Modelling Australia’s Carbon Treasure: A Supercomputing Solution for Climate Action

The Problem: Understanding Hidden Carbon Stores

Australia faces a critical challenge for climate change, sustainability, and biodiversity conservation: understanding and mapping the vast stores of carbon hidden beneath its diverse landscapes. From the eucalyptus forests of Tasmania to the mangrove swamps of Queensland’s coast, soil organic carbon (SOC) represents one of the planet’s largest carbon reservoirs—larger than all terrestrial vegetation and atmospheric carbon combined. However, until recently, scientists lacked comprehensive, high-resolution maps showing where this carbon is stored across Australia’s 7.7 million square kilometres.

This knowledge gap hampers effective carbon accounting, sustainable land management, and conservation planning. Without reliable baseline measurements, Australia cannot effectively track carbon losses from wildfires, land-use changes, or climate impacts. Nor can we identify areas where protecting soils and ecosystems can deliver the greatest co-benefits for carbon storage, productive landscapes, and habitat conservation. The 2019–2020 bushfires, which burned 5.8 million hectares and released vast amounts of stored carbon, highlighted the need to integrate carbon accounting into broader strategies for resilience, sustainability, and biodiversity protection.

Everyday Impact: Why Soil Carbon Matters

While invisible to most Australians, soil carbon directly affects daily life in numerous ways. Healthy carbon-rich soils improve agricultural productivity, supporting the food security that feeds the nation’s 26 million people. These soils also enhance water retention and quality, reducing flood risks and ensuring reliable water supplies for cities and farms. Coastal blue carbon ecosystems—seagrass beds, mangroves, and tidal marshes—protect shoreline communities from storm surges while supporting fisheries that provide seafood to Australian tables.

Beyond these practical benefits, soil carbon strengthens ecological resilience by supporting healthy soils and habitats that underpin biodiversity and represent a powerful natural climate solution. When protected, these carbon stores help Australia meet its international climate commitments under the Paris Agreement. When degraded through poor land management or extreme weather, they become sources of greenhouse gas emissions, accelerating climate change and its impacts on Australian communities

The Supercomputing Solution

Researchers at the Soil & Landscape Science Laboratory, Curtin University and collaborating institutions tackled this massive mapping challenge using advanced machine learning powered by supercomputing resources. The team harmonised data from 6,767 soil sampling sites across Australia—the largest dataset of its kind—representing every ecosystem from arid deserts and tropical rainforests, to mangroves and seagrasses. 

The computational complexity was enormous. The researchers employed various modern methods in a staged workflow. First, they used soil spectroscopy and deep learning to measure the organic carbon in the soils more cost-effectively. Then these spectroscopic measurements were modelled with spatially explicit environmental data (e.g. from remote sensing) using the regression tree method, CUBIST. CUBIST is a sophisticated machine learning algorithm that segments Australia into distinct regions based on climate, vegetation, terrain, and soil properties. To capture carbon variation at multiple scales, they used discrete wavelet transforms to decompose topographic data into different spatial frequencies, a computationally intensive process.

Supercomputers proved essential for several reasons. The deep learning of soil carbon with soil spectra made from 1000s of frequencies required massive parallel processing power. Then, the machine learning model had to process 29 different environmental variables across 90-meter resolution pixels covering the entire continent, with billions of calculations. The researchers ran 30 Monte-Carlo iterations to quantify uncertainty, multiplying the computational load. The Pawsey Supercomputing Centre provided the parallel processing power necessary to complete these calculations in reasonable time frames.

Breakthrough Results

The supercomputing analysis revealed that Australia stores 27.6 gigatonnes of carbon in terrestrial soils and 0.35 gigatonnes in blue carbon ecosystems within the top 30 centimetres alone. The high-resolution maps show dramatic variation, from carbon-poor desert soils holding 7.6 tonnes per hectare to carbon-rich temperate forests storing up to 582 tonnes per hectare.

Importantly, the study identified that while tall eucalyptus forests and coastlines have the highest carbon density, vast grasslands and woodlands hold the largest total carbon stocks due to their extensive coverage. This finding has immediate policy implications: protecting these ecosystems may be more important for climate goals than previously recognised.

Transforming Climate Policy

This supercomputer-generated soil science now provides Australia with unprecedented tools for sustainable land management, biodiversity conservation and climate action. Land managers can identify carbon-rich areas requiring protection, while also identifying degraded areas with the greatest potential for soil restoration, water regulation, and habitat recovery. Policymakers can make evidence-based decisions about where conservation efforts will yield co-benefits, supporting agriculture, safeguarding terrestrial and coastal marine ecosystems, and meeting environmental commitments simultaneously. The research demonstrates how modern supercomputing transforms soil and environmental science, enabling comprehensive analysis of complex systems that would be impossible with conventional computing approaches.

This breakthrough exemplifies the critical role of high-performance computing in tackling 21st-century soil and environmental challenges, helping turn big data into actionable knowledge for climate resilience, sustainable development, and the protection of Australia’s unique natural heritage and our planet’s future.

Authors: Lewis Walden, Oscar Serrano, Mingxi Zhang, Zefang Shen, James Z. Sippo, Lauren T. Bennett, Damien T. Maher, Catherine E. Lovelock, Peter I. Macreadie, Connor Gorham, Anna Lafratta, Paul S. Lavery, Luke Mosley, Gloria M. S. Reithmaier, Jeffrey J. Kelleway, Sabine Dittmann, Fernanda Adame, Carlos M. Duarte, John Barry Gallagher, Pawel Waryszak, Paul Carnell, Sabine Kasel, Nina Hinko-Najera, Rakib Hassan, Madeline Goddard, Alice R. Jones & Raphael A. Viscarra Rossel