Spectral analysis as proxy for lineament spatial distribution: validation and case study

Spectral analysis (SA) is a technique commonly used in signal and image processing that makes use of the Fast Fourier Transform to compute the 2D power spectrum, which is a representation of the magnitude of each frequency component of the signal or image. SA can be similarly performed on a topographic map, and the orientation, frequency and magnitude (or power) of general topographic trends can be automatically retrieved and displayed in the 2D power spectrum. Recent studies have shown that spectral analysis can be successfully used to characterize repetitive and spatially homogeneous features or landforms, such as ridge and valley or glacial lineations. However, although these repetitive features dominate the 2D power spectrum, all the topographic information of the map is still present. Therefore, SA can be used on heterogenous and complex topographic map as a proxy for lineament analysis.Lineament analysis is broadly used in a wide number of applications which include tectonic studies, exploration for groundwater, hazard evaluation for tunnel excavation, rockfalls or waste repository etc. Here, we propose a new methodology for lineament analysis based on spectral analysis and we demonstrate that this is a fast and effective way to derive lineament spatial distribution from images that can be visualized as rose diagrams. To validate our methodology, we stochastically generated 1000 synthetic lineament networks and numerically compared the rose diagrams derived from the power spectra to known lineament distribution. The comparison held a similarity of 94%.The methodology was also applied to the Oslo region and compared to automatically extracted lineaments from OttoDetect software (developed by the Geological Survey of Norway). Results on three pre-selected areas characterized by different topographic patterns showed similarity of 97%, 95%, and 90%, respectively.One of the pitfalls of spectral analysis is the lack of positioning on the original map of the signatures in the power spectrum. To locate the main signature on the map, we used the orientation of the main signatures from the power spectrum and used cross-correlation and clustering methods on topographic profiles.