<p>Sandy cobble soil exhibits pronounced heterogeneity. The assessment of the uncertainty surrounding its properties is crucial for the analysis of settlement characteristics resulting from volume loss during shield tunnelling. In this study, a series of probabilistic analyses of surface and subsurface settlements was conducted considering the spatial variability of the friction angle and reference stiffness modulus, under different volumetric block proportions (<i>P</i>_v) and tunnel volume loss rates (<i>η</i>_t). The non-intrusive random finite difference method was used to investigate the probabilistic characteristics of maximum surface settlement, width of subsurface settlement trough, maximum subsurface settlement, and subsurface soil volume loss rate through Monte Carlo simulations. Additionally, a comparison between stochastic and deterministic analysis results is presented to underscore the significance of probabilistic analysis. Parametric analyses were subsequently conducted to investigate the impacts of the key input parameters in random fields on the settlement characteristics. The results indicate that scenarios with higher <i>P</i>_v or greater <i>η</i>_t result in a higher dispersion of stochastic analysis results. Neglecting the spatial variability of soil properties and relying solely on the mean values of material parameters for deterministic analysis may result in an underestimation of surface and subsurface settlements. From a probabilistic perspective, deterministic analysis alone may prove inadequate in accurately capturing the volumetric deformation mode of the soil above the tunnel crown, potentially affecting the prediction of subsurface settlement.</p>