GNGTS – Atti del 23° Convegno Nazionale / 14.02 G. Dal Moro, M. Pipan, P. Gabrielli, E. Forte and M. Sugan Department of Geological Environmental and Marine Sciences, University of Trieste RAYLEIGH WAVE DISPERSION CURVE INVERSION VIA GENETIC ALGORITHMS AND POSTERIOR PROBABILITY DENSITY ESTIMATION Surface wave dispersion has been shown to be a valuable tool for seismic investigation of subsurface structures (Park et. al., 1999; Xia et al., 1999; Dal Moro et al., 2004). Dedicated inversion procedures are highly recommended, due to the complexity of the problem, but have been generally scarcely considered. The multi-modal and highly non-linear nature of surface wave dispersion curve inversion represents a prohibitive problem for common linear inversion procedures based on the Jacobian matrix because of failures due to local minima. Fig. 1 highlights such aspect through the data regarding a 6-layer synthetic model. Various V S and thickness values were adopted for the second and third layers while the values for the other layers were set to their proper values and the rms values (objective function, ObjFN) between the actual and calculated dispersion curves plotted versus the V S and thickness ratios. Fig. 1 - Multimodality of the surface wave dispersion curve inversion problem. In abscissas the ratios between the shear-wave velocities (VS) of two overlaying strata and their thicknesses (THK), in ordinate the objective function. Arrow indicates the position of the real model.