Exploring the role of biotite characteristics on regolith production

Regolith production, the conversion of intact rock into a nutrient-bearing, water-holding, and life-supporting substrate, is of vital interest in geomorphology, hydrology, and soil science. For several decades, the expansion of biotite by oxidation has been implicated as an important step in the conversion of rock to regolith, yet questions remain about how an intact rock's biotite characteristics control the nature and rate of regolith production. As a first step towards this aim, we inspected nine rock and five rindlet samples from a quartz diorite core from the Luquillo Critical Zone Observatory using high-resolution back-scattered electron imagery and electron dispersive spectroscopy. We trained a machine learning algorithm to map thin-section mineralogy from electron dispersive spectroscopy data and to measure biotite abundance, size, shape, and distribution. The automated mineralogical maps agree well with manually identified phases on a pixel-by-pixel basis and overall abundance. Automated maps show ~96% agreement for pixel classification across all phases and ~90% agreement for biotite when compared to manually delineated maps from two rock samples. We applied two automated feature extraction techniques to back-scattered electron imagery to measure 2D microfracture density and orientation. Our preliminary analysis of high-resolution back-scattered electron imagery shows that incipient expansion of biotite occurs at a depth of at least ~13 m, suggesting that biotite weathering and microfracturing of surrounding minerals begins several meters below the bedrock-saprolite interface in this core. Together, these imaging techniques and machine learning algorithms hold promise for identifying biotite and microfracture characteristics and hence discerning the microscopic controls of biotite weathering on regolith production.