Characterization of Ground Conditions at Seismic Stations in the North Caucasus Using Machine Learning Methods

To extend the capabilities of using records of local earthquakes (for constructing regional ground motion prediction equations, assessing seismic hazard, etc.), the classification of seismic stations in the North Caucasus by the ground conditions was performed. A technique has been developed that allows assessment of ground conditions by comparing spectra of weak earthquakes selected in narrow ranges of magnitudes and hypocentral distances, at different stations. The use of machine-learning methods showed the complexity of the problem, but at the same time, the application of logical operations and techniques allowed us to find the most effective approaches to solve it. As a result, 70 seismic stations of the North Caucasus were classified according to the ground conditions; the conditions were characterized by one dimensionless parameter based on the calculation of spectral characteristics. We are planning to refine the estimates in the future.