Joint Inversion of Noise Horizontal-to-Vertical Spectral Ratio and Multimode Dispersion Curves: Method and Verification

Abstract Obtaining the S-wave velocity structure of geotechnical layers is critical for assessing local site effects during earthquakes. Although intrusive approaches provide accurate results, their high costs and logistical constraints limit widespread application. Nonintrusive approaches using surface collected seismic data are more practical but often suffer from significant non-uniqueness due to the inherent nonlinearity and multiparameter nature of geophysical inversion. This study addresses these challenges by introducing a joint inversion method that combines noise horizontal-to-vertical spectral ratios (NHV) and multimode Rayleigh-wave dispersion curves (MRDCs) with a genetic simulated annealing algorithm. Validation through three synthetic cases and field data from Mirandola, Italy, demonstrates that the joint inversion approach provides superior constraints on shallow and deep S-wave velocity structures compared to single-dataset inversions using only NHV or MRDCs, leveraging their complementary sensitivity kernels. At the Mirandola site, the method effectively mitigated errors from data computation and frequency band limitations, the latter referring to the narrower frequency range (e.g., 3–60 Hz) typically available in practical inversion compared to the broader range (5–100 Hz) often assumed in theoretical studies. This significantly improved inversion accuracy. These findings highlight the potential of joint inversion techniques to enhance the precision of seismic site-effect evaluations.