Project Supervisors
- Dr Suzana Ilic, Lancaster University
- Prof Aneta Stefanovska, Lancaster University
- Mr Michael Thomas, Reliable Insights
- Dr Bryan M. Williams, Lancaster University
Application Deadline
29 January 2026
Overview and Background
Rogue waves, exceptionally high ocean waves, whose height exceeds twice the significant wave height, are rare, short-lived events that pose serious risks to shipping, fishing, and maritime infrastructure, including offshore platforms and wind turbines. Understanding their formation and improving prediction are essential for safe marine operations.
Despite advances in theoretical and experimental studies, the physical mechanisms driving rogue wave formation in real seas remain poorly understood, making prediction challenging. This PhD project aims to address these gaps by analysing extensive field data, developing advanced non-linear dynamic techniques, and utilising high-performance computing. The aim is to improve understanding of rogue wave dynamics and enhance forecasting capabilities, thereby contributing to the safety and resilience of marine operations.
Methodology and Objectives
The PhD project will address the following questions: How can data processing and modelling be accelerated to enable non-linear analysis and modelling with higher spatial and temporal resolution? Which of the sea state parameters predicted by existing operational wave models are useful for detecting the formation of rogue waves? How do the formation and predictability of rogue waves depend on physical conditions?
Teaser 1
This data-intensive project aims to accelerate novel time-localised analysis methods to investigate physical mechanisms underlying rogue waves and predict their occurrence.
- O1: Exploit GPU Accelerated Computing to parallelise algorithms for time-localised phase coherence and couplings between waves recorded in many spatial points, enabling scaling to higher-resolution and near real time analysis.
- O2: Isolate the mechanisms leading to the formation of rogue waves using algorithms developed in O1.
- O3: Develop in-situ feature detection for automated analyses exploiting GPU and assess the relationship between the occurrence of rogue waves and their characteristics from time-series measured under different physical conditions.
- O4: Develop a time-series-based prediction modelling approach, using the relationships identified in O2-3 and assess its ability to predict the occurrence of rogue waves.
Methods
The numerical modelling and algorithms for time-series analysis will exploit GPU Accelerated Computing; exascale will then allow near real-time practical applications. The Multiscale Oscillatory Dynamics Analysis (MODA) toolbox for non-linear and time-localised phenomena in time-series (e.g. phase coherence, coupling and wave energy exchange [3&4]) will be parallelised and used to identify rogue wave mechanisms. Automated pattern analysis and feature engineering will be applied using Tangent to detect anomalous sea surface elevations and enable an easily deployable computationally light forecasting solution using processed data. The methods will be first applied to laboratory data (e.g. [1]) and then to publicly available field measurements (e.g. Free Ocean Wave Dataset with more than 1.4 billion wave measurements). The newly developed prediction modelling approach will be systematically validated with measured data.
Teaser 2
This is a data-intensive project focused on the computational optimisation of time series analyses for dynamic systems and the relationship between rogue wave properties and environmental conditions.
- O1: Assess the current performance of the numerical tools included in MODA and Tangent in terms of their relevance for detecting the mechanisms of rogue waves and their computational efficiency.
- O2: Optimise the algorithms of the tools identified in O1 with multiple GPU to improve time to improve computation time and experimental throughput, enabling large-scale ensemble time-series analyses.
- O3: Develop and apply a GPU version of MODA to field measured data to isolate mechanisms that lead to the formation of rogue waves.
- O4: Assess the relationship between the occurrence of rogue waves and concurrent ocean and atmospheric data.
Methods
The Multiscale Oscillatory Dynamics Analysis (MODA) toolbox offers several high-order methods for time-series analysis, some based on wavelets. The high computational demands of uncertainty evaluation methods limits their use for operational purposes. Optimised algorithms, GPU-acceleration and Exascale facilities will enable higher resolution and practical applications. MODA will identify the mechanisms underlying rogue wave formation using field measured time-series of surface elevations (e.g. Free Ocean Wave Dataset). The concurrent environmental data (e.g. surface ocean currents, wind and atmospheric pressure) will be collated either from field measurements or from the operational forecast models provided by meteorological offices. The correlation between the occurrence of rogue waves and environmental parameters, as well as ‘causal’ relationships between the identified mechanisms and the environmental conditions, will be investigated using the Tangent modelling engine which can be incorporated into predictions in the future.
References & Further Reading
- Luxmoore, J.F., Ilic, S. and Mori, N., 2019. On kurtosis and extreme waves in crossing directional seas: a laboratory experiment. Journal of Fluid Mechanics, 876, pp.792-817.
- Mori N., Waseda, T., Chabchoub A.(eds.) (2023) Science and Engineering of Freak Waves, Elsevier (https://doi.org/10.1016/C2021-0-01205-0).
- Newman, J., Pidde, A. and Stefanovska, A., 2021. Defining the wavelet bispectrum. Applied and Computational Harmonic Analysis, 51, pp.171-224.
- Stankovski, T., Pereira, T., McClintock, P.V. and Stefanovska, A., 2017. Coupling functions: universal insights into dynamical interaction mechanisms. Reviews of Modern Physics, 89(4), p.045001.
- Yang X., Rahmani H., Black S., Williams B. M. Weakly supervised co-training with swapping assignments for semantic segmentation. In European Conference on Computer Vision 2025 (pp. 459-478). Springer, Cham.
- Jiang Z., Rahmani H., Black S., Williams B. M. A probabilistic attention model with occlusion-aware texture regression for 3D hand reconstruction from a single RGB image. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2023 (pp. 758-767).
- Jiang Z., Rahmani H., Angelov P., Black S., Williams B. M. Graph-context attention networks for size-varied deep graph matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022 (pp. 2343-2352).