The stationary Markov process model is widely used to predict the geological conditions in tunnel excavation projects. However, the validity of the stationary assumption made in the model is questionable. The prediction error caused by the assumption has not been investigated in previous studies. In this study, the significance of a stationary Markovian zone detection is evaluated by comparing the predicted geological conditions with the real soil layer distributions in boreholes. A new method is proposed to detect the stationary Markovian zones in a tunnel-covered area. Borehole data from Perth, Australia are collected to illustrate the significance of the stationary Markovian zone detection and the effectiveness of the proposed method. The results show that the stationary assumption leads to considerable errors in the predicted number, location, and thickness of soil layers. The proposed method is robust with respect to the starting point of a detection.