By Romaric Limacher, Geophysicist at Murphy Geospatial, a Woolpert Company.
We are proud to release the white paper via Wiley “Towards Electrical Resistivity Tomography Interferometry: Leveraging Spatial Redundancy with Machine Learning and Information Theory,” written by Romaric Limacher from our Subsurface Engineering department in Ireland, based in our Kilcullen office.
The white paper will be featured in a special edition of Near-Surface Geophysics, later this year, an international journal for the publication of research and development in geophysics applied to the near surface.
At Murphy Geospatial, a Woolpert Company, our integrated approach to subsurface investigations drives innovation. This newly published paper highlights how our geophysics team, embedded within operates across the full project lifecycle, from field acquisition through to data processing and interpretation. This end-to-end involvement provides a practical, ground-level understanding of the data, enabling us to develop innovative approaches that push the boundaries of near-surface geophysics.
Subsurface investigations . Electrical Resistivity Tomography (ERT) is a well-established method used to image the near surface and support subsurface investigation, but traditional ERT surveys can require extensive data acquisition, making them resource-intensive to deliver.
Murphy Geospatial has investigated whether machine learning can help reduce this acquisition effort by identifying the hidden structure within existing ERT datasets. The aim is simple but significant: to understand whether the use of legacy data can support faster, more efficient surveys without compromising confidence in the results.
Key takeaways
- Machine learning can help identify predictable patterns within Electrical Resistivity Tomography datasets.
- Predictive modelling may reduce the amount of ERT data that needs to be collected in the field.
- Legacy data can support subsurface investigation, monitoring and wider data-led geospatial workflows.
Using Existing Geophysical Survey Data More Effectively
Using five years of ERT surveys collected across Ireland, the study explored whether resistivity measurements at a given depth could be inferred from neighbouring measurements. In practical terms, this means testing whether parts of a dataset can be predicted reliably from the surrounding information.
By applying concepts from information theory, the research showed that ERT data contains significant internal redundancy. Put simply, many measurements are not entirely independent; they follow patterns shaped by the underlying geology. Recognising these patterns creates an opportunity to predict missing or reduced measurements with a high degree of confidence.
The work is based on what is believed to be the first national-scale ERT database created by merging datasets across an entire country. Built from surveys across Ireland, this database provides a distinctive foundation for machine learning and innovation in near-surface geophysics.
Applying Machine Learning to Improve Electrical Resistivity Tomography Acquisitions
To test whether these patterns could be used in practice, a full machine learning workflow was developed. This combined:
- Data preprocessing and dimensionality reduction
- Multiple regression algorithms (including neural networks and ensemble methods)
- Data augmentation techniques to improve robustness
The models were trained to predict resistivity values at a given level using neighbouring measurements. This approach allowed the team to assess how much of the survey information could be inferred accurately, rather than acquired directly in the field.
What the Results Mean for Subsurface Efficiency
The results demonstrated strong predictive capability, with several findings that point towards practical efficiency gains:
- Around 70–80% of resistivity values can be inferred with less than 5% error
- Several machine learning models achieved high accuracy (relative error below 2%)
- Predictions were successfully validated on an independent survey, confirming real-world applicability.
In practical terms, these results suggest that a substantial proportion of ERT survey information may be inferred from surrounding measurements. This creates potential to reduce field effort, improve the handling of incomplete datasets and support more stable subsurface interpretation.
Reducing Field Time in Subsurface Surveys
The research shows that ERT datasets contain structured, predictable information that can be used to:
- Reduce the number of measurements required in the field
- Fill gaps in incomplete or noisy datasets
- Improve the stability of subsurface imaging and inversion
For clients and project teams, the value is clear: faster survey delivery, better use of existing data and the potential to maintain data quality while reducing the amount of information that needs to be collected on site. These benefits can support wider subsurface investigation, monitoring and, where relevant, complementary activities such as utility detection.
A Step Towards ERT Interferometry
The study also introduces the concept of applying ideas inspired by seismic interferometry to resistivity data. Although the underlying physics differs, the principle is relevant: existing measurements may contain enough structure to reveal or reconstruct additional information.
This marks a promising step towards a more data-led approach to ERT, where machine learning can help extract value from survey datasets and open new pathways for innovation in near-surface geophysics.
Using Legacy Geophysical Data to Support Smarter Investigation
While the results are promising, further development will strengthen the approach. Future work should focus on:
- Larger and more diverse datasets
- Application to different geological settings
- Continued integration of machine learning into geophysical workflows
More broadly, this work highlights the opportunity for Murphy Geospatial, a Woolpert company, to make greater use of extensive legacy datasets accumulated over the past 10–15 years. Beyond ERT, similar machine learning techniques could be applied to other geophysical methods used across the business, supporting more efficient workflows, stronger interpretation and continued innovation in subsurface investigation.
To read more about the published paper, click here (subscription required). To discuss how geophysical survey data can support a future project, contact the Murphy Geospatial team.
For more information on our subsurface surveys offering click here.