The characterization of landslide heterogeneity in urbanized area using geophysical and machine learning methods: a case study from Cieszyn, Poland

Landslides in urban areas pose special challenges for engineering geology. Because of the high risk they pose, they require special attention. In the presented work, the key novelty is an approach using geophysical imaging methods and unsupervised machine learning to identify a high-risk landslide in an urban area. It proved insufficient in the case presented here, and the proposed approach made it possible to identify the slip surface much more accurately. The results obtained were verified and supplemented with borehole data. Combining model generation based on machine learning can be applied as a new solution.

The research presented concerns the analysis of the stability of a slope located in the centre of the city of Cieszyn (Voivodeship, Silesia, Poland). The research used geophysical methods, including electrical resistivity tomography, refraction seismic and multichannel surface wave analysis. The essence of the study was to identify the geological structure and determine the slip surface of the rock masses, which are expected to answer whether further urbanization and development of the area is possible on the studied slope and whether the recognized landslide threatens lower-lying structures. As a result of the research, the object of study was recognized, and the effectiveness of the assumed cost-effective methodology was presented. The described example and used approach can broadly apply to similar research problems in the Carpathian region and for imaging similar geotechnical problems in other parts of the world.