Shear velocity model appraisal in shallow surface wave inversion

Uncertainty of surface wave dispersion from numerical modeling and field repeatability tests shows a strong dependence on the spread length resolution limitation with frequency. High frequency uncertainty is about 1% and rises nonlinearly at low frequency to over 30%, where Gaussian distribution is invalid. The major source of dispersion error is additive noise to seismograms, static shifts and geophone placement, with geophone coupling and tilt effects being minimal. Monte Carlo ensemble analysis shows the half-space shear velocity to become unresolvable in the presence of realistic dispersion error. Shear velocity of a shallow LVL under a caprock is well resolved, but becomes poorly resolved when at depth underlying a buried HVL, as are the thicknesses of a LVL or HVL. Shear velocity 'smearing' between interfaces occurs at depth, where velocity gradients are more likely to be estimated instead of sharp contrasts, as well as an underestimate of half-space shear velocity. Gaussian error propagation is invalid, with shear velocity likelihood showing broad and/or bi-modal distribution. A 5 in thick, stiff layer at 20 in depth is found to be physically non-resolvable with dispersion measured from an array at the surface.