A shallow sedimentary structure (10 to 15 m) of alluvial sediments over a shaly bedrock is investigated with ambient vibration arrays to derive the shear-wave velocity profile. Four arrays with distinct geometries are deployed on the site. Three array processing methods are used to derive either the dispersion curve or the auto-correlation curves. These curves are then inverted with the neighbourhood algorithm, which is well-suited to problems affected by strong non-uniqueness. Additionally, information from boreholes, classical refraction and active surface wave experiments are analysed to check the validity of the array results.
Of the three processing techniques, only the f-k method provided coherent dispersion curves for the three arrays and proved to be the more robust. The results provided by the high resolution technique globally agree with the first method but exhibit unexpected sharp variations of the dispersion curve at some frequencies. Finally, the auto-correlation technique was only usable for one array and appeared to be very sensitive to uncorrelated noise.
The refraction results provide a 
profile with its uncertainties. For the inversion of ambient vibrations,
profiles are uniformly chosen within this uncertainty interval. With only the vertical component, a good definition of 
down to 8 or 10 m is achieved. Below, the experimental uncertainties are too great to obtain a correct estimation of the velocity. The introduction of the H/V peak frequency as a supplementary constraint improves the final results by reducing the posterior uncertainty about the depth of the basement.
This example clearly demonstrates that any single method and any single array aperture is not valid for a reliable determination of the
profile. On the contrary, we used all available methods to find out the robust features and to discard the contradictory results. The use of the horizontal components would probably improve the determination of the deeper structure.