Preliminary Results from Applying Machine Learning to Classical Barrovian Metamorphism: Advancing Petrographic Standardization

Petrographic analysis is a cornerstone of petrologic and mineralogical investigations within geology and is widely debated for its reliance on expert observation and interpretation, which can leave these methods being time-intensive and subject to observer bias. As geoscience increasingly moves toward large, image-based datasets, questions remain about how traditional petrography can be standardized and enhanced using computational approaches without losing geological context. Although machine learning has been applied in various geologic disciplines, its use in classical metamorphic sequences remains limited. A clear gap exists in the application of machine learning techniques to petrographic analysis of Barrovian metamorphic rocks, despite their importance as a benchmark for understanding progressive regional metamorphism. Previous studies focus on qualitative petrographic interpretation, leaving Barrovian zones largely unexplored through automated image classification. This project addresses that gap by integrating supervised and unsupervised machine learning with petrographic analysis of thin sections collected from Barrovian metamorphic zones in Scotland. Thin sections will be examined under plane- and cross-polarized light to document characteristic mineral assemblages and textures, including chlorite, biotite, garnet, staurolite, kyanite, and sillimanite. Petrographic images will be treated as multispectral datasets and classified using ERDAS Imagine, while ArcGIS Pro will be used to construct a digital metamorphic zone map and analyze spatial relationships. This project argues that machine learning can effectively complement traditional petrography by increasing efficiency, reducing subjectivity, and identifying subtle mineralogical and textural patterns that may be overlooked through manual analysis, advancing modern metamorphic research methods.