English：

Due to the sparsity of borehole data in the early stage of geotechnical survey, the stratum modelof the engineering area needs a large amount of borehole data to construct. The MCMC model can processsparse survey borehole data and has the advantages of simple parameters and quantifiable formationuncertainty, so it is widely used in geological modeling and geological uncertainty analysis. However, thetraditional Markov chain assumes that the transition probability of the stratum is constant throughout thespace⁃time range. This uniformity assumption has great limitations in stratum simulation, and MonteCarlo is relatively rigid in simulating the discontinuity and variability of stratum sequences. Therefore, astratum simulation method based on local Markov chain⁃adaptive importance sampling ( LMC⁃AIS) isproposed. This method uses the unified depth interval fragmentation processing of boreholes in the studyarea to construct the stratum transition probability matrix, randomly simulate the stratum state, and multi⁃fragment superposition to establish a stratum model suitable for complex uneven strata. The uncertainty ofthe geological model is quantitatively evaluated by using the variance of the stratigraphic sequencegenerated based on the LMC⁃AIS model. The results indicate that compared with the traditional method,this method not only considers the inhomogeneity of formation change but also quantifies the uncertainty information prediction.