Full text loading...
Creating quantitative 3D ground models for offshore wind sites relies on predicting Cone Penetration Test (CPT) and soil parameters from geological, geophysical, and geotechnical data using a combination of computationally intensive machine learning and geostatistical models, and simpler statistical methods. While simple models serve as a fast and interpretable baseline, they may fail to capture lateral soil variability or the non-linear effects of stress. To overcome these limitations, we propose a simplified geostatistical approach for CPT parameter prediction utilising stress-normalised parameters, as successfully applied at the IJmuiden Ver Gamma Wind Farm site. The primary benefits of the model are speed, stability, and interpretability, allowing for rapid model updates and validation of the interpreted geological structure itself. Within the broader ground modelling workflow, this approach acts as a powerful complementary tool, generating a robust, low-frequency background model, while machine learning techniques capture more subtle details and complex relationships between geophysical and geotechnical data.