The development of offshore wind farms necessitates precise subsurface characterization to mitigate foundation design challenges posed by paleo-channel structures. Site investigation is primarily based on grids of 2D seismic lines limiting the ability to capture 3D subsurface complexity. This study employs Multiple-Point Statistics using the direct sampling approach to assess the probability distribution of paleo-channels away from 2D seismic lines, considering geological interpretation as conditional data. Two training images are created from interpreted lines to simulate channel distribution under varying line spacings (150 m, 500 m, and 1000 m), with training images and conditional data from the same and different areas. Results indicate that a line spacing of 150 m yields high accuracy in replicating known channel distributions. Wider spacings (500 m) showed diminished accuracy but maintained general channel system integrity, while 1000 m spacing led to significant connectivity loss and unreliable predictions, particularly when training images and conditional data were derived from different regions. The findings highlight the critical role of seismic line density and the appropriate selection of training images in enhancing the reliability of paleo-channel predictions. This work underscores the importance of integrating advanced statistical methods for effective subsurface assessment in renewable energy development.
Siemann et al. (Wed,) studied this question.