During the early stage of evaluation of the gas field development with less well data，the established sedimentary facies model has great uncertainty only based on well data information，which is difficult to meet the accuracy requirements of geological research. Thus it is becoming more and more important to be constrained by the multi-seismic attributes for sedimentary facies modeling. Firstly under the model of Bayes，the local prior probability distributions was built，and then the probability was updated into the posterior distributions with the neighboring well data and seismic attributes data. During the updating，Markov assumption could be used to solve the problem of instability of cross matrix for different variables，and at last the sedimentary facies could be modeled through random sampling from posterior distributions of modeling points. Taking the example of the Su10 area in the north of Sulige gas field，the relationship between seismic attributes and sedimentary facies was analyzed firstly and four types of seismic attributes were chosen including RMS amplitude，average instantaneous frequency，effective bandwidth and attenuation. Considering the problem encountered in the computation of multi-variable cross-covariance，the fusion of multi-seismic attributes was calculated to obtain their fusion probability based on Markov-Bayes model，by which a sedimentary facies model could be built. The result demonstrates that simulation result keeps an over 80 percent agreement with a sedimentary facies model drawn in hand. The prediction error is considered to be within 1% through the analysis of cross-validation.