrounds in geophysics, data science, statistics, geology and prospecting to tackle the Explorer Challenge.

Approach: OreFox used its proprietary artificial intelligence systems to analyse the data supplied by OZ Minerals as well as open source data obtained through Geoscience Australia and the SARIG database. They compiled data stacks of geochemistry, gravity, magnetics and radiometrics, then allowed their Prospector AI and Hunter AI systems to evaluate the most prospective regions within the OZ Minerals tenement areas. They also utilised our Target AI to eliminate potential targets further and refine their suggestions.

Winning this prize proves that out-of-the-box methods, while scary and unknown, can actually produce incredible results. OreFox is the commercialisation of research undertaken by our METS company Quantum Geology. With four paying customers now on board we are just getting started in our mining revolution. Future plans are to expand into Western Australia and then take the company global."
Avant Data Solutions

Insights Prize (A$25,000): Avant Data Solutions - Zhen Wang, Andrew Vinciguerra, and Christiaan.

Background: Avant's multidisciplinary team consists of Andrew and Zhen on data science and programming and Christiaan on geological domain expertise.

Approach: The team took a heavily data driven approach with verification and interpretation using geology. The challenge was tackled by first analysing and exploring the data in detail and finding what data might be overlooked. Geophysical spatial data and the geochemistry of the region and drill holes were considered in detail in their methods.

Winning the Insights prize is great validation of our approach and will allow us to further develop our solution for commercial use. We plan to continue developing our unique solution, and invest in additional research and development, as well as offering our expertise in data science and machine learning to mining companies."
Team Phar Lap

Data Hound Prize (A$25,000): Team Phar Lap - Matthias Helbig, Dr. René Kahnt, Rohan Lloyd, Dr. Holger Eichstaedt, Ruky Siu, and Bill Marr.

Background: Team Phar Lap consists of a mathematician, a physicist, a German trained geologist and ecologist, a pilot, and a US trained geologist, offering a latticework of geosciences and data science.

Approach: The consortium decided on a mixed approach between geological interpretation and data crunching with a strong focus on controlled learning. Additionally, they decided to collect high resolution hyperspectral surface expression data, which was used for a re-interpretation of the geologic structures and machine learning process.

Winning the Data Hound prize means a great deal to Team Phar Lap as it's really the first time combining surface expression and geophysics data using complex machine learning models. While most of our clients are struggling with bridging the gap between these two worlds, the Explorer Challenge gave us a chance to develop and verify methods with rich data from both worlds. Winning this award gives credence to our methods and provides a proof of concept to the mining exploration industry”.

Fusion Prize (A$25,000): SRK Consulting - Mark Rieuwers, Ben Jupp, Bert De Waele, Antoine Caté, Ali Shaban, Ron Uken, Erwann Lebrun, and Michael Cunningham.

SRK Consulting

Background: Team SRK consists of qualified structural geologists across SRK Consulting offices in Perth, Melbourne, Toronto and Vancouver – one of these (Dr Antoine Caté, SRK Toronto) having branched into the development of applications of Machine Learning for the mining exploration sector as a Postdoc at the Université du Québec.

Approach: Their approach included the re-interpretation and/or value-add of the provided and available datasets followed by a multi-pronged and integrated targeting approach applying data-driven machine learning (based on a balanced random forest algorithm) and weights of evidence to guide a set of knowledge-driven mineral systems informed fuzzy inference solutions. This resulted in three highly-ranked IOCG targets and seven secondary targets.

Winning the Fusion Prize is validation of a great collaborative effort by the global SRK team that we believe was rooted in solid geoscience, combined with cutting-edge machine learning and more traditional data science techniques. We have already had enquiries from interested clients and will be presenting some of our work from the Explorer Challenge at the Australasian Exploration Geoscience Conference (AEGC 2019) in Perth this September.

The submissions to the Explorer Challenge displayed an amazing range of analytical and geological approaches. From cutting edge machine learning to advanced physical modelling, the submissions represented thousands of hours of work developing and applying robust techniques applicable to the problem of target generation. Bringing data scientists and geologists together for this challenge has resulted in novel ways to apply modern data science techniques, such as machine learning, to geological problems in a meaningful and explainable way.

OZ Minerals Chief Executive Officer, Andrew Cole, said that despite exploring extensively around Prominent Hill over the past decade, they still hadn’t hit a deposit.

We needed to find fresh insights and ways of working smarter with our data. The innovators who participated in the Explorer Challenge have provided approaches to mineral exploration that we never would have imagined internally, including ways to fuse datasets together, combining multiple layers of information, and making predictions based on the extensive datasets," he said.

Reviewing the diverse range of solutions that have come back from this process has been truly remarkable."

Unearthed Industry Lead – Crowdsourcing, Holly Bridgwater previously worked for a decade as a geologist in resource exploration and definition. She believes that crowdsourcing will transform the lengthy and intensive exploration process.

We are extremely excited by the incredible range of solutions submitted by these pioneers that can generate high quality exploration targets in an efficient way” said Bridgwater.

Many industry professionals and mining companies are beginning to realise that their true competitive advantage in exploration is speed, not necessarily data or technological intellectual property. I think that the ability that the crowd gives you to generate new ideas, develop solutions, and automate processes, is something that can make a big difference and provide that competitive advantage,” she said.

Crowdsourcing open data in this way provides a unique opportunity to combine hundreds of models and targets from independent experts into one exploration model, boosting confidence in the targets generated. This previously unseen model to combine diverse thought into consensus, could significantly increase discovery rates.

---

Thank you to Explorer Challenge competition partners: Government of South Australia, Department for Energy and Mining for access to their open file data platform SARIG (South Australian Resource Information Gateway) that served as an important resource available to contenders, Esri Australia for providing complimentary licenses to their leading GIS mapping software – ArcGIS – to support the entrants in their explorations throughout Australia, and Amazon Web Services for their complimentary data storage and computation support.

Congratulations to all next-gen pioneers who took part in the Explorer Challenge. It was a phenomenal effort from all teams on what was a very difficult challenge. Unearthed hopes to be bringing another exploration challenge to the crowd soon. Watch this space for full profiles on the Explorer Challenge winning teams and updates on results of the OZ Minerals drilling program of the top targets.