Preliminary Seismic Signature Analysis of the Tagliamento River During Flood Events Using Machine Learning Algorithms

Identifying the seismic signature of rivers (e.g., flow and bedload) is a significant challenge due to the varying responses of the investigation site and the hydrodynamic parameters controlling river streams during flood events. Moreover, environmental noise, such as wind and rain components, is not always easily distinguishable from the signal generated by river motion, given their overlapping frequency ranges.

We analysed the seismic signature of the Tagliamento River, located in Friuli-Venezia Giulia (Northeast Italy), recognised as one of the "last large natural alpine rivers in Europe." This river is characterised by significant water level rises and gravel sediment transport during extreme meteorological events. Using data from level gauges and pluviometric sensors alongside seismic stations installed along the river, we examined the relationship between increasing water levels, rainfall indices, and the amplitude of seismic waves recorded by seismometers during multiple flood events from 2018 to 2024.

Additionally, we performed detailed analyses, including cross-correlation, time-of-concentration calculations, and seismic signal polarisation, to better characterise river behaviour. This preliminary study aims to understand the seismic signals generated by the turbulent flow of the river and the transported bedload using the collected data. Subsequently, we propose to develop an empirical model for water level estimation, enabling the evaluation of hydrogeological hazards during upstream floods with the assistance of machine learning algorithms.