For the first time

an application of Multiple-point geostatistics in 3D internal architecture modeling of point bar is proposed in this paper.Data from dense well patterns (well spacing of 60 m) offer us the statistical distribution characteristics of interbeds within the point bar.Training images are produced by interactive processing of the statistical results for quantitative characterization of the interbeds

such as their thickness

dip

frequency

density and horizontal spacing.A typical point bar is first choosed for vertically projecting the interbeds interpreted from single well data to its top surface along the surfaces of interbeds

and then a 3D spinning cube is calculated in combination with the depositional patterns of the point bar for strike characterization of the interbeds.The Snesim algorithm is used for 3D architectural modeling of lateral accretion

and the results are compared with those from sequential indicator simulation (SIS).It is concluded that SIS may represent the characteristics of interbeds

but it is inadequate to characterize their continuity.Besides

a lack of constraints from training images forbids satisfactory quantitative characterization.In contrast

the model based on Snesim is under the dual control of training images and 3D spinning cube

thus it can quantitatively characterize the distribution

geometry and spatial structure of the interbeds.The data of watered-out reservoirs reveal that the architecture model built with multiple-point geostatistics method can guide the mapping of remaining oil.