Regional‐Scale Detection of Fault Scarps and Other Tectonic Landforms: Examples From Northern California

Fault scarps and fault‐related landforms provide important information about fault zone activity over timescales that are not captured by instrumental measurements or historic records. Semiautomated methods for delineating these landforms using topographic data from light detection and ranging (lidar) and spaceborne imaging systems offer the opportunity to characterize fault zones on a global scale. We present a computationally efficient method for extracting scarp‐like landforms from high‐resolution (≤2 m), regional‐scale (≥ 100‐km‐long) digital topographic data sets. We identify fault‐related landforms using a curvature template based on the diffusion model for scarp degradation and extract scarp heights and morphologic ages at each pixel. The method was applied to the GeoEarthScope Northern California data set, an airborne lidar acquisition imaging nearly 2,500 km2 of the northern San Andreas Fault system, by adapting the algorithm to use cloud computing resources. Template results and fault trace mapping show spatial agreement in active fault zones with clear topographic expression, including detection of fault scarps, shutter ridges, and elongated drainages. Comparison of the method against field‐based morphologic dating of scarps along the southern San Andreas reveals a trade‐off between template window size and morphologic age contrasts resolved between strike‐slip fault scarps of different relative ages. Detection performance suggests that window size and orientation constraints may play a key role in improving the accuracy of methods for semiautomated fault zone mapping. As data availability grows, these methods could constrain key earthquake simulation parameters such as damage zone width or rupture length and improve fault maps worldwide. Key Points A cloud‐based template matching algorithm enables detection and relative dating of scarp‐like landforms in large topographic data sets Performance of pixel classification methods using detected landforms varies with changes in window size and orientation constraints Comparison to field‐based morphologic dating reveals that the variability of morphologic age estimates decreases with increasing window size