Construction of average S-wave velocity maps using simplified microtremor method: Case study of area with pyroclastic flow deposits to the east of Aso caldera, Japan

Summary In this study, we compare the usability of a simplified microtremor-based empirical method and a conventional microtremor method based on an inversion analysis of a subsurface velocity structure model for constructing a map of average S-wave velocity (AVS) values. In the simplified (empirical) method, the phase velocities of Rayleigh waves, which can be obtained by processing a microtremor array, at wavelengths of 13, 25, and 40 m are regarded as AVS values from the ground surface to depths of 10, 20, and 30 m (Vs̄ ₁₀,Vs̄ ₂₀,\ and\ Vs̄ ₃₀), respectively. Microtremor array surveys were conducted at 173 observation points within a 15 km × 17 km area east of Aso caldera, Kyushu, Japan (target area). AVS values are obtained by applying the empirical method to the phase velocities obtained at each observation point. The AVS values at an observation point (located near the centre of the target area) with velocity logging data are verified by a comparison with those based on the velocity logging data (i.e. overestimations by 6 per cent at maximum). It is found that for the entire target area, the spatial distribution of the obtained AVS values is consistent with the geological distribution. The AVS values within areas of the Aso-3 ignimbrite are 30–40 per cent larger than those within areas of thick soil and tephra on the strongly consolidated Aso-4 ignimbrite. In addition, the AVS values of the Aso-3 deposits are more than 10 per cent larger than those of the Aso-4 deposits and about 10 per cent smaller than those of geological units older than the Aso-3 deposits. We also apply a conventional (i.e. inversion) method to the phase velocity data at each observation point to obtain a one-dimensional S-wave velocity (Vs) structure model from which we deduce AVS values. The deduced AVS values at the velocity logging point are underestimated by -8 per cent, with differences from the AVS values obtained using the empirical method reaching 13 per cent. The average systematic difference between the two methods is 15 per cent, as determined from a statistical analysis. None the less, a strong correlation is found between the methods, with an average correlation coefficient of 0.94, with no evidence showing that either method is more accurate. The empirical method can be used to construct an AVS map if overestimation is carefully considered. This analysis also reveals that the average maximum survey depths of the one-dimensional Vs structures based on the inversion method are only 23±10 m, making them often insufficient to map Vs̄ ₂₀ and Vs̄ ₃₀ (the ratios of the available to total numbers of data points are only 60 and 21 per cent, respectively). In contrast, the empirical method can determine Vs̄ ₁₀,Vs̄ ₂₀,\ \ and\ Vs̄ ₃₀ at more than 80 per cent of all sites. The construction of AVS maps using the empirical method is effective in terms of the simplicity and reliability of planning, observational efficiency, and simplicity of data processing, which support a practical and objective approach to seismic assessments.