Multi-point geological statistics is one of the hot spots in current study

and the creation of training image is a key part determining the accuracy and reliability of reservoir simulation.Based on the shallow high frequency seismic data

the morphology and quantitative aspects of shallow channel deposits are studied and used as prototype models for further investigation of morphology and scale statistics of channel sand body-ies in deeper oil reservoiors.Quantitative three-dimensional training image is established on the basis of high-precision seismic inversion data.Training image grids are selected and Petrel is applied to analyze its impact on multi-point geological statistics.Results show that simulation is consistent with the well point data

and the spatial morphology and sizes of different channels are constrained by the quantitative characteristics of training image.Simulation results can best reveal the geometric characteristics and spatial confi-guration of sedimentary facies when the grid size of 3-D training image is consistent with the actual grid of simulation model.It provides a new method to create the training image

and may serve as a reference example for the simulation of other types of deep-water sedimentary facies.