How reliable are 1D models in reproducing the site seismic amplification functions?: a case study in the Val d’Agri Basin, Italy

Predicting ground motion over large areas seems to be the frontier in the enhancement of seismic hazard maps at the national level. Simplified approaches based on 1D numerical simulations have gained traction due to the limited availability of detailed geological and geotechnical data. The Italian Project PRIN-SERENA (mapping Seismic site Effects at REgional and National scAle) aims to improve ground motion amplification maps by deriving site-specific amplification functions (FAs) through numerical modelling and validating these models against experimental data to ensure reliability and practical applicability.

In the framework of WP6-SERENA project, we evaluated the suitability of these simplified methods in 21 seismic stations located an intermontane basin in the southern Apennines (Val d’Agri, southern Italy), characterised by high seismicicity and complex geology. The amplification functions (FA) were experimentally estimated by applying a nonparametric single-step generalized inversion (Klin et al., 2018) to a database of about 2000 waveforms from local and regional events (up to 400 km), providing robust azimuthal coverage for each station.

For the 21 sites experimental FAs were estimated for three period bands of engineering interest: 0.1 – 0.5 s, 0.4-0.8 s and 0.7-1.1 s. These estimates were compared with those derived from the 1D stochastic modelling results by using the NC92soil software (Acunzo et al., 2024) and the subsurface seismo-stratigraphic and mechanical models as input data. The modelled FAs exhibited a narrow amplification range (0.7-1.5), while experimental FAs showed a broader range (1-5). Sites that have the same stratigraphic and mechanical subsurface model have identical modelled FAs, while experimental FAs have very different values reflecting the actual geological complexity of the sites. This discrepancy is most evident at sites with thick surface soils (>100 m), where the 1D models were unable to fully account for subsurface conditions.

For the sites with pronounced discrepancies, deterministic modelling incorporating detailed subsurface characteristics and a higher shear velocity contrast between sediments and bedrock was performed. For sites with low sediment cover (<30 m), modelled FAs aligned closely with experimental results. At sites with medium sediment cover (up to 100 m), modelled FAs were generally lower than the observed values. For deep basin sites (depth > 200 m), deterministic modelling could not reproduce the experimental findings.

These results highlight the limitations of simplified 1D modelling, which primarily accounts for stratigraphic amplification but fails to capture the full complexity of the local seismic response, influenced by path and source effects, as the experimental approach does. Simplified numerical estimates of FAs risk underestimating or misrepresenting seismic site responses, particularly in geologically complex settings such as valleys and intermontane basins. This study underscores the importance of integrating experimental data into seismic hazard assessments to account for the complexity of local seismic responses.

 

Acunzo, G. et al. 2024. NC92Soil: A computer code for deterministic and stochastic 1D equivalent linear seismic site response analyses. Computers and Geotechnics, 165, p.105857.

Klin, P. et al. 2018. GITANES: A MATLAB package for estimation of site spectral amplification with the generalized inversion technique, Seismol. Res. Lett. 89, no. 1, 182–190.