Effect of lateral variation and model parameterization on surface wave dispersion inversion to estimate the average shallow structure in the Paraná Basin

Abstract

The average layered structure of the intracratonic Parana Basin, SE Brazil, is investigated with surface-wave group velocities from a small regional earthquake recorded by two broadband stations. Rayleigh and Love waves in the period range 1-4.2 s are used to infer average properties down to about 4 km. Genetic algorithm techniques are used to find the best fitting 1-D S-wave model. The inverted 1-D models show fair correlation with the average properties of the propagation paths as expected from geology and borehole information. However, different S-wave velocity models are obtained for the different inversion parameterizations. Since lateral heterogeneities are expected along the paths, several synthetic tests are performed with heterogeneous propagation paths. For approximately homogenous path (i.e., little lateral variation), the main features of the average synthetic model can be retrieved for different model parameterizations. For strong lateral variations, however, the average dispersion curve can produce very different 1-D inverted models depending on the parameterization. Also, the 1-D inverted models may differ significantly from the average properties of the inhomogeneous path, and wrong depths to interfaces may be inferred. For real data inversions, it is then suggested that various different parameterizations should be tested. If the resulting models show consistent features, this probably indicates homogeneity in the propagation path. But, if very different and unstable features are obtained in the 1-D inversions, then strong lateral variation may be present in the propagation path, and the average 1-D model may not represent average properties along the path.