Sea State Estimation Based on Measurements from Multiple Observation Platforms

Ocean waves represent the main source of dynamic excitation of surface vessels and floating marine structures. The scarcity of wave data from vast areas of the world’s oceans is yet an ongoing problem. The crucial quantification and mitigation of uncertainties inherent to wave monitoring is increasingly recognised by the shipping, offshore, and renewable energy industries, as well as in coastal engineering.

Fusing data from multiple and heterogeneous observation platforms enables reducing the uncertainties of measurements from individual platforms. In this framework, the ship-as- a-wave-buoy concept shows high relevance, simply by accounting for the sheer number of ships in transit on the open sea. The wave-induced responses of modern vessels are often monitored by inexpensive off-the-shelf sensors, therefore constituting data that can be used for estimating sea states in near real-time, in a fundamentally similar way to traditional wave-rider buoys. In this thesis, the technical functionality and feasibility of a network of wave sensor systems, integrating vessels as “sailing wave buoys”, is investigated through several case studies, making use of synthetic data of vessel motions, sample results from model-scale experiments, and real-world data collected by seagoing vessels.

The first part of the PhD work addresses the data-driven estimation of vessel hydrodynamic models that describe the dynamics of wave-structure interactions. Such models are needed for the execution of the wave-buoy analogy. Two main methods are proposed; one is a tuning procedure to improve the fidelity of precomputed models in real operations, whereas the other is a parametric calibration method that allows a model identification when the knowledge of the hull geometry is limited to the vessel’s main dimensions.

In the second part, the hydrodynamic model calibration method is adapted for the estimation of the response amplitude operators of a wave-propelled unmanned surface vehicle. This study is motivated by the need to provide cost-efficient hydrodynamic models on an ad-hoc basis, for the purpose of enhancing directional wave spectrum estimates from this new kind of ocean observation platform. Moreover, a novel method is proposed to conveniently handle the Doppler shift of wave frequencies when the platform is moving forward in the seaway at relatively low speeds.

The third part of the work introduces new developments for collaborative sea state estimation methods, in which various in-situ observation platforms experiencing the same sea state are considered as parts of a single-input multi-output system. The problem of blind system identification is dealt with in a data fusion approach to simultaneously perform the estimation of the on-site wave spectrum and calibration of the hydrodynamic models. This idea develops the self-sufficiency of the network with auto-calibration capabilities, resulting in an improvement in the accuracy and precision of the local sea state estimates.

The fourth and final part is dedicated to the spatial estimation of wave data within a larger geographical domain, integrating the wave buoy analogy into a machine learning- driven framework for wave nowcasting. Surrogate models assimilating vessel and buoy observations are established to spatially map the sea states encountered across a whole region, at a fraction of the computational time normally needed for the execution of state- of-the-art physics-based wave models.

Overall, the PhD thesis demonstrates that vessel response-based estimates of the sea state, combined with measurements from other met-ocean sensors, can be integrated into observation networks to improve the reliability and availability of wave data. The work aims to improve the sampling and modelling of the ocean wave environment around the globe, thereby contributing to scientific understanding, as well as to the sustainability and operational safety of ocean-based human activities.