Physics-based and data-driven investigations into rapid wave formations in oceans

Abstract

Extreme waves, also known as “rogue waves,” can be observed in a range of media including oceans, microwave cavities, and optical fibers. In ocean scenarios, these are monstrous waves, that are significantly larger than the preceding and subsequent waves, appearing out of nowhere. These waves can have devastating impacts on ships and offshore oil rigs. The motivation behind this dissertation stems from efforts to better understand the conditions leading to such sudden, extreme energy localizations in oceans. In the past, different mechanisms leading up to the formation of extreme ocean waves have been theorized. These mechanisms have paved the way for wave tank experiments in controlled conditions for validation of the different proposed mechanisms. However, appropriate scalability can be a significant limitation for the results obtained from such experimental efforts. Numerical modeling, as an alternative, is a highly feasible avenue to accurately understand and observe different rapid wave formations in practical scenarios. Extending this to a more practical front, it would be beneficial to develop a systematic forecasting network of extreme waves through a well-developed data-driven strategy. This can be done by leveraging information on different sea state parameters. The aim of this dissertation is to effectively develop a more informed understanding of the processes leading to rapid wave formations in oceans through wave modeling efforts and prepare a framework for developing an extensive and flexible rogue wave prediction network by using a data-driven outlook. To that end, in this dissertation, a Lagrangian N-particle numerical scheme is applied to develop a numerical wave tank. Furthermore, the aim of optimizing the numerical scheme for rapid wave simulations through investigation of kernel functions and numerical dissipation schemes is pursued. The developed numerical scheme is utilized for wave tank simulations to understand rapid wave formations in different scenarios, with a special focus on understanding behavior of sidebands in modulated plane waves in intermediate and deep sea waters. Wave modeling with substantial data reserves in data repositories of wave monitoring and navigation systems is leveraged in conjunction with artificial intelligence (AI) principles. In particular, machine learning processes are used to develop a wave prediction strategy with a special focus on imputing of missing values in wave monitoring records. In addition, machine learning and deep learning principles are applied to forecast occurrence of rogue waves over different horizons using ocean buoy data. The work presented in this dissertation can form a basis for developing prediction and forecasting schemes for rapid wave formations in oceans.