TEXTURAL CONSTRAINTS AND IMAGING TECHNIQUES: BIAS IN EXTRACTING CRYSTAL SIZE DISTRIBUTIONS

Crystal size distributions (CSDs) in volcanic textures represent the interaction of crystal nucleation, growth, and magma dynamics in pre-, syn- and post-eruptive environments. The goal of CSD analysis is to provide insight into magma chamber processes which are not themselves directly observable. Although considerations for crystal shapes (aspect ratios), length measurements, intersection-probability effects, and cut-section effects are routinely taken into account, there exists no protocol for extracting CSDs from volcanic textures. To examine the method- and image-based limitations imposed by texture, we obtain CSDs from two texturally diverse basaltic lavas using approaches that consider texturally imposed biases of 2D image types. High-resolution backscatter electron, electron backscatter diffraction, and energy-dispersive X-ray imaging are used to demonstrate that 1) plagioclase phenocrysts (L ≥ 150 μm) and groundmass (L < 150 μm) suggest multiple aspect ratios, while the latter also displays greater nucleation densities as crystal size population are considered over increasingly smaller crystal lengths; 2) complex crystal clusters must be manually dissected into their discrete crystal components to minimize coarse-size skewing; 3) localized electron backscatter diffraction reveals mild preferred orientation in complex clusters and groundmass, the latter confirmed by small variations in CSDs between orthogonal backscatter electron images; and 4) small variations in aspect ratios for CSD calculations potentially create shifts in CSD curves. Our results suggest that to produce robust CSDs, extracting crystal lengths for plagioclase crystals from 2D images requires textural compensation through manual tracing as well as the use of high-resolution scanning electron imaging measurements for thousands of plagioclase groundmass crystals which permit deciphering of late-stage eruptive behavior. A texture-focused protocol that considers the above must be established, as results from CSD analyses across multiple studies using multiple approaches cannot otherwise be directly compared.