Detachment of the Subducting Slab in Ancient Subduction Zones Discerned Using Machine Learning

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 4
• Main Authors: Ou, Quan, Zheng, Dongyu, Nichols, Alexander R. L., Wang, Kun, Qian, Sheng‐Ping, Zi, Feng, Ma, Anlin, Zeng, Yun‐Chuan, Huang, Zongying, Liao, Jia, Hou, Zixi
• Format: Journal Article
• Language: English
• Published: Wiley 01.12.2024
• Subjects: Bangong‐Nujiang ocean; basalts; machine learning; simplification of basalt types; slab detachment; trace elements
• DOI: 10.1029/2024JH000400
• ISSN: 2993-5210

Identification of complex geochemical signatures of basalts from various tectonic settings using trace elements contents is still challenging due to uncertainties in existing classification diagrams and machine‐learning attempts. To address this, we trained a machine‐learning model using a random forest classifier on trace element concentrations of a global basalt‐dataset, categorized by types of tectonic plate boundary—destructive or constructive—at which they formed. Achieving an accuracy exceeding 98%, the model efficiently extracts the distinctive characteristics from each basalt type. When applied to the Bangong‐Nujiang suture zone in central Tibet, the model reveals that the basalts exhibited features of both boundary types prior to 108–107 Ma before transitioning to solely destructive characteristics. This transition is likely to be caused by the detachment of a descending oceanic slab, aligning with existing geological evidence. This case study highlights the promising potential of machine learning models, trained on simplified basalt types, in more accurately tracing lithospheric evolution. Plain Language Summary Determining where volcanic rocks erupted using just their composition is difficult. The current methods, including diagrams and computer models, often provide ambiguous results. To improve this, we created a new computer model that uses a technique called a random forest classifier. We fed the trace element compositions of basalts from across the world into a computer, along with whether the rocks formed at a destructive (e.g., western Pacific margin) or constructive (e.g., mid‐ocean ridge) plate boundary. Our new model correctly identified the geochemical signatures of basalts from the two types over 98% of the time. We then tested the model using basalts from the Bangong‐Nujiang suture zone in central Tibet, which has a complex tectonic history with evidence for the existence of both destructive and constructive plate boundaries. Our model shows that up until about 108–107 million years ago, basalts in this region had features of both types of plate boundary, after which their characteristics became solely destructive. This shift could be due to a subducted oceanic plate detaching and sinking in the mantle, which matches other geological evidence. This shows how powerful our new model can be in understanding how Earth's lithosphere evolves. Key Points A global trace element data set of basalts from destructive and constructive plate boundaries was used to train the machine learning model The trained models clearly differentiate basalts from the two boundaries with an accuracy higher than 98% Applied to Bangong‐Nujiang suture zone, a simplification of basalt types 108–107 million years ago, marking subducting slab's detachement