EGU General Assembly 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Ambient noise tomography of post-subduction setting in northern Borneo enhanced with machine learning

Joseph Fone1, Simone Pilia2, Nicholas Rawlinson1, and Song Hou3
Joseph Fone et al.
  • 1Bullard Laboratories - Department of Earth Science, University of Cambridge, Cambridge, United Kingdom (
  • 2Department of Earth and Environmental Sciences, University of Milan-Bicocca, Milan, Italy
  • 3CGG, United Kingdom

Given that subduction is an important driver of plate tectonics on Earth, it is notable that the effects of subduction termination are often complex and poorly understood. Northern Borneo is a prime example of a post-subduction environment, where two subduction zones have terminated within the last 20 Ma. The region however has seen very few seismic studies likely due to the low levels of seismicity in the region compared to the rest of Southeast Asia and due to the challenging deployment environment. The goal of the northern Borneo Orogeny Seismic Survey (nBOSS) network, which operated between 2018 and 2020 and consisted of 47 broadband instruments, was to provide constraints and answer first order questions about the structure of the lithosphere and asthenosphere in this post-subduction setting. Waveform data from this network were supplemented with data recorded by 33 permanent instruments operated by the Malaysian meteorological authority, METMalaysia. In this study we produce the first model of the crustal shear wave velocity structure under northern Borneo using surface wave ambient noise tomography to try and better understand the effects of subduction termination on the crust and to better understand the present day structure of the crust in this region which has not been imaged in this way before. We use a trans-dimensional tomography to produce variable resolution 2D Rayleigh wave phase velocity maps in the period range 2-30 seconds sampled every 2 seconds. Then to produce the final 3D shear wave velocity model a series of 1D inversions were used in combination with a neural network that is trained to find a generalised solution to the 1D inverse problem for this data set. This helps to prevent artefacts forming in the final model as a result of there being no lateral correlations in the 1D inversions by providing the more region specific trained neural network to perform the bulk of the 1D inversions. The result is a model that shows a detailed 3D shear wave velocity structure of the crust that matches expected velocity anomalies from known geological features. This includes the large sedimentary basins in the region, which are revealed as large slow velocity anomalies. Our new model agrees with results from other methods used to study this region, including receiver functions and surface wave tomography.

How to cite: Fone, J., Pilia, S., Rawlinson, N., and Hou, S.: Ambient noise tomography of post-subduction setting in northern Borneo enhanced with machine learning, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7668,, 2022.