Data for: 'How is sea level change encoded in carbonate stratigraphy?'

The history of organismal evolution, seawater chemistry, and paleoclimate
is recorded in layers of carbonate sedimentary rock. Meter-scale cyclic
stacking patterns in these carbonates often are interpreted as
representing sea level change. A reliable sedimentary proxy for eustasy
would be profoundly useful for reconstructing paleoclimate, since sea
level responds to changes in temperature and ice volume. However, the
translation from water depth to carbonate layering has proven difficult,
with recent surveys of modern shallow water platforms revealing little
correlation between carbonate facies (i.e., grain size, sedimentary bed
forms, ecology) and water depth. We train a convolutional neural network
with satellite imagery and new field observations from a 3,000 km2 region
northwest of Andros Island (Bahamas) to generate a facies map with 5 m
resolution. Leveraging a newly-published bathymetry for the same region,
we test the hypothesis that one can extract a signal of water depth
change, not simply from individual facies, but from sequences of facies
transitions analogous to vertically stacked carbonate strata. Our Hidden
Markov Model (HMM) can distinguish relative sea level fall from random
variability with ∼90% accuracy. Finally, since shallowing-upward patterns
can result from local (autogenic) processes in addition to forced
mechanisms such as eustasy, we search for statistical tools to diagnose
the presence or absence of external forcings on relative sea level. With a
new data-driven forward model that simulates how modern facies mosaics
evolve to stack strata, we show how different sea level forcings generate
characteristic patterns of cycle thicknesses in shallow carbonates,
providing a new tool for quantitative reconstruction of ancient sea level
conditions from the geologic record. Supporting data to reproduce the work
in Geyman et. al (2021).