Page 8 - i1052-5173-32-9

A

C

B C D E

Figure 4. Example of a reproducible, quantitative lithofacies data set. (A) The lower Ordovician Kinblade Formation outcropping in Ardmore Oklahoma (GPS
location: 34.372821, –97.145353) is a 791 m succession of carbonate strata containing 1,922 beds. (B) Following bed-by-bed field study, sampling, and geo-
chemical measurement, we epoxy 1 cm chips from each sample for efficient grinding, polishing, and imaging. The chip size is chosen to best encapsulate
2
the dominant grain, fossil, and bedform sizes in the data set. The resulting ~2,000 images (examples C–E, shown here in red-green-blue space, but all are
8-channel multispectral images) now are a documentation of the lithofacies in the measured section at Ardmore and ready for image analysis, classifica-
tion, segmentation, and interpretation.

data that quantify spatio-temporal patterns tend to be documented primarily in field sions. We also appreciate the constructive feedback
in grain, crystal, and fossil characteristics, notes or illustrative outcrop/sample photo- received from reviewer Sarah Jacquet and editor
while allowing new tests of geochemical graphs. For corroboration or expansion upon James Schmitt.
interpretations (e.g., Geyman and Maloof, previous outcrop-based studies, this system
2021; Ahm et al., 2019). requires workers to visit the locality them- REFERENCES CITED
At the same time, amassing a standard- selves. Instead, open access to standardized Ahm, A.S.C., Bjerrum, C.J., Blättler, C.L., Swart,
ized, multispectral image library with petrographic image collections will allow P.K., and Higgins, J.A., 2018, Quantifying early
annotated examples (Deng et al., 2009) of broader groups of researchers to measure marine diagenesis in shallow-water carbonate
sediments: Geochimica et Cosmochimica Acta,
geologic features from many localities will and interpret features in rock formations from v. 236, p. 140–159, https://doi.org/10.1016/j.gca
help train more general machine learning around the world, enhancing both reproduc- .2018.02.042.
models for petrography. These collections ibility (Baker, 2016) and diversity, equity, Ahm, A.S.C., Maloof, A.C., Macdonald, F.A., Hoff-
of slab and thin-section images are a first and inclusion (Fernandes et al., 2020) in geol- man, P.F., Bjerrum, C.J., Bold, U., Rose, C.V.,
step toward the goal of automated routines ogy. Although our archives are not contin- Strauss, J.V., and Higgins, J.A., 2019, An early
diagenetic deglacial origin for basal Ediacaran
to measure features in rock samples from uous records like IODP cores, they benefit “cap dolostones”: Earth and Planetary Science
pictures. Curated image libraries also can from the added spatial context available at Letters, v. 506, p. 292–307, https://doi.org/ 10.1016/
serve as a classroom tool for teaching petrog- rock outcrops and provide a zoomed-in per- j.epsl.2018.10.046.
raphy, and student work to classify images spective to supplement constantly improv- Ahm, A.S.C., Bjerrum, C.J., Hoffman, P.F., Mac-
can provide training examples for machine ing aerial survey techniques (Shah et al., donald, F.A., Maloof, A.C., Rose, C.V., Strauss,
J.V., and Higgins, J.A., 2021, The Ca and Mg iso-
learning—a crowdsourcing technique that 2021). In concert with satellite, drone-derived, tope record of the Cryogenian Trezona carbon
has seen recent success in several fields and hand-held imagery, our pipeline for sys- isotope excursion: Earth and Planetary Science
(e.g., van den Bergh et al., 2021). tematic imaging, classification, and mea- Letters, v. 568, 117002, https://doi.org/10.1016/
In addition to improving lithofacies data, j.epsl.2021.117002.
we see our camera and petrographic images surement of rock samples can form an Baker, M., 2016, Reproducibility crisis: Nature,
important layer in multiscale digitization
v. 533, p. 353–366.
as a vehicle to improve access and reproduc- and interpretation of physical rock properties. Becker, K., Austin, J.A., Exon, N., Humphris, S.,
ibility in geology. Open access to archives Kastner, M., McKenzie, J.A., Miller, K.G., Suye-
like Integrated Ocean Drilling Program hiro, K., and Taira, A., 2019, 50 Years of Scien-
(IODP) cores has expanded the number of ACKNOWLEDGMENTS tific Ocean Drilling: Oceanography (Washing-
The authors would like to thank NSF EAR-
people producing complementary data sets 1028768 and Princeton University for funding. ton, D.C.), v. 32, p. 17–21, https://doi.org/10.5670/
oceanog.2019.110.
and provided for deeper, more reproducible Arnab Chatterjee and Peter Siegel at Digital Transi- Bergmann, K.D., Zentmyer, R.A., and Fischer,
studies of Earth’s climate and oceans in tions, Nolan Greve and Jeremy Brodersen at Smart- W.W., 2011, The stratigraphic expression of a
recent geologic periods (Becker et al., 2019). Vision Lights, Mark Harfouche and Gregor Horst- large negative carbon isotope excursion from the
A similar framework should exist for rock meyer at Ramona Optics, and Dave Crawford and Ediacaran Johnnie Formation, Death Valley:
outcrops that span deeper into Earth’s his- Jenifer Powell at Moxtek provided support with the Precambrian Research, v. 188, p. 45–56, https://
camera, lights, light table, and polarizers, respec-
doi.org/10.1016/j.precamres.2011.03.014.
tory, where, currently, the observations that tively. We thank Bolton Howes, Cedric Hagen, Bishop, J.L., Bell, J.F., III, and Moersch, J.E., edi-
form geologic maps and stratigraphic sections Brad Samuels, and Michael Eddy for useful discus- tors, 2019, Remote Compositional Analysis:
8 GSA TODAY | September 2022