Research Project “On Certain Ground”: Uncertainty Quantification in Ground Modelling for Offshore Wind

On Certain Ground pioneers a Bayesian statistical framework for subsurface characterization, connecting statistical analyses, machine learning (ML), and physics-based models to quantify and propagate uncertainties from data acquisition to engineering design. Central to this work is the development of a comprehensive uncertainty taxonomy that systematically categorizes uncertainties across data types and modelling stages, enabling robust fusion of underexploited datasets - high-resolution seismic data, laboratory and field geotechnical measurements, and expert geological insights - to inform prior models for seismic inversion and other physics-based analyses. Faced with large, complex offshore datasets, the industry looks towards machine learning, despite reliability concerns. By embedding modern uncertainty-aware methodologies, this project unlocks ML’s potential to accelerate computationally intensive and tedious tasks (e.g., predictive models, seismic inversion) while advancing fundamental understanding of soil behavior through interpretable correlations. These advancements accelerate site characterization, enhance decision-making for renewable energy infrastructure, and establish a foundation for reliability-based geotechnical design.