Deep learning of terrain morphology and pattern discovery via network-based representational similarity analysis for deep-sea mineral exploration

Abstract

The race to explore valuable metals in the deep ocean recently emerged, and nations now seek to secure prospective areas for minerals that may support the low-carbon transition, from electric vehicles to wind turbines. Yet, the deep seafloor remains unexplored and vast, which asserts the need for technological advances in exploration. As key areas for new mineral discoveries often reside in vast zones of undersea eruptions, it becomes crucial to examine seafloor processes and spatial patterns to elucidate the nature of the geological phenomena and their complex interactions. Especially, seafloor mounds provide important information about surface changes, sometimes attributable to mineral accumulations at the seafloor. This study applies a 2-step method to investigate these mounds: (1) semantic segmentation with an encoder-decoder convolutional neural network, then (2) morphological similarity analysis and clustering of segmented features by exploiting convolution signals generated by the model with computer vision algorithms and data processing procedures. The study uses high-resolution bathymetric data of a mid-ocean ridge, which includes a known polymetallic mineral occurrence (case study). The model segmented 1,659 features and achieved accuracy up to 84% pixel-wise, and 80% object-wise, using data combination of bathymetry and terrain attributes as input. Clusters reveal morphological patterns that are immediate aftermaths of diverse eruption mechanisms. Eventually, some clusters may be targeted for undiscovered mineral occurrences.

Deep learning of terrain morphology and pattern discovery via network-based representational similarity analysis for deep-sea mineral exploration