Researchers at Heriot-Watt University advocate for physics-integrated AI models to address high-stakes decisions in carbon storage, geothermal energy, and nuclear waste disposal, stressing the need for validated, decision-ready tools.
Machine learning is moving deeper into the geoenergy sector, but a new paper from researchers at Heriot-Watt University argues that the technology will only be useful if it is designed around the realities of the subsurface rather than the convenience of the dataset. Writing from the Institute of GeoEnergy Engineering and the Subsurface Energy Transition and Innovation Centre, Menke and colleagues say the field now spans carbon storage, geothermal systems, subsurface hydrogen, critical minerals and nuclear waste disposal, all of which demand decisions that are both fast and defensible.
The central challenge, they argue, is that these projects are still forced to make high-stakes calls from sparse and uneven evidence. In practice, operators must move from screening to appraisal, operations, monitoring and long-term stewardship with limited field data and strong physical constraints. That creates a need to balance performance against containment, pressure management, induced seismicity, energy and water use, and eventual closure obligations.
According to the paper, four weaknesses repeatedly hold back progress: thin and biased labels, a tendency to treat uncertainty as an optional extra, poor links between scales from pores to basins, and weak quality assurance for regulator-facing use. The authors say these problems are especially acute because the sector is not just trying to predict outcomes, but to produce decisions that can stand up to audit, scrutiny and climate-credibility claims.
Their response is to argue for machine-learning methods that are built to work with physics, not around it. That includes hybrid physics-ML models, probabilistic uncertainty quantification, structure-aware representations and multi-fidelity or continual learning. The paper links these approaches to four practical areas: imaging and process digital twins, multiphase flow and near-well conformance, monitoring and inverse problems including deformation and microseismicity, and basin-scale portfolio management.
The broader institutional context matters too. Heriot-Watt’s Institute of GeoEnergy Engineering presents itself as a major centre for subsurface engineering and applied geoscience, with research spanning carbon capture and storage, geothermal energy and renewable energy storage. The university also says its wider School of Energy, Geoscience, Infrastructure and Society brings together research-led teaching across multiple institutes, underlining the extent to which geoenergy is increasingly being framed as an interdisciplinary systems problem rather than a narrow reservoir issue.
For industrial decarbonisation teams, the message is straightforward. Machine learning may help de-risk geoenergy deployment, but only if it is benchmarked, validated and reported in ways that regulators, investors and operators can trust. Menke and colleagues close by calling for better standards, stronger validation and policy support so that ML tools can be used not just to generate outputs, but to support reproducible, auditable and decision-ready inference in the subsurface.
- https://paulhcleverley.com/2026/05/05/machine-learning-for-sustainable-geoenergy/ – Please view link – unable to able to access data
- https://arxiv.org/pdf/2603.14907v1.pdf – This paper by Menke et al. (2026) from the Institute of GeoEnergy Engineering and Subsurface Energy Transition and Innovation Centre at Heriot-Watt University, Edinburgh, UK, discusses the challenges in geoenergy projects such as CO₂ storage, geothermal energy, subsurface hydrogen generation/storage, critical minerals extraction, and nuclear waste disposal. The authors highlight four main bottlenecks: scarce and biased data, uncertainty treated as an afterthought, weak scale-bridging from pore to basin, and insufficient quality assurance for regulatory compliance. They propose machine learning approaches like hybrid physics-ML, probabilistic uncertainty quantification, structure-aware representations, and multi-fidelity/continual learning to address these issues, with applications in digital twins, multiphase flow, monitoring, and basin-scale portfolio management. The paper concludes with a call for benchmarks, validation, reporting standards, and policy support for reproducible and defensible ML in sustainable geoenergy.
- https://www.hw.ac.uk/uk/schools/energy-geoscience-infrastructure-society/research/ige.htm – The Institute of GeoEnergy Engineering (IGE) at Heriot-Watt University is a leading centre of excellence in geoenergy engineering and applied geoscience. Recognised for the quality of its teaching, training, and research, IGE is ranked 3rd in the UK and 12th in the world for subsurface engineering (2023 QS World Universities Ranking). The institute integrates geoscience and engineering to understand the subsurface for the benefit of the energy industry, pioneering world-class research and training across a broad range of subsurface energy challenges, including oil and gas, carbon capture and storage, geothermal energy, and renewable energy storage, to deliver innovation for the low-carbon transition, sustainable energy, and responsible resource management.
- https://www.hw.ac.uk/uk/schools/energy-geoscience-infrastructure-society/about-us.htm – The School of Energy, Geoscience, Infrastructure and Society (EGIS) at Heriot-Watt University provides a collaborative research and teaching environment organised into five institutes: Institute of Sustainable Built Environment (ISBE), Institute of Life and Earth Sciences (ILES), Institute of GeoEnergy Engineering (IGE), TUI, and ISPHERE, which have been replaced by the Institute for Place, Environment and Society. EGIS offers research-led programmes of undergraduate and postgraduate study, many of which provide professionally-accredited qualifications. The school is highly rated in national surveys, and its graduates are highly sought after by employers worldwide. EGIS prides itself on its international activities, with teaching and research carried out across the globe.
- https://energisimulation.site.hw.ac.uk/about/ – The Energi Simulation Centre at Heriot-Watt University is a focal point for collaborative research, industry partnerships, and academic excellence in exploring sustainable energy solutions. The centre aims to address complex challenges facing the energy sector, from optimising resource utilisation to mitigating environmental impacts. Through state-of-the-art facilities and an interdisciplinary approach, the Energi Simulation Centre contributes to advancing knowledge and expertise in the field of fluid flow simulation, supporting the transition to a low-carbon world.
- https://energisimulation.site.hw.ac.uk/research/paceset/ – PACESET is the PETRONAS Centre of Excellence in Subsurface Engineering and Energy Transition at Heriot-Watt University. The centre focuses on advancing research and innovation in subsurface engineering and energy transition, addressing complex challenges in the energy sector. Through collaborative research and industry partnerships, PACESET aims to optimise resource utilisation and mitigate environmental impacts, contributing to the global transition to a low-carbon economy.
- https://www.hw.ac.uk/study/postgraduate/sustainable-geoenergy – The MSc Sustainable Geoenergy at Heriot-Watt University equips students with the knowledge and skills to manage geoenergy resources in a holistic, safe, and cost-effective manner. The programme addresses key concepts in subsurface energy and CO₂ storage, geothermal energy, and transitional gas, integrating geoscience, engineering, and political and societal aspects into the learning experience. It prepares graduates for a wide range of career opportunities in sectors such as the traditional oil and gas industry, emerging geothermal industry, wider energy sector, government agencies, non-government organisations, and academia.
Noah Fact Check Pro
The draft above was created using the information available at the time the story first
emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed
below. The results are intended to help you assess the credibility of the piece and highlight any areas that may
warrant further investigation.
Freshness check
Score:
8
Notes:
The article references a paper by Menke et al. (2026) from Heriot-Watt University, published in March 2026. The blog post was published on 5 May 2026, indicating timely reporting. However, the blog’s non-commercial nature and lack of editorial oversight may affect the freshness score.
Quotes check
Score:
7
Notes:
The article includes direct quotes from the referenced paper. While the paper is accessible online, the blog post does not provide direct links to the original source, making independent verification of the quotes challenging.
Source reliability
Score:
5
Notes:
The primary source is a peer-reviewed paper from Heriot-Watt University, which is reputable. However, the blog post is from a non-commercial research blog with limited reach and no editorial oversight, raising concerns about source reliability.
Plausibility check
Score:
8
Notes:
The claims about machine learning applications in geoenergy are plausible and align with current research trends. However, the lack of direct links to the original paper and the blog’s non-commercial nature warrant caution.
Overall assessment
Verdict (FAIL, OPEN, PASS): PASS
Confidence (LOW, MEDIUM, HIGH): MEDIUM
Summary:
The article provides a timely summary of a recent academic paper on machine learning applications in geoenergy. While the primary source is reputable, the blog’s non-commercial nature and lack of editorial oversight, along with challenges in verifying quotes and limited independent verification, reduce the overall confidence in the content’s reliability.
