Dimensionless parametric analysis and statistical characterization of 2D seismic aggravation factors in sedimentary basins

Sedimentary basins can amplify earthquake ground motion beyond levels predicted by one-dimensional (1D) site response analysis. This two-dimensional (2D) excess, quantified through aggravation factors (AGFs), is absent from current seismic design codes such as Eurocode 8, which rely exclusively on 1D schemes. A database of 13,266 2D finite element simulations and 252,054 complementary 1D analyses were generated using OpenSeesPy for trapezoidal basins with a flat floor, covering 67 symmetric and asymmetric configurations, 9 stratigraphic profiles, and 22 spectrum-compatible seismic input motions. The database covers deposit thicknesses of 50–200 m and is restricted to coarse-grained granular materials; this may overestimate hysteretic dissipation and underestimate AGFs for basins with significant fine-grained layers, and results outside the explored thickness range lie beyond the validated scope of the study. Seventeen governing variables were organized into 14 physically motivated dimensionless parameters spanning basin geometry, soil and bedrock properties, and seismic input characteristics. Statistical dependencies were assessed using Kendall rank-correlation analysis across three levels of spatial resolution and four spectral period bands; as a methodological limitation, only monotonic dependencies are detected, and non-monotonic parameter–AGF relationships may appear as negligible even when physically meaningful. The width-normalized position emerges as the dominant descriptor of AGF variability, followed by wavelength-normalized geometric parameters, resonance proximity, and impedance contrast, while basin shape ratio exhibits the strongest positive association with amplification severity. These findings provide a statistically robust and physically interpretable characterization of the parameter space governing 2D basin aggravation, constituting a step toward code-compatible predictive models.