Assessment of global variations in the velocity–porosity relationship of marine sediments and rocks using the legacy data of scientific ocean drilling programs
Abstract
The physical properties of sediments and rocks are routinely measured during scientific ocean drilling programs, as such data enable the recovered materials to be linked to geophysical observations and provide constraints on subseafloor lithology. However, lithological descriptions are not consistently linked to physical properties datasets for older drilling expeditions, thereby limiting global assessments of trends in lithology-dependent physical properties across various tectonic settings. In this study I compiled publicly available physical property data collected onboard, including the lithology-labeled datasets of recent expeditions, and developed a machine learning model to assign broad lithological context to unlabeled samples. Lithological labels were simplified into six classes: biogenic/carbonate sediment, siliciclastic sediment, volcaniclastic sediment, basaltic rock, gabbroic rock, and ultramafic rock. A bagged decision tree model was trained using porosity, bulk density, grain density, P-wave velocity, and magnetic susceptibility, and was applied to unlabeled data. The model achieved an overall accuracy of ~83% on the test dataset. Results shows that global trends in velocity–porosity relationships vary with lithology and show large internal variability. Applying a linear mixed-effects model, which statistically separates systematic effects shared across the dataset from group-specific variability, the deviations from the global trend are attributed primarily to variations between drilling areas and heterogeneity within each drilling area, rather than the tectonic setting. These results demonstrate that regional geological processes and local heterogeneity should be considered when utilizing physical property data for modeling and interpreting geophysical structures.
Related
- Relationship between Meteorological Parameters and Temporal Variations of Microseismic Processes at the Belarusian Antarctic Research Station
- Assessment of trace metal contamination in mangrove sediments—a case study from Guadeloupe Island, Caribbean Sea
- Assessment of the Seasonal Variations in Heavy Metals Concentration in the Ughelli Central Market River, Delta State, Nigeria
- Spatial distribution and pollution assessment of potentially toxic elements in stream sediments of the Luis Moya region, Central Mexico: Evidence from mining-affected areas
- Assessment of potentially toxic elements level in surface sediments of rivers in a territory of strong industrial legacy
BibTeX
@misc{akamatsu2026assessment,
title = {{Assessment of global variations in the velocity–porosity relationship of marine sediments and rocks using the legacy data of scientific ocean drilling programs}},
author = {Yuya Akamatsu},
howpublished = {EarthArXiv},
year = {2026},
month = sep,
doi = {10.31223/x55r5f},
url = {https://eartharxiv.org/repository/view/14938/},
note = {CC BY-ND 4.0},
}
RIS
TY - UNPB
TI - Assessment of global variations in the velocity–porosity relationship of marine sediments and rocks using the legacy data of scientific ocean drilling programs
AU - Yuya Akamatsu
PY - 2026
DA - 2026/09/11
PB - EarthArXiv
DO - 10.31223/x55r5f
UR - https://eartharxiv.org/repository/view/14938/
AB - The physical properties of sediments and rocks are routinely measured during scientific ocean drilling programs, as such data enable the recovered materials to be linked to geophysical observations and provide constraints on subseafloor lithology. However, lithological descriptions are not consistently linked to physical properties datasets for older drilling expeditions, thereby limiting global assessments of trends in lithology-dependent physical properties across various tectonic settings. In this study I compiled publicly available physical property data collected onboard, including the lithology-labeled datasets of recent expeditions, and developed a machine learning model to assign broad lithological context to unlabeled samples. Lithological labels were simplified into six classes: biogenic/carbonate sediment, siliciclastic sediment, volcaniclastic sediment, basaltic rock, gabbroic rock, and ultramafic rock. A bagged decision tree model was trained using porosity, bulk density, grain density, P-wave velocity, and magnetic susceptibility, and was applied to unlabeled data. The model achieved an overall accuracy of ~83% on the test dataset. Results shows that global trends in velocity–porosity relationships vary with lithology and show large internal variability. Applying a linear mixed-effects model, which statistically separates systematic effects shared across the dataset from group-specific variability, the deviations from the global trend are attributed primarily to variations between drilling areas and heterogeneity within each drilling area, rather than the tectonic setting. These results demonstrate that regional geological processes and local heterogeneity should be considered when utilizing physical property data for modeling and interpreting geophysical structures.
ER -