WP1 - Uncertainty classification and modelling approaches
Large offshore wind farm sites present a complex challenge for ground modelling due to their vast spatial extent, heterogeneous seabed and sub-surface conditions, and relatively limited availability of high-quality data. These constraints introduce significant uncertainty, making robust and integrated ground modelling essential for reliable site characterisation and design.
Ground modelling for offshore wind is an inherently interdisciplinary field that integrates geophysical, geological, and geotechnical data. Because it requires close collaboration across these domains, the initial step towards uncertainty quantification involves developing a clear understanding of the individual tasks within each discipline and how they interrelate.
The next step is to define the flow of information - that is, the input and output data associated with each task - together with the key sources of uncertainty, followed by a classification of the uncertainties according to a taxonomy suitable for quantitative modelling. This process will be undertaken by a multidisciplinary team of experts, coordinated by a data analyst to ensure coherence and consistency as well as practical relevance.
Once the main uncertainties are identified, we will summarise ranges of potantial modelling approaches and provide recommendations tailored to each task. These approaches encompass classical and Bayesian statistics, machine learning, deep learning and surrogate modelling among others. All modular recommendations will be aligned with the Bayesian data analysis framework for propagation of uncertainties for this strategic project. Previous work by Feng et al. (2025) has already demonstrated the value of Bayesian methods in deriving soil properties from CPT data (see Figure XZ).
Figure XZ: Examples of CPT-based Gmax predictions across the TNW site. Red curves show the predicted Gmax profiles; the red dashed curve is the best estimate. Additional red curves denote uncertainty ranges. Unit boundaries are indicated by black dashed lines.
Reference:
Feng Y., Bozorgzadeh N., Klinkvort R. (2025). Bayesian hierarchical modelling for high-resolution probabilistic characterisation of small strain shear modulus from offshore CPT. Computers and Geotechnics 188, 107584, ISSN 0266-352X, https://doi.org/10.1016/j.compgeo.2025.107584.