Reliable prediction of tunnelling-induced ground deformation is essential for assessing differential settlement of surface buildings and deformation of subsurface structures in densely built urban areas. This study establishes a multi-source engineering-evidence framework for parameter selection in the modified Peck model. An interval database containing 39 published correction-coefficient records is compiled to investigate the maximum-settlement correction coefficient α and settlement-trough-width correction coefficient β, while an independent dataset comprising 24 international shield-tunnelling cases and 84 surface and subsurface observations is used to evaluate depth-dependent settlement characteristics. Descriptive statistics, the Mann–Whitney U test, Cliff’s δ, bootstrap resampling, interval Monte Carlo simulation, and case-level trend analysis are employed. The median values α are 0.650 for soft soil and 0.550 for composite strata. Although the difference is not statistically significant (p=0.150,δ=0.302), interval Monte Carlo analysis gives a probability of 0.907 that the median α is higher in soft soil, indicating a directional ground-condition dependence. In contrast, β shows substantial between-group overlap and no robust ground-condition dependence. With increasing normalized observation depth z/T, i(z)/i(0) decreases strongly (ρ=-0.898), whereas S_max(z)/S_max(0) generally increases, with consistent within-case trends in 17/17 and 20/22 cases, respectively. The results support treating α as a ground-conditioned prior and β as an updateable project-specific parameter rather than prescribing either as a fixed constant. The proposed framework provides a transparent basis for settlement assessment of the built environment and highlights the need for depth-dependent corrections when subsurface structures are involved.