EP060-0001
AN UPDATED GLOBAL PREDICTION OF MARINE UNIT THICKNESSES FOR PRESENT TO MIDDLE MIOCENE SEDIMENTS USING OCEAN DRILLING PROGRAM DATA.

Wednesday, 16 December 2020
Poster
Angelena Luciano, ASEE Science and Engineering Apprenticeship Program, U.S. Naval Research Laboratory, Washington, DC, United States, Taylor Runyan Lee, US Naval Research Laboratory, Geology and Geophysics, Washington, DC, United States, Benjamin J Phrampus, US Naval Research Laboratory, Washington, DC, United States, Jeffrey Obelcz, U. S. Naval Research Laboratory, Geology and Geophysics, Stennis, United States, Warren T Wood, Naval Research Laboratory, Stennis Space Ctr, MS, United States and Adam D Skarke, Mississippi State University, Department of Geosciences, Mississippi State, MS, United States
Abstract:
Isopach maps represent the stratigraphic thickness of sedimentary units, commonly bound by geologic
time horizons. The variability in isopach thicknesses are a result of a variety of geological processes
including terrestrial sediment flux, glacial cycles, and bathymetric interactions. Because marine isopachs
represent sediment deposition and may serve as an important sink to sequester carbon in recent and past
geologic time, there exists an increasing need to understand unit thicknesses on a global scale.


The first global scale age-constrained marine unit isopach estimates were completed this year (Lee et al.,
in revision). In this first demonstration, each isopach used ~550 depth vs. age estimates from the Deep
Sea Drilling Project (DSDP) with a k-nearest neighbor machine learning algorithm to predict global unit
thickness and produce isopach maps of recent geologic epochs. Since the first iteration of global isopachs,
for each isopach we have added ~150 new observed depth vs. age data compiled from the Ocean Drilling
Campaign (ODP).


For each set of isopach estimates, we linearly interpolate between age-depth observations to determine the
depth to mid-Pleistocene, Pliocene and middle Miocene sediments. The differences in depth between each
epoch are used to determine isopach thicknesses between the present and middle Miocene. These isopach
observations are used to geospatially predict isopachs at all marine locations, using machine learning, and
a large library of previously assembled predictors. The predicted isopach thicknesses and total depths are
validated in the machine learning process using 10-fold cross validation.


Here, we present both the first and second iterations of a globally complete marine isopach map for
geologic time epochs dating back to middle-Miocene. More specifically, we will address how the addition
of new data has changed the global distribution of isopachs as well as the error in final prediction.
Predictive capability in isopach predictions is expressed as a function of the absolute median error and R 2
value calculated through the 10-fold cross validation of observed vs. predicted values. In general, for each
isopach, the new ODP observations converge within the distribution of the DSDP histogram (i.e.
increased redundancy of observation values).

Previous Abstract | Next Abstract >>