EP 148: Simulation Is Not Estimation - rethinking uncertainty in mining

In this episode of Fresh Thinking by Snowden Optiro, Jamie Oppelaar (Senior Consultant) and Tarrant Elkington (General Manager) discuss a growing challenge in mining: resource models are getting bigger and bigger. More blocks, more attributes, more simulations. While this can improve geological representation, it can also slow down optimisation, scheduling and everyday planning decisions. So how do we strike the right balance between detail and usability? Jamie and Tarrant explore why models are expanding, the practical consequences for engineers, and how geologists and planners can work together to build fit-for-purpose models that support better decisions without unnecessary complexity.  Key moments in this episode 00:00 Introduction: Why resource models keep getting larger 03:40 The three main drivers of model size 07:10 How large models slow optimisation and planning workflows 09:30 When higher resolution models are actually justified 14:20 Practical ways to reduce model size without losing value 20:00 Strategies engineers use to simplify models for optimisation 24:55 The key takeaway: the best model is not the biggest model  If you work with block models, resource estimation, or mine planning, this conversation highlights practical ways to improve workflow efficiency while maintaining the geological detail that matters.  If you would like to contact Tarrant and Jamie: contact(at)snowdenoptiro.com 🎧 Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available as an audio podcast on all major podcast platforms, and in video format on Snowden Optiro's YouTube channel. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of Fresh Thinking by Snowden Optiro, our General Manager Dr Tarrant Elkington speaks with Senior Consultant Dr Gregory (Greg) Zhang about a common question in the mining industry: what is the real industry standard in mineral resource estimation? Drawing on experience reviewing projects across a range of commodities and deposit types, Greg discusses current practices in data management, QAQC, geological domaining, estimation methods such as kriging, and resource reporting. The conversation also explores how practices are evolving and the emerging role of simulation methods and machine learning in resource estimation. Key topics: 00:01 – Introduction and the motivation behind the study 03:30 – Data quality and common database issues 07:20 – QAQC and confidence in data 10:30 – Geological domaining and understanding deposits 16:20 – Estimation methods and industry trends 20:10 – Resource reporting and common pitfalls 25:00 – Future trends in resource estimation If you would like to contact Tarrant and Greg: contact(at)snowdenoptiro.com 🎧 Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available as an audio podcast on all major podcast platforms, and in video format on Snowden Optiro's YouTube channel. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of Fresh Thinking, Susan Havlin (Managing Consultant APAC) speaks with Dr Gregory Zhang (Senior Consultant APAC) about using Excel as a transparent learning environment for geostatistics. Rather than relying on black-box software, they explore how building the logic step by step in a spreadsheet can strengthen intuition, improve data discipline and deepen technical confidence. Whether you are a geologist, resource modeller or mining professional wanting to better understand what sits behind the software, this episode breaks down the thinking that underpins robust estimation. 🔎 In this episode: 00:01 – Why the Excel World Championship is more relevant to geologists than you might think 02:02 – Why Excel is a powerful tool for learning geostatistics 06:27 – Building variograms from scratch: pairs, distances and continuity 09:24 – Understanding kriging as a system of equations 12:03 – Simple validation routines to keep yourself honest 13:02 – Can Excel help you understand machine learning? If you want to move beyond "clicking the kriging button" and truly understand the mechanics behind resource estimation, this conversation is for you. If you would like to contact Susan and Greg: contact(at)snowdenoptiro.com 🎧 Fresh Thinking by Snowden Optiro is available as an audio podcast on all major podcast platforms, and in video format on Snowden Optiro's YouTube channel. If you need further assistance in using Excel, we have 2 courses that might interest you: 1-day Essential Excel Skills for Geologists and our 1-day Advanced Excel Skills for Kriging & Simulation. Both courses are presented by Dr Gregory Zhang. Link to training courses: https://snowdenoptiro.com/professional-development/ Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of Fresh Thinking, Ian Glacken (Executive Consultant) and Dr Gregory Zhang (Senior Consultant) discuss one of the most important topics in modern resource modelling: uncertainty. From the long-standing debate of simulation vs estimation, to the growing focus on geological uncertainty and hierarchical truncated plurigaussian simulation, this episode explores where the mining industry stands — and where it still needs to improve. ⏱ Key Topics 00:00 – Introduction: Simulation in mining and why uncertainty matters 01:18 – Simulation vs estimation: what's the real difference? 02:16 – The "space of uncertainty": grade, geology, parameters and planning 03:34 – Are we misleading stakeholders by ignoring uncertainty? 05:34 – Has grade uncertainty been solved? Two-point vs multipoint simulation 06:39 – Data uncertainty: blast holes, nugget effect and sampling quality 07:30 – Parameter uncertainty and spatial bootstrapping 09:43 – Geological uncertainty: why it's fundamentally different 10:44 – Plurigaussian and hierarchical truncated plurigaussian simulation (HTPGS) 12:51 – Industry adoption and emerging trends 13:08 – The "holy grail" of a comprehensive uncertainty model 14:47 – Digital twins and the future of integrated uncertainty modelling 👍If you enjoyed this discussion, please Subscribe to Fresh Thinking wherever you listen to podcasts, for more mining-focused technical discussions across the mine value chain. If you would like to contact Ian Glacken and Dr Gregory Zhang: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms. It is also available in video format on Snowden Optiro's YouTube channel: https://youtu.be/-DnskmSDnn4 Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/ #Mining #ResourceModelling #Geostatistics #Simulation #Geology #MiningEngineering

In this episode of Fresh Thinking by Snowden Optiro, Susan Havlin (Managing Consultant) speaks with Dr Tarrant Elkington (Snowden Optiro General Manager) about extending generative AI into coding practical engineering solutions. They discuss how mining engineers and scientists can use AI tools to prototype bespoke solutions, automate niche problems, and build simple applications without being expert programmers. If you've ever thought about using AI, but didn't know where to start, this conversation is for you. Key moments: 0:36 – Why generative AI and coding is relevant for mining professionals 2:41 – What "AI coding" actually involves and how it works 4:22 – Real mining examples: road design optimisation and waste dump planning 6:46 – Does it really work first time? Debugging and validation 8:24 – What you learn through the process 9:37 – When it won't work and why starting small matters 11:06 – Practical advice for engineers and scientists wanting to begin Generative AI is changing how we work. The question is not whether to use it, but how to use it well. Listen now and see how you could start building your own tools. If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain. If you would like to contact Tarrant Elkington or Susan Havlin: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on most audio podcast platforms and also as a video podcast on Snowden Optiro's YouTube channel: https://youtu.be/v3dQDfaU-vg 👍 Like, comment, and subscribe for more technical mining insights from our global consulting team. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

Gregory Zhang (Senior Geology Consultant) to talk about what convolution actually means in a geostatistical and geological context, and how it shows up in tools ranging from moving averages to convolutional neural networks (CNNs). Building on part one of the conversation, they focus on practical use, common pitfalls, and how to combine modern machine learning approaches with classic geostatistics without losing geological control. Key moments 00:02 – What convolution means in day-to-day geology and resource work 00:51 – Moving averages, detrending and kriging in practice 01:30 – Using external data like magnetics and structure with kriging 02:09 – Common pitfalls: edge effects, over-smoothing and stationarity 05:14 – Letting geological structure guide smoothing 05:17 – A worked example using folded geology and copper grades 06:11 – Where CNNs fit, and how they compare to traditional geostatistics 07:28 – Tools and software for convolution, kriging and CNN workflows 09:05 – When not to use moving averages or CNNs 10:39 – Practical sanity checks after detrending 11:05 – A clear framework for combining geostatistics and machine learning If you work in resource estimation, geostatistics, or applied geology, this episode is packed with practical insights you can take straight back to your own workflows. 👍 If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain, from out global consulting team. If you would like to contact Jamie or Gregory: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms including Snowden Optiro's YouTube channel. Link to video podcast: https://youtu.be/z5Gz1lYQ0Jg Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of Fresh Thinking, Ian Glacken and Dr Gregory Zhang explore why mineral resource models and grade control models often fail to reconcile and how convolutional neural networks (CNNs) can help bridge that gap. This is Part 1 of a 2 part series. The discussion focuses on early-stage resource estimation, sparse drilling data, and how machine learning can learn from dense grade control information to improve confidence, reduce bias, and support better decision-making. Importantly, this approach is positioned as a supplement to sound geology and geostatistics, not a replacement. Speakers Ian Glacken – Executive Consultant Geology - Snowden Optiro Dr Gregory Zhang – Senior Consultant Geology - Snowden Optiro Key moments in the episode 00:22 Why resource models and grade control often don't match 02:55 Systematic bias and the limits of traditional reconciliation 04:22 Using CNNs to learn relationships between sparse and dense data 06:02 Treating resource models as 3D images for machine learning 07:18 What CNNs can actually predict in a mining context 10:07 Using grade control areas as calibration zones 12:43 Why AI should supplement, not replace, good geology This is Part 1 of a two-part conversation on applying machine learning responsibly in mineral resource estimation. If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain. If you would like to contact Ian or Gregory: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms. 👍 Like, comment, and subscribe for more technical mining insights from our global consulting team. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

Ian Glacken (Executive Consultant geology) and Dr Gregory Zhang (Senior Consultant Geology) explore how machine learning and convolutional neural networks (CNNs) can be used to bridge the gap between grade control data and resource estimation, and why treating resource models as static can hold operations back. Key discussion points: ⏱ 00:00 Introduction and context for using CNNs with grade control data ⏱ 00:55 Why machine learning must be an ongoing, iterative process ⏱ 03:18 Handling multivariate data and complex geological relationships ⏱ 04:00 Practical considerations: data quality, alignment, and validation ⏱ 06:02 Trust, interpretability, and keeping geologists in the loop ⏱ 07:19 Using lithology and categorical data in CNN models ⏱ 08:58 How an operation can get started with these techniques ⏱ 11:58 Final thoughts on machine learning as a decision-support tool If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain. If you would like to contact Ian or Gregory: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms including the video format on the Snowden Optiro YouTube channel: https://www.youtube.com/playlist?list=PLZm0zjSNmpo27fX_tfI79Yzhxy3VXjvMt  👍 Like, comment, and subscribe for more technical mining insights from our global consulting team. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of Fresh Thinking, Senior Consultant Jamie Oppelaar is joined by Dr Gregory Zhang, Senior Consultant at Snowden Optiro, to explore what convolution really means in a geostatistical and mine planning context. Jamie and Greg walk through how it already shows up in everyday resource estimation workflows. From moving averages and detrending through to kriging residuals and convolutional neural networks, this episode focuses on practical application, not theory for theory's sake. This is Part 1 of a two-part discussion, laying the foundations for how convolution links classical geostatistics with modern machine learning approaches. Key topics and timestamps  00:01 – Introduction to convolution - geostatistics and machine learning 00:21 – Separating geological trend from local variability in grade data 02:26 – Detrending, residuals, and the link to universal kriging 03:15 – Moving averages as a defensible trend model using convolution 03:41 – What convolution actually is, explained practically 05:08 – Choosing window sizes using variograms and cross-validation 05:57 – Stationarity, variograms, and kriging residuals 07:19 – How convolution relates to convolutional neural networks (CNNs) 09:22 – A step-by-step workflow mine geologists can apply with existing data 13:19 – Summary of key ideas and what's coming in Part 2 If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain. If you would like to contact Gregory or Jaimie: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms. 👍 Like, comment, and subscribe for more technical mining insights from our global consulting team. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/

In this episode of the Fresh Thinking podcast, Dr Gregory Zhang (Senior Geology Consultant) is joined by Dr Leon Lorenzen (Executive Consultant Metallurgy) to talk about the practical realities of geometallurgy and why linking geology, mineralogy and processing really matters. The conversation draws on real project experience, highlighting the risks of disconnects between resource models and plant performance, especially for complex ore bodies. A thoughtful discussion for geologists, metallurgists and mine planners working with increasingly challenging deposits. Key moments from the episode: 1:45 – Why geometallurgical models are often created but not used in operations 3:50 – How hardness, mineralogy and recovery should feed into daily plant decisions 6:40 – The challenge of limited data and confidence in geomet modelling 9:10 – Why sulphides, preg-robbing carbon and mineral associations matter 16:50 – When a geometallurgical model really is essential, and when it may not be If you enjoyed this episode, please Subscribe for more mining-focused technical discussions across the mine value chain. If you would like to contact Gregory or Leon: [email protected] Listen on the go: Fresh Thinking by Snowden Optiro is rapidly becoming the best mining podcast globally, and is available on all major podcast platforms - including YouTube: https://www.youtube.com/playlist?list=PLZm0zjSNmpo27fX_tfI79Yzhxy3VXjvMt  Like, comment, and subscribe for more technical mining insights from our global consulting team. Snowden Optiro: Snowden Optiro is a resources consulting and advisory group that provides independent advice, consulting and training to mining and exploration companies, their advisors and investors. We help mine developers to advance their projects, mining companies to improve their operations and their professionals, and investors to de-risk their investments by the provision of quality advice, training and software in the field of Mineral Resources and Mineral/Ore Reserves. Explore more: https://snowdenoptiro.com/