Semi-automated fault extraction and quantitative structural analysis from DEM data, a comprehensive tool for fault network analysis

Mapping and characterisation of crustal faults represent one of the contemporary challenges for both tectonic understanding and seismic hazard assessment. Given the high resolution of satellite-derived digital elevation models and remote-sensing imagery, the development of an automatic method of fault extraction is a critical turning point. Here we present a Python-based, open-source workflow,  which is able to extract and characterize individual faults as well as entire fault networks from various datasets. 

Our workflow consists of four main steps: (1) The DEM contains different types of noise, which we reduce using Gaussian smoothing. (2) Then we use the Canny edge detection to highlight topographic discontinuities, such as faults. (3) These edges are simplified in single pixel-wide lines through the skeletonization algorithm. (4) Finally, we create a network consisting of nodes and edges from this skeleton. After a few post-processing steps we obtain a fault network of the sample area. 

We use the toolbox to study faulting in the East African Rift system, especially the Magadi Natron basin. The workflow was applied to a TanDEM-X digital elevation model with 12 m horizontal resolution and the Copernicus GLO-30 dataset with 30 m average horizontal resolution. The strike analysis shows four main directions from distinct fault populations. Moreover, we derive the fault displacement distribution throughout the basin, which allows us to calculate the total orthogonal extension of each geological unit and to compute the overall amount of extension of the region during geologically recent times.

Our workflow is designed to evaluate topographic data of target sites in nature, it can, however, also be used to analyze analogue models and numerical simulations. To this aim, specific functions can be added in a modular way to suit the particularity of the area and of available data types. This workflow allows us to imagine a very wide range of applications and subjects of interest.