[Purposes] The randomness of karst cave distribution and development in karst areas, coupled with the inherent spatial variability of geotechnical parameters, renders it challenging to predict the impacts of tunnel construction in such environments. [Methods] Taking the Batang pipe jacking project for water intake in Xinhua County, Loudi City, Hunan Province as a case study, this paper predicted karst cave locations by using geophysical prospecting data with small data spacing and sparse borehole data according to the Markov random field theory and constructed a stratigraphic type model. On the basis of this model, the geotechnical parameter random fields were then generated through the Karhunen-Loève transform and imported into the FLAC3D numerical model by using the non-invasive finite difference method. The FLAC3D numerical model was employed to explore the law of surface deformation induced by pipe jacking construction in karst areas. Several sets of geotechnical parameters were used to investigate the influence of spatial variability of geotechnical parameters on the surface deformation law by using stochastic analysis. [Findings] The results indicate that based on the Markov random field theory, the effective prediction of karst cave distribution can be realized with borehole and geophysical prospecting data. Compared with the traditional numerical model analysis of homogeneous strata, the stochastic analysis considering the spatial variability of geotechnical parameters by using the random field theory can provide confidence intervals of surface deformation. The random field parameters mainly influence the size of confidence intervals of surface deformation, but have no influence on the type of surface deformation. [Conclusions] Studying the influence of pipe jacking construction in karst areas with the help of the random field theory can guide actual construction processes, which is of some significance for the construction of projects similar to tunnels.