Accurate prediction of strong ground motion is essential for post-earthquake emergency response and seismic risk mitigation, particularly in tectonically complex regions such as the Sichuan–Yunnan area of southwest China. In physics-based ground motion simulations, subsurface structure is represented by crustal velocity models ranging from simplified one-dimensional to more detailed three-dimensional descriptions, yet the extent to which model choice influences engineering-relevant shaking predictions has remained unclear.
In a new study published in Science China Earth Sciences, researchers from Southern University of Science and Technology investigated how different velocity models affect strong ground motion simulations using the 5 September 2022 Mw 6.6 Luding earthquake as a real-world example. The team employed a finite-fault rupture model derived from seismic observations and conducted numerical simulations up to 1 Hz using nine representative velocity models that differ in data sources, resolution, and construction strategy.
Rather than focusing on waveform similarity or travel-time misfits, the study evaluated model performance using peak ground velocity (PGV), a parameter widely used in earthquake engineering and damage assessment. PGV predictions were analyzed at individual stations, across regional statistical distributions, and through spatial patterns of shaking intensity to assess the practical applicability of each velocity model.
The results show that most three-dimensional velocity models can reasonably reproduce observed PGV at frequencies below 0.3 Hz, while simulation errors increase at higher frequencies. Overall, 3D velocity models outperform traditional one-dimensional models in predicting both the amplitude and spatial distribution of ground shaking. However, systematic differences persist among individual models, with some tending to overestimate shaking intensity and others to underestimate it. These discrepancies are primarily linked to differences in shallow velocity structure, modeling strategy, and whether surface topography is included.
Importantly, the researchers found that averaging PGV predictions from multiple velocity models significantly reduces systematic bias and yields results that are more consistent with observations. This multi-model approach provides a robust and practical strategy for rapid post-earthquake ground motion assessment, especially when uncertainties in subsurface structure cannot be avoided.
The study offers quantitative guidance for selecting and applying velocity models in strong ground motion simulations and highlights the value of multi-model strategies for improving the reliability of earthquake shaking predictions in complex tectonic regions.
See the article:
Li T, Zhang W. 2026. Assessment of the impact of different 3D crustal velocity models on strong ground motion simulations in the Sichuan-Yunnan region. Science China Earth Sciences, 69(1): 348–365, https://doi.org/10.1007/s11430-025-1743-3
