A document by Kazuhide Mimura. Click on the document to view its contents.

Deep-sea mud enriched in rare-earth elements (REEs), termed REE-rich mud, is a promising seafloor mineral resource. A decade of surveys has revealed that the mud with the highest REE concentration occurs in the pelagic realm of the western North Pacific Ocean, with two layers of elevated REE concentration. Previous analyses of sediments have revealed multiple periods of significant REE enrichment, with the first (youngest) REE enrichment triggered by global cooling during the Eocene–Oligocene climate transition. However, the depositional mechanism of older REE peaks remains unclear. Fish debris is the major host of REE in deep-sea sediments. In this study, the microfossils of fish teeth and denticles, called ichthyoliths, were observed to constrain the depositional ages of REE-enriched layers with unknown genesis. Empowered by deep learning, more than 40,000 ichthyoliths were observed, and the second (older) REE enrichment was revealed to have occurred in the early Eocene, when the Earth’s climate was exceedingly warm. The warm ocean may have enhanced the efficiency of trophic transfer, leading to an increased supply of fish debris and thus REE, to the seafloor. Therefore, the Paleogene Hothouse might have been advantageous for producing valuable seafloor mineral resources.

Seafloor massive sulfide deposits have attracted attention as a mineral resource, as they contain a wide variety of base, precious, and other valuable critical metals. Previous studies have shown that signatures of hydrothermal activity can be detected by a multi-beam echo sounder (MBES), which would be beneficial for exploring sulfide deposits. Although detecting such signatures from acoustic images is currently performed by skilled humans, automating this process could lead to improved efficiency and cost effectiveness of exploration for the seafloor deposits. Herein, we attempted to establish a method for automated detection of MBES water column anomalies using deep learning models. First, we compared the “Mask R-CNN” and “YOLO-v5” detection model architectures, wherein YOLO-v5 yielded higher F1 scores. We then compared the number of training classes and found that models trained with two classes (signal and noise) exhibited superior performance compared with models trained with only one class (signal). Finally, we examined the number of trainable parameters and obtained the best model performance when the YOLO-v5l model with a large trainable parameters was used in the two-class training process. The best model had a precision of 0.928, a recall of 0.881, and an F1 score of 0.904. Moreover, this model achieved a low false alarm rate (less than 0.7%) and had a high detection speed (20−25 ms per frame), indicating that it can be applied in the field for automatic and real-time exploration of seafloor hydrothermal deposits. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. 2023.2.24: This work was published by IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. https://ieeexplore.ieee.org/document/10052641