Abstract: The existing spatial random field theories can better depict the uncertainty of a single geotechnical parameter, including the intrinsic randomness, global variability and local singularity. The current challenge is how to analyze the correlated variables in multiple soil layers for soil-tunneling mechanics. The coefficient of cohesion, internal friction angle and compression modulus are considered in the reliability index analysis of ground surface settlement due to excavation of shallow-buried tunnels. At first, the geotechnical parameters of multiple non-stationary soil layers are converted into unified anisotropic spatial random field using the local detrending method, and the statistical parameters are defined. Then, the uncertainty is simplified into aleatory randomness, and the response surface method is introduced into the reliability index analysis. The co-sequential Gaussian discretization is designed for the spatial random field and numerical analysis. Finally, the failure probability (i.e., reliability index) of ground surface settlement is calculated stochastically by using the classical and subset Monte-Carlo simulations, respectively. This approach is applied into the four shield tunnel paralleling zones of the 5^th and 6^th metro lines between West Huanhu Road station and West Binguan Road station in Tianjin of China. The results show that the reliability index considering geotechnical parameters as the classically probabilistic variables is smaller than that of spatial random field assumption. It may support substantially the construction control and design optimization in complex shallow-buried shield tunneling projects.