Researchers utilize 2D electrical resistivity imaging and borehole data to estimate N60-value of soils with k-means clustering. Credit: Institute of Rock and Soil Mechanics, Chinese Academy of Sciences Source Link: https://www.sciencedirect.com/science/article/pii/S1674775525004251

Thailand’s northern regions, characterized by complex geology and active fault systems, experience frequent landslides that threaten both lives and critical infrastructure. In 2022, a slope failure occurred along Highway No. 1088 in Chiang Mai Province, northern Thailand. When a research team led by Professor Shinya Inazumi from Shibaura Institute of Technology conducted geotechnical investigations to determine the collapse’s cause, they encountered a critical limitation.

While 2D electrical resistivity imaging covered the entire slope area, borehole data were available only along certain survey lines, leaving substantial portions of the site uncharacterized. This gap created significant uncertainties in identifying soil layer boundaries, distinguishing different geomaterials, and assessing relative soil stiffness across the failure zone. The team recognized that while drilling provides precise data at specific points, its high cost and time requirements make comprehensive site coverage impractical, particularly in challenging mountainous terrain.

In this regard, advances in machine learning offer new possibilities for pattern recognition in geophysical data.

Now, the team presents a breakthrough methodology that significantly reduces the need for expensive and time-consuming borehole drilling in geotechnical investigations. Their findings were made available online and have been published in the Journal of Rock Mechanics and Geotechnical Engineering on 26 September 2025.

By combining 2D electrical resistivity imaging with machine learning, specifically k-means clustering analysis, this study successfully estimates soil strength parameters (N60-values) across large areas using data from just four strategically placed boreholes. The key innovation lies in developing a robust power equation (R2 = 0.9467) that correlates subsurface electrical resistivity with soil stiffness, validated with high accuracy (MAE = 3.94, RMSE = 5.21).

“Notably, our machine learning algorithm successfully classified subsurface materials into three distinct competency levels: low-competency materials such as loose sand, moderate-competency materials such as medium-stiff clay, and high-competency materials such as stiff-hard clay,” remarks Prof. Inazumi.

The proposed approach dramatically reduces the number of required boreholes while maintaining reliable subsurface characterization, as well as minimizing ground disturbance, surface degradation, and groundwater contamination risks associated with extensive drilling.

Moreover, it provides continuous subsurface information across entire study areas, not just discrete borehole points, enabling comprehensive identification of weak zones for landslide mitigation and slope stability analysis. This integrated geophysical-geotechnical-machine learning framework offers a scalable, practical solution for foundation design, slope stability assessments, and infrastructure planning, particularly valuable in challenging terrains where traditional drilling is difficult or impractical.

Prof. Inazumi highlights the potential real-life applications of their work: “Highway departments can assess soil stability along roadways and identify weak zones prone to slope failure, as demonstrated in this study’s investigation of Highway No. 1088 in northern Thailand following a 2022 landslide. Construction companies can optimize foundation designs by mapping subsurface soil strength variations across building sites, reducing overdesign in strong soil areas while ensuring adequate support in weaker zones.”

Furthermore, this work can enable government agencies to conduct comprehensive landslide hazard mapping in mountainous regions, identifying vulnerable slopes for proactive stabilization measures. It can also aid engineers in assessing seepage pathways and embankment integrity in existing dams without compromising structural stability through excessive drilling, as well as help municipal planners in conducting preliminary site investigations across large development areas cost-effectively, prioritizing detailed drilling only in critical locations identified through resistivity surveys.

Lastly, emergency management agencies can create regional subsurface strength maps for earthquake liquefaction susceptibility assessment and infrastructure resilience planning through this research.

Overall, the development of an integrated framework that can extend limited borehole data across entire investigation areas is expected to make comprehensive subsurface characterization both feasible and affordable.

More information: Ratchadakorn Chumkhiao et al, Using 2D electrical resistivity imaging and borehole data to estimate N60-value of soils with k-means clustering for subsurface geomaterials categorization, Journal of Rock Mechanics and Geotechnical Engineering (2025). DOI: 10.1016/j.jrmge.2025.05.030

Citation: Geophysical-machine learning tool developed for continuous subsurface geomaterials characterization (2025, October 21) retrieved 21 October 2025 from https://phys.org/news/2025-10-geophysical-machine-tool-subsurface-geomaterials.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.