<p>With the ongoing expansion of underground road construction in numerous Chinese cities, the Artificial Ground Freezing (AGF) method used in subway tunnel construction frequently causes unavoidable deformation of surrounding structures and strata. In severe cases, such deformation may lead to accidents. Consequently, the accurate prediction of ground deformation during freezing construction is of considerable importance. This study integrates the development process of frozen walls with the three-dimensional mirror image method to establish a time-dependent prediction model for ground surface deformation during tunnel freezing construction. A dedicated calculation program was developed using mathematical software Maple. The accuracy of the proposed approach was assessed through an engineering case study, and its validity was verified by comparing with numerical simulation results and field monitoring data. The spatial distribution and temporal evolution of surface displacements during active freezing were obtained. The findings indicate that the distribution patterns of surface displacements predicted by different methods are broadly consistent with the measured values. The measured values lie between those obtained from the proposed method and the numerical simulation method, thereby confirming the validity of the proposed theoretical prediction model. For all prediction methods, the maximum vertical displacement occurs at the ground surface directly above the tunnel centre. However, the peak magnitudes of vertical displacement at this location differ markedly, and the ultimate surface deformation is influenced by frost heave.</p>