In recent years the 3D geological models are widely used in describing deep reservoirs compared with the conventional two dimensional modeling approaches. The main reason of such methodology is the better agreement of the actual reservoir with the three dimensional geological attributes particularly the petrophysical and lithological ones. Nowadays producing quantitative models for optimizing petroleum upstream activities is inevitable. It is very well known that traditional geostatistical methods based on variograms may fail to generate the complex, curvilinear, continuous and interconnected facies distributions that are often encountered in real geological media, due to their reliance on two-point statistics. Multiple Point Geostatistics (MPG) overcomes this constraint by using more complex point configurations whose statistics are retrieved from training images. Hence it is very important to employ professional geostatistical techniques such as multiple point geostatistics in simulating reservoir properties. Although well log data has a good vertical resolution in refining subsurface reservoir hetrogenities but due to their large horizontal separation are unable to discriminate lateral variations in comparable resolution. Therefore using well log data combined with 3d seismic cube as an aid in simulating 3d static reservoir properties would be highly advantageous for compensating lateral continuity. The main objective of this work is to apply multiple point geostatistical simulation of 3d reservoir facieses of Asmary reservoir in parts of one of south Iranian oil fields followed by comparing the results with the conventional methods of facies modeling such as sequential indicator simulation (SISIM). This reservoir is considered as shallow depth reservoir with high recovery capabilities among other carbonate and sandstone reservoirs. The modeling approach begins with developing a suitable method for wav elet extraction capable of tying the 3d seismic data to that of well log data. Then the model base and neural network approaches are used in inverting poststack seismic data in Hampson Rassell software environment. Results show that the Neural networks outperforms the model base approach in respecting geological realities (77% correlation against 74%). Therefore the final neural network model was selected as the secondary data in all geostatistical simulation methods used afterward. Then the reservoir facieses were simulated using MPG and Sequential Indicator Simulation (SISIM) using petrel 2011 software and the results were compared. Results showed that the multiple point geostatistical simulations could reconstruct the continuitity of heterogeneous subsurface structures owing to its better usage of all available information (including conceptual geological information) in the form of Training Images. Finally, the effective porosity was simulated using Sequential Gaussian Co-Simulation using 3d acoustic impedance as secondary data (Co-SGSIM). Through cross validation it is shown that the estimated porosity is in a very good agreement with the core sample and other log related data. As concluding remarks it is emphasized that the preprocessing steps outlined in this research work is required in order to get acceptable results. Also it is recommended to not only get more well log data to improve inverting seismic reflection but to use more attributes and logs as input variables in conditioning multiple point geostatistical approaches.