Probabilistic seismic response analysis of tunnel linings considering coupled rock mass property and earthquake excitation uncertainties

• Random field theory is adopted to characterize the spatial variability of rock parameters. • Improved Clough–Penzien power spectrum model is used to generate stochastic nonstationary earthquake excitations. • The difference in mechanical behavior of lining between uncertainty conditions is compared. • The influence of key earthquake excitation parameters on the seismic performance is discussed. Tunnel lining seismic performance is significantly influenced by the spatial variability of geological parameters and the uncertainty of earthquake excitation factors, which are conventionally treated in isolation. This study proposes a novel probabilistic framework that integrates random field theory with an enhanced Clough–Penzien spectrum to concurrently model both uncertainty sources. The approach offers a more realistic and integrated assessment of seismic risk for tunnels under complex geological and loading conditions. The case analysis of a railway project reveals that considering both spatial variability of rock mass and uncertainty in seismic excitation leads to significant increases in internal forces and their variability, with mean values rising up to 278.9% and coefficients of variation (COV) up to 262.8%, compared to single-factor random analyses. The non-normal distribution of responses under seismic uncertainty, combined with the broader dispersion from rock variability, necessitates integrating both random factors for reliable seismic performance assessment of tunnels. Parametric studies demonstrate spectral parameters, including initial circular frequency ( ω 0 ), equivalent damping ratio ( ξ 0 ), and peak acceleration ( a max ), significantly influence results: increasing ω 0 and ξ 0 markedly reduces both the mean and COV of lining mechanical response-by up to 83.5% and 82.5%, respectively-potentially underestimating failure risk and underscoring the need to adopt lower-bound values in design for enhanced safety. Meanwhile, a max positively correlates with mean structural response, while variability in internal forces follows distinct trajectories; moreover, the interaction between rock spatial variability and seismic uncertainty raises failure probabilities by 3%–38%, emphasizing the necessity of integrating both randomness sources, especially in high-intensity seismic regions.