Gaussian process (GP) models provide a flexible probabilistic framework for characterizing spatial variability in soils from limited and heterogeneous measurements. By enabling probabilistic spatial interpolation and the fusion of multiple direct and indirect data sources, GPs offer a powerful approach for inferring subsurface conditions while explicitly quantifying uncertainty. The resulting uncertainty can be propagated naturally into probabilistic and reliability assessments of geotechnical systems.

This lecture introduces the fundamental concepts underlying GP-based modelling and illustrates their application to probabilistic site characterization for offshore wind farms. Geological information and cone penetration test (CPT) data are jointly incorporated to infer soil stratification and cone penetration resistance across a three-dimensional seabed. Categorical stratigraphic data are encoded as Gaussian variables via a Dirichlet transform, enabling their integration within the GP framework. To address the high spatial dimensionality and large data volumes typical of offshore sites, computational scalability is achieved through approximate variational inference using inducing points, which enforce sparsity in the spatial correlation structure and significantly accelerate inference. Stratigraphic information is further embedded in the prediction of cone penetration resistance by training independent GP models for each soil type, enabling effective identifcation of soil-type-specific correlation lengths. Finally, the resulting predictive distributions are propagated through a reliability analysis of monopile foundations, demonstrating how GP-based site characterization can directly support reliability-based monopile design.

Iason Papaioannou is a Privatdozent and Deputy Group Leader of the Engineering Risk Analysis Group at the Technical University of Munich (TUM). He has been appointed Assistant Professor at the National Technical University of Athens (NTUA), where he will take up the position in April 2026. His research interests include random fields and spatial variability, reliability assessment, data driven uncertainty quantification, uncertainty-based sensitivity analysis, Bayesian methods and probabilistic machine learning. He is interested in applications from a wide range of engineering fields, including structural and geotechnical engineering, hydrogeology and material science. Dr. Papaioannou studied civil engineering at NTUA and computational mechanics at TUM and received his doctoral degree from TUM in 2012. He received the GEOSNet Young Researcher Award 2015 for his research on geotechnical safety and reliability and the Early Achievement Award 2022 of the International Association for Structural Safety and Reliability (IASSAR). He is an editorial board member of Georisk, Structural Safety and GEOAI. He serves as a member of the ISSMGE Technical Committee (TC304) on Engineering Practice of Risk Assessment and Management, and the Working Group (WG2.15) on Geotechnical Reliability Methods of the German Geotechnical Society.