The complex tectonic setting of the central‐western Mediterranean has interested geoscientists for decades, but its geodynamic evolution remains a matter of debate. We rely on 807 seismometers from southern Europe and northern Africa to measure Rayleigh and Love phase velocities in the period range ∼5–200 s, based on teleseismic earthquakes and seismic ambient noise. By nonlinear joint inversion of the phase‐velocity maps, we obtain a 3‐D shear‐wave velocity (V_S) model of the study area. At shallow depths, our model correlates with surface geology and reveals the presence of a sedimentary cover in the Liguro‐Provençal basin, as opposed to the Tyrrhenian basin where this is either very thin or absent. At ∼5‐km depth, high velocities below the Magnaghi, Vavilov, and Marsili seamounts point to an exhumed, scarcely serpentinized mantle. These are replaced by lower velocities at larger depths, likely connected to the presence of partial melt. At 50–60‐km depth, a very heterogeneous structure characterizes the Tyrrhenian basin, with low velocities pointing to the presence of fluids due to the lateral mantle inflow from the Ionian slab edges, and higher velocities associated with a relatively dry upper mantle. Such heterogeneity disappears at depths ≳75 km, replaced by more uniform velocities which are ∼2% lower than those found in the Liguro‐Provençal basin. We infer that, at the same depths, the Tyrrhenian basin is characterized by a larger concentration of fluids and possibly higher temperatures.

Summary: To improve our understanding of the Earth’s interior, seismologists often have to deal with enormous amounts of data, requiring automatic tools for their analyses. It is the purpose of this study to present SeisLib, an open-source Python package for multiscale seismic imaging. At present, SeisLib includes routines for carrying out surface-wave tomography tasks based on seismic ambient noise and teleseismic earthquakes. We illustrate here these functionalities, both from the theoretical and algorithmic point of view and by application of our library to seismic data from North America. We first show how SeisLib retrieves surface-wave phase velocities from the ambient noise recorded at pairs of receivers, based on the zero crossings of their normalized cross-spectrum. We then present our implementation of the two-station method, to measure phase velocities from pairs of receivers approximately lying on the same great-circle path as the epicentre of distant earthquakes. We apply these methods to calculate dispersion curves across the conterminous United States, using continuous seismograms from the transportable component of USArray and earthquake recordings from the permanent networks. Overall, we measure 144 272 ambient-noise and 2055 earthquake-based dispersion curves, that we invert for Rayleigh-wave phase-velocity maps. To map the lateral variations in surface-wave velocity, SeisLib exploits a least-squares inversion algorithm based on ray theory. Our implementation supports both equal-area and adaptive parametrizations, with the latter allowing for a finer resolution in the areas characterized by high density of measurements. In the broad period range 4–100 s, the retrieved velocity maps of North America are highly correlated (on average, 96 per cent) and present very small average differences (0.14 ± 0.1 per cent) with those reported in the literature. This points to the robustness of our algorithms. We also produce a global phase-velocity map at the period of 40 s, combining our dispersion measurements with those collected at global scale in previous studies. This allows us to demonstrate the reliability and optimized computational speed of SeisLib, even in presence of very large seismic inverse problems and strong variability in the data coverage. The last part of the manuscript deals with the attenuation of Rayleigh waves, which can be estimated through SeisLib based on the seismic ambient noise recorded at dense arrays of receivers. We apply our algorithm to produce an attenuation map of the United States at the period of 4 s, which we find consistent with the relevant literature.

We evaluate, by numerical tests, whether surface‐wave attenuation can be determined from ambient‐noise data. We generate synthetic recordings of numerically simulated ambient seismic noise in several experimental setups, characterized by different source distributions and different values of attenuation coefficient. We use them to verify that the source spectrum can be reconstructed from ambient recordings (provided that the density of sources and the attenuation coefficient are known) and that true attenuation can be retrieved from normalized cross correlations of synthetic signals. We then apply the so validated method to real continuous recordings from 33 broadband receivers distributed within the Colorado Plateau and Great Basin. A preliminary analysis of the signal‐to‐noise ratio as a function of azimuth reveals a SW‐NE preferential directionality of the noise sources within the secondary microseism band (6–8 s), consistent with previous studies. By nonlinear inversion of noise data we find the attenuation coefficient in the area of interest to range from ∼ 1 × 10⁻⁵ m⁻¹ at 0.3 Hz to ∼ 4.5 × 10⁻⁷ m⁻¹ at 0.065 Hz, and confirm the statistical robustness of this estimate by means of a bootstrap analysis. The result is compatible with previous observations based on both earthquake‐generated and ambient Rayleigh waves. In this regard, the method proves to be promising in accurately quantifying surface‐wave attenuation at relatively high frequencies.

Distributed acoustic sensing (DAS) offers a cost effective, nonintrusive method for high-resolution near-surface characterization in urban environments where conventional geophysical surveys are limited or nonexistent. However, passive imaging with DAS in urban settings presents challenges such as strong diurnal traffic noise, nonlinear array geometry, and poor fiber coupling to the ground. We repurposed a dark fiber in Melbourne, Australia, into a 25 km DAS array that traces busy arterial roads, tram routes, and orthogonal sections. By employing noise cross correlation and array beamforming, we calculated dispersion curves and successfully inverted for a near-surface shear-wave velocity model down to 100 meters. Stationary seismic sources are maximized by selecting daytime traffic signals, thereby recovering surface waves and reducing interference from acoustic waves from man-made structures in the subsurface. Poorly coupled channels, which are linked to fiber maintenance pits, are identified through cross-correlation amplitudes. The dispersion curve calculation further considers the channel orientation to avoid mixing Rayleigh and Love waves. Using a trans-dimensional Markov chain Monte Carlo sampling approach, we achieved effective model inversion without a prior reference model. The resulting near-surface profile aligns with mapped lithology and reveals previously undocumented lithological variation.

SUMMARY: We compile a data set of Rayleigh-wave phase velocities between pairs of stations, based on teleseismic events located on the same great circle as the two stations. We validate our observations against dispersion estimates based on ambient-noise cross correlations at the same station pairs. Discrepancies between the results of the two methods can in principle be explained by deviations in the wave propagation path between earthquake and receivers, due to lateral heterogeneity in the Earth’s structure, but the latter effect has, so far, not been precisely quantified nor corrected for. We implement an algorithm to measure the arrival angle of earthquake-generated surface waves and correct the dispersion measurements accordingly. Application to a data set from the Central-Western Mediterranean shows that the arrival-angle correction almost entirely accounts for the discrepancy in question, decreasing significantly the velocity bias for a wide range of periods.