Weighted F-K for surface wave analysis

Tests based on surface waves (SWM) consist of three steps: acquisition, processing and inversion. The acquisition is the observation of Rayleigh wave propagation in time and space: the vertical velocity at the free surface is usually recorded in two or more points at known distances from the source. To assess the dispersion characteristics on a wide frequency range, the source has to produce either a single broad-band signal or a series of monochromatic signals over a wide frequency band. The processing extracts from field records the information about Rayleigh waves dispersion. The inversion estimates site mechanical properties as a function of depth from the experimental dispersion characteristics. The amount of information extracted by the processing and its uncertainty, which strongly depend on the acquisition parameters, affect the reliability of the final result. Several processing techniques can be used; the basic principle is the same and some of them are formally identical. Nevertheless their different sensitivity to coherent and random noise leads to peculiar quality assessment and filtering capabilities. I compare here different processing techniques for dispersion curve extraction, and propose a new algorithm based on f-k processing and coherence between signal pairs. The procedure reduces the effects of noise by discarding low coherence signals and allows a great increase of the useful frequency band. Moreover non-uniformly spaced arrays can be processed and poor quality signals are filtered out automatically.