Broadband ground motion simulation is a vital research area in earthquake engineering, providing a critical basis for seismic hazard assessment and engineering design. Despite significant advancements in simulation techniques, existing deterministic physical models still face major limitations, particularly in accurately reproducing high-frequency ground motion characteristics. These shortcomings stem primarily from constraints in source model refinement and the substantial computational resources required for broadband, high-accuracy simulations. However, accurately capturing these high-frequency components is crucial for the seismic design of structures and for ensuring infrastructure safety during earthquakes.

To address these challenges, this study integrates the GP15.4 kinematic hybrid source model with a regional site model that incorporates surface topography within the spectral element method （SEM） framework. This integrated approach combines the broadband strengths of the GP15.4 model with the high accuracy and computational efficiency of SEM. The GP15.4 model achieves its broadband capability by superimposing high-frequency stochastic elements onto a low-frequency deterministic slip model. Specifically, the deterministic slip distribution is converted into the wavenumber domain via a two-dimensional Fourier transform. High-frequency stochastic components, conforming to a von Karman autocorrelation function spectrum, are then generated and superimposed. An inverse Fourier transform reconstructs a spatially heterogeneous slip distribution. Rupture time and rise time for each subfault are derived from this slip distribution, with consideration given to depth-dependent rupture velocity variations and the introduction of random perturbations to better simulate high-frequency radiation. The final rupture model is implemented in the SPECFEM3D spectral element code, enabling a complete simulation from fault rupture to seismic wave propagation.

To validate the proposed methodology, we first conducted a simulation for a region in western China using a flat-surface model. The results, compared against NGA-West2 ground motion prediction equations （GMPEs）, predominantly fell within their expected variability ranges. Subsequently, the model was applied to simulate the 2016 M_W6.5 Norcia, Italy earthquake. Comparisons of acceleration time histories and response spectra at ten observational stations demonstrated the method’s effectiveness and accuracy. These validations confirm that the GP15.4 model within the SEM framework significantly enhances the realistic simulation of ground motion characteristics that are challenging for traditional deterministic methods. Furthermore, wavefield snapshots provide visualization of seismic wave propagation from the fault to the surface, elucidating key features such as the concentration and rapid attenuation of near-fault motions and the effects of surface topography.

This study establishes a framework applicable to other seismically active regions worldwide. By combining an advanced hybrid source model with SEM, this approach ensures a comprehensive understanding of seismic wave propagation, making it a valuable tool for both theoretical research and practical engineering applications. With ongoing advancements in computational power, this method holds promise for delivering even more accurate simulations across broader frequency ranges and under more complex site conditions, thereby offering substantial support for future seismic risk assessment and resilient infrastructure design.