SEDIMENTATION THROUGHOUT THE GLOBAL OCEAN: NEW DETAILED MAPPINGS AND STATISTICS

Composition of the seabed has proven difficult to map in detail on the scale of whole ocean-basins. This project, dbSEABED, has the largest, best integrated collection of seafloor materials data worldwide, over 5 million observations. It uses specially designed data entry workflows then advanced computing methods to obtain the best possible coverages and detail. The collection includes numerous geologically dated core/dredge samples, as well as surficial contemporary samplings.

Direct observation and proxy methods (such as acoustics) are both in demand to describe the seafloor. However dbSEABED focuses on direct observation results since these have rich content for parameters like consolidation, color, odor, structures and components. (They are also ground-truthing for the proxy methods.) Uniquely, the system handles and integrates word-based data, which is necessary to achieve good coverage (85% of seabed direct observations are word-descriptions), for reporting unusual materials, and to convey ideas on of processes and origins. The heterogeneous data integration involves statistical testing, geostatistics, text information extraction (Natural Language Processing, NLP; machine learning, ML), and fuzzy set theory (FST). Validation of results is via cross-calibration with accompanying numerical analytical data.

Because of the unprecedented detail and geographic and stratigraphic (subbottom, cored) coverage of the dataset, it is a powerful resource for statistical stratigraphy. This is the analysis of the spatial-temporal trends and variances in sediments across the environmental zones of the oceans, whether by latitude, water depth, source proximality, water bodies, or geologic time epoch. We present examples of this form of analysis.