A deep-learning approach to quantify and analyze rock trait distributions for the geomorphic study of rocky fault scarps
Zhiang Chen, Chelsea P. Scott, Devin Keating, Amanda Clarke, Jnaneshwar Das, & Ramon ArrowsmithSubmitted April 25, 2022, SCEC Contribution #11844
Research on fault scarp evolution has focused on geometric indicators such as scarp slope, height, and length, but rock trait distributions can provide a new perspective to understand rocky fault scarp development. We apply a deep learning model to segment and identify rocks based on a structure-from-motion orthomap and digital elevation model of a rocky fault scarp in the Volcanic Tablelands, Eastern California. By post-processing the deep learning inference results, we build a semantic rock map (geospatial database of rock polygons) and analyze rock trait distributions. The resulting semantic map has nearly 230,000 rocks between 2 cm and 250 cm in effective diameter. Heatmaps indicate rock size spatial distributions on the fault scarp and surrounding topographic flats. Median grain size changes perpendicular to the fault strike with largest grains exposed on and down slope from the scarp footwall. Correlation analyses of the segmented fault scarp illustrate the relationship between rock trait statistics and fault scarp geomorphic characteristics. Local fault scarp heights correlate with the median grain size, the ratio between the number of small and large rocks, and the mean grain size of the largest rocks. These correlations indicate that larger (locally higher) fault scarps have accumulated more geomorphic fracturing and tectonic faulting. The positive correlation between local fault scarp height and sorting suggests that rocks on a larger fault scarp are less well sorted. Correlation between fault scarp height and rock orientation statistics supports a particle transportation model in which larger fault scarps will have relatively more rocks with long axes parallel to fault scarp strike because rocks have a larger distance to roll and orient the long axes. Our work demonstrates a data-driven approach to geomorphology based on rock trait distributions, which promises a greater understanding of fault scarp formation and tectonic activity.
Citation
Chen, Z., Scott, C. P., Keating, D., Clarke, A., Das, J., & Arrowsmith, R. (2022). A deep-learning approach to quantify and analyze rock trait distributions for the geomorphic study of rocky fault scarps. Earth Surface Processes and Landforms, (submitted).