Combined ambient seismic noise tomography and H/V analysis to decipher the shallow subsurface in Saxon Lusatia (eastern Germany)

Saxon Lusatia (eastern Germany) is considered a region with particularly low seismic background noise. It is therefore earmarked as a possible location for a so-called Low Seismic Lab as part of the newly founded German Center for Astrophysics (DZA) and for the Einstein Telescope, the new generation of gravitational wave detectors.As part of the preliminary site investigations, several temporary seismic networks (a total of almost 400 stations) were operated in the area between Bautzen, Kamenz, and Hoyerswerda in 2024 and 2025. The main objectives were to create a 3D model of the subsurface (shear wave velocity; ambient noise tomography) using the seismic ambient noise field, and to investigate the spatial-temporal distribution of seismic noise (and noise sources).Following the general approaches to analyzing ambient seismic noise, we started with a division of the data sets (vertical component data) into hourly segments, followed by bias removal and trend correction, as well as spectral brightening and 1-bit normalization. These pre-processed hourly segments were then used to calculate cross-correlations. Finally, these individual hourly cross-correlations were stacked to obtain the final empirical Green's functions for every station pair.In the next step, the Rayleigh dispersion curves were determined interactively for a large number of cross-correlations. A general observation for the FTAN displays was that in almost all cases, the energy content of the selectable dispersion curves is very frequency-limited (typically 1.5–4 Hz) and that the data is noisy. This suggests that the tomographic resolution of the subsurface structures will be quite limited. Given the expected model complexity with a strongly varying layer of unconsolidated sediments of variable thickness (1–200 m) on top of high-velocity granodiorite, we focused our dispersion curve analysis on traces with offsets < 2 km.The inversion was performed using a Bayesian statistical method, namely a transdimensional hierarchical Monte Carlo search using Markov chains and a Metropolis/Hastings sampler. This is a full tomographic inversion technique that can be used to derive the 3D distribution of shear wave velocity and the associated uncertainty. Given the difficult initial situation with regard to the data (noise, band-limited), we extended the inversion of the dispersion curves to include H/V data from 128 three-component stations.Using seismic ambient noise data (dispersion curve and H/V data), we were able to successfully create a three-dimensional model of the shallow (