Broad-band ambient noise characterization by joint use of cross-correlation and MUSIC algorithm
Résumé
Several days of passive seismic broad-band recordings (vertical component) from a dense 3 × 6 km array installed near Chémery (France), with about 100 seismometers, are analysed for wavefield characterization between 0.1 and 3 Hz. Backazimuth is determined by using the Multiple Signal Characterization (MUSIC) algorithm at frequencies below 1 Hz, and non-coherent cross-correlation beamforming above 1 Hz, since the latter is less sensitive to aliasing issues. A novel method of determining the wavefield velocity is introduced, consisting of processing a cross-correlation common-offset gather by the MUSIC algorithm. The fundamental and three higher modes of Rayleigh waves (R0, R1, R2 and R3) are identified under 1 Hz. Above 1.5 Hz, the Lg phase is detected, while R0 and R1 are also present. Roughly between 1 and 1.5 Hz, a quicker phase, probably Pg, is detected. Both Pg and Lg are dominant during night time, suggesting they have a natural origin, which is also consistent with their backazimuth pointing towards the Atlantic. Large scale 2-D spectral-element simulations using deep- and shallow-water ocean sources confirm the possibility of the Lg phase excitation. Thus, even above 1 Hz, natural sources can explain the major part of the ambient noise energy during quiet time periods.