Abstract
Three-dimensional stratigraphic models provide a visual and quantitative description of complex geological structures and phenomena, which can support a variety of applications including urban subsurface space development and utilization, geological disaster prediction and evaluation, and more. In the field of three-dimensional stratigraphic modeling, multiple-point geostatistics (MPS) shows advantages in the automatic characterization of heterogeneous structures. Nevertheless, in many real-world situations, complex geological structures hardly meet the stationarity assumption of MPS. In this work, we present a novel three-dimensional stratigraphic modeling approach via an adaptive search strategy-based MPS simulation, named AS3DR. The adaptive search strategy is taken to capture precise local training images of the current node in MPS simulation. Moreover, a probability aggregation mechanism is proposed to fuse multiple statistics in different training images to a combined probability distribution. Various synthetic experiments and a real application are performed to verify the performance of AS3DR. The results show that AS3DR selects and fuses training images from adjacent cross sections to accurately capture and reproduce multiple-point patterns in the three-dimensional reconstruction of complex stratigraphic structures. Overall, our approach mitigates the negative influence of the strong non-stationarity of complex geological structures in MPS methods and extends the application scope of MPS to three-dimensional stratigraphic modeling.