Tracking the spatiotemporal variations of statistically independent components involving enrichment of rare-earth elements in deep-sea sediments

Abstract

Deep-sea sediments have attracted much attention as a promising resource for rare-earth elements and yttrium (REY). In this study, we show statistically independent components characterising REY-enrichment in the abyssal ocean that are decoded by Independent Component Analysis of a multi-elemental dataset of 3,968 bulk sediment samples from 101 sites in the Pacific and Indian oceans. This study for the first time reconstructs the spatiotemporal variations of the geochemical signatures, including hydrothermal, hydrogenous, and biogenic calcium phosphate components that were closely involved in the formation of REY-rich mud over the past 65 million years. An underlying key factor of significant REY-enrichment is a sufficiently low sedimentation rate that enables the mud to accumulate REY from seawater. In the early Cenozoic, a remarkably small supply of aeolian dust, compared with any other time and region, facilitated the deposition of very high-grade REY-rich mud in the South Pacific. This indicates an important link between the genesis of the seafloor mineral resources and Earth’s dynamic phenomena such as climate change and plate tectonics.

Statistic and Isotopic Characterization of Deep‐Sea Sediments in the Western North Pacific Ocean: Implications for Genesis of the Sediment Extremely Enriched in Rare Earth Elements

“…The computational protocol for performing ICA is specified in previous studies (Iwamori et al, 2010;Iwamori & Albarède, 2008;Yasukawa et al, 2016), and this study follows the described procedure. First, the observed data matrix X is whitened, that is, centered according to the mean of each variable, then uncorrelated using an ordinary PCA algorithm and scaled by the standard deviations along the PCs.…”

“…Geochemistry, Geophysics, Geosystems (Plank & Langmuir, 1998;Yasukawa et al, 2015Yasukawa et al, , 2016. Therefore, IC1 appears to reflect, at least partially, the variation in sedimentation rates.…”

“…Because the Minamitorishima area is far from active regions discharging hydrothermal fluids (e.g., the East Pacific Rise), only the first two components were extracted as key components of the REY-rich mud in this study. Yasukawa et al (2016) highlighted a low sedimentation rate that enables the sediment to accumulate REY sufficiently from overlying seawater as a key environmental factor for the genesis of REY-rich mud. They quantitatively discussed the relationship between bulk REY contents and a depositional flux of REY from the water column to the seafloor, estimating that relatively high bulk contents of REY (ΣREY + Ce > 1,000 ppm) in deep-sea sediments require sedimentation rates less than~0.5 m/Myr.…”

“…Therefore, previous studies concluded that BCP incorporates REY from seawater after deposition on seafloor and/or from pore water during early diagenesis after burial (Liao et al, 2019;Kashiwabara et al, 2018;Takahashi et al, 2015;Yasukawa et al, 2015Yasukawa et al, , 2016. In any case, a high concentration of REY in the sediment generally requires a low sedimentation rate (Sa et al, 2018;Yasukawa et al, 2016). However, the formation mechanism of the extremely REY-rich mud has not been fully understood.…”

Classification of geochemical data based on multivariate statistical analyses: Complementary roles of cluster, principal component, and independent component analyses

“…In geochemistry, petrology‐mineralogy, and geology, although such approaches seem slightly less common, various multivariate statistical methods have been applied to unravel the data structures including trends, groups, and related end‐members. These include linear regression, cluster analysis, discriminant analysis, principal component analysis, factor analysis, and independent component analysis: e.g., application to various petrological problems [ Le Maitre , ]; identification of mantle geochemical structures [ Zindler et al ., ; Allègre et al ., ; Hart et al ., ; White and Duncan , ; Iwamori and Albarède , ; Stracke , ]; classification and source identification of sediment [ Pisias et al ., ; Yasukawa et al ., ] or volcanic rocks [ Brandmeier and Wörner , ]; and rock‐tectonic setting association [ Agrawal et al ., ; Snow , ; Vermeesch , ; Verma et al ., ]. Additionally, advanced methods of supervised machine learning have been applied recently to identify Tsunami deposits [ Kuwatani et al ., ] and tectonic discrimination of igneous rocks based on PetDB and GEOROC databases [ Petrelli and Perugini , ].…”