DETERMINISTIC SEA WAVE AND SHIP MOTION FORECASTING: FROM REMOTE WAVE SENSING TO PREDICTION ERROR ASSESSMENT

Presently, the assessment of the operational safety of ships and offshore structures is typically addressed within a statistical framework, both at design stage and for the specification of operational limits. Recently, however, the availability of new remote sensing technologies is paving the way for complementary approaches based on deterministic predictions of sea waves and ship motions. In this respect, the marine wave radar is considered as the key asset for the deterministic prediction of wave elevation. Indeed, the possibility of measuring the sea surface, almost instantaneously and for large areas, can be used to forecast the wave elevation at the location of the operating units. Eventually, the coupling of deterministic wave forecast with suitable ship motion models opens the possibility for giving anticipated prediction and guidance. The application of this emerging approach can be beneficial to those short-time offshore operations requiring sea wave or ship motions to be forecast in the time horizon of tens of seconds to minutes. This envisages the possibility of development of finely tuned early warning, hazard control and support decision systems. One of the main aspects of this chain of models, which is oftentimes overlooked, is the importance of providing the forecast with a consistent assessment of the prediction error. Moreover, the additional sources of uncertainty coming from the wave measurement and from the inversion of the wave radar images are also seldom accounted for. In this thesis, the whole chain of models, that from the wave radar measurement leads to the ship motion prediction, is investigated. The first step is the proposal of a novel technique for the inversion of wave radar images that can consistently account for those regions of the sea surface that cannot be uniformly illuminated because of the shadowing effect. The adoption of a linear least square fitting approach, provided with a regularization technique, allows the proposed inversion method the needed flexibility to address the shadowed regions as missing data. Afterwards, the assessment of the error associated with deterministic wave predictions is addressed. A novel semi-analytical procedure is proposed which allows estimating the ensemble variance of prediction error, in a simple and flexible way, naturally embedding the characteristics of the linear fitting and propagation procedures. The approach also allows the inclusion, in the estimation of prediction error, of the effect of measurement error coming from the radar inversion techniques. The same technique for the estimation of wave elevation prediction error is then extended to linear ship motions, using linear transfer functions. In fact, the developed framework can deal with any linear transformation of the wave elevation, resulting in the definition of a sound measure of the prediction error of linear responses. The method can be used to provide deterministic predictions with confidence intervals, as well as for a consistent setup of the whole forecasting chain. The developed models are then tested for a set of application examples considering both linear and nonlinear wave fields. In this latter respect, a high order spectral method has been implemented to provide more realistic wave elevation fields. Example applications regarding linear ship motions have also been carried out. The proposed inversion procedure has shown interesting results for synthetic radar images generated from both linear and moderately nonlinear wave fields. However, further investigations are needed to reduce the high computational cost required for the inversion. The proposed approach for wave and ship motion prediction error, instead, can represent a convenient novel sound method for the consistent setup of deterministic prediction procedures, remaining, however, limited to those scenarios where nonlinearities play a minor role.