Reading Earth's surface history from image archives

Abstract

Earth's first animal-built reefs, which arose over 500 million years ago during the Cambrian period, resemble modern coral reefs in terms of both organism-scale characteristics and large-scale structural features. Local fauna appear to have relied the habitats provided by these reefs, and so these homologies between reef environments across eons may have contributed to the original rise of animal biodiversity seen in the Cambrian. These main conclusions presented within this dissertation would not be possible without first reconsidering the methods by which we interpret past physical environments preserved in sedimentary rocks. Reading physical environmental characteristics from the rock record in a manner that furnishes true analogy between Earth's deep past and near future is challenging. Unlike the chemical and ecological components of ancient environments, which have received considerable attention in recent decades to improve insights through the development of statistical and laboratory techniques, physical characteristics have been interpreted in a similar fashion for over 150 years. Factors like water depth and wave energy often are read through qualitative descriptions of features like grains, bedforms, and fossil morphologies. While this approach has led to many discoveries about the nature of Earth history, it prevents quantitative study at the intersection of physical, chemical, and biological components and truly linking Earth's environmental evolution across time. Additionally, geologists' descriptions of physical rock properties are not reproducible and are particular to the researcher that made them. All the metadata that accompany a single qualitative assessment, such as the geologists' biases, the scientific paradigms of the era in which they are working, or their internal points of reference for a descriptor like `very coarse,' are unrecorded and cannot be removed from the raw physical data. These confounding factors make qualitative characterizations imprecise and limit the ability of future geologists to build from past work in a continuous scientific dialog. Following an introduction, I present three chapters that investigate Cambrian reefs through new methods for quantitative and reproducible geophysical measurements made through capture and analysis of petrographic images. Because reefs are perhaps the most physically-complex sedimentary deposits in the rock record, they are difficult to constrain through traditional qualitative assessments, and so unlocking the ability to study these fossil structures and the habitats they provided to past animals is a particular benefit of quantitative geophysical sedimentology. The first of these three chapters investigates the three-dimensional (3D) morphology of the animals building Earth's first reefs, and, through numerical measurements, compares them to modern corals. The two chapters that follow present a high-resolution, multi-spectral petrographic camera that captures important spectral properties of fossils and minerals to enhance the amount of geologically-relevant information available in a single image. Combining this camera with a laboratory process to expedite the preparation of petrographic samples, I perform a combined ecological, chemical, and physical survey of a set of Cambrian reefs. In this set of studies, I discover that these ancient structures were built by heavily-calcified, photosymbiotic animals that had advantageous morphologies for transferring nutrients and waste between the pelagic and benthic realms. This suite of fossil reefs shows comparable chemical properties compared to modern examples, and associated fauna preferentially occur in and around the reefs as opposed to other contemporaneous environments. In all, these chapters point to the rich conclusions we can reach when studies adequately utilize analogy between Earth's deep past and present to test specific environmental hypotheses. In the fifth chapter of this thesis, I take the lessons of observing Earth's deep past through bespoke image analyses, and apply them to a question of modern environmental change. Specifically, the Landsat satellite image record holds over four decades of data that tell the story of recent surface evolution. Tracking changes in vegetation in this record may hold important historical information to give context and corroboration to accounts of human rights abuses. These changes are subtle, and require analysis that considers both natural surface conditions and how small anthropogenic changes may be manifest in images. This final chapter presents a new, interpretable method to parse this satellite image record and detect instances of land use change in a human rights application.