Quantifying and analysing rock trait distributions of rocky fault scarps using deep learning

We apply a deep learning model to segment and identify rock characteristics based on Structure-from-Motion orthomap digital elevation of rocky fault scarp in the Volcanic Tablelands, Eastern California, USA. By post-processing results, we build semantic map analyse trait distributions. The resulting contains nearly 230 000 rocks with effective diameters ranging from 2 250 cm. Rock distributions provide new perspective development extend past research geometry including slope, height length. Heatmaps indicate size spatial surrounding topographic flats. Median grain changes perpendicular trace, largest downslope proximal footwall. Correlation analyses illustrate relationship between statistics geomorphology. Local correlates median ( R $$ {R}^2 = 0.6), mean 0.8) ratio number small large 0.4). positive correlation local standard deviation suggests that higher are less well sorted. analysis orientation supports particle transportation which locally scarps have relatively more long axes parallel trace because larger distance roll orient axes. Our work demonstrates data-driven approach geomorphology distributions, promising greater understanding formation tectonic activity, as many other applications for granulometry is an indicator process.