Relating process to form in jointed-bedrock fault scarps using morphologic variability, a novel metric derived using supervised landform classification.

Understanding the evolution of landform morphology through time requires identifying and quantifying the underlying structural and geomorphic processes. However, given the complexity of natural systems, isolating the effects and rates of specific processes can be a daunting task. We present a new morphologic variability metric, which measures the frequency and degree of change in landform-profile shape and can provide insights into the evolution of complex landforms that traditional approaches might overlook. We apply our methodology to a common, but morphologically complex type of landform: fault scarps in jointed bedrock. Using point clouds derived from structure-from-motion photogrammetry, we computed the morphologic variability along jointed-bedrock fault scarps from four field sites, located in southwestern Iceland, northern and central California, and southeastern Hawaii. Using maximum scarp height as a proxy for scarp maturity, we find that morphologic variability decreases as scarps mature, suggesting that scarp-profile form evolves towards a common morphology as scarps mature. We also compare scarp morphologic variability to other parameters that vary along the scarp, including fracture characteristics. Using dominance analysis to quantitatively determine which of these characteristics had the most impact on the scarps morphologic variability, we analyze a post-glacial Icelandic fault scarp. The parameters with the highest relative importance were the facing direction of the fracture planes (14%) and the spacing between horizontal fractures (10%). This suggests that processes that are direction-dependent and the heterogeneity of the primary rock structure are important contributors to variation in form of this young scarp. Since dominant processes are likely to change as the scarp matures, the relative importance of a given parameter can be used to fingerprint the evolutionary stage of the landform, as demonstrated by our analyses of the significantly more mature Medicine Lake Volcano and Hilina Pali escarpments. Our method of harnessing statistics and machine learning to quantify changes in landform shape and process can be applied to many other types of landforms at fine and broad scales, considering both evolutionary stage and setting.