Excitability of Ocean Rogue Waves – Numerical Prediction and Early Warning by combining Wave Physics, Numerical Simulation and Data Driven Methods

对海洋上的异常巨浪进行短期预测被广泛认为是近海作业和航运的一个重要目标。目前发表或使用的途径和方法都不能令人满意。目前的预测时间在一分钟或更短。可能造成这种情况的最重要原因是无法将与振幅相关的色散效应适当地整合到计算和模拟环境中，而计算和模拟环境的数值效率足以具有实时能力。此外，从数据源中提取的初始条件中的小误差在预测过程中也会迅速增长，这对于考虑通常是混沌甚至弱湍流的波浪动力学是不可避免的。目前的技术水平远远低于我们所需要的水平，也远远低于我们在非线性波动物理方面的知识所表明的可能的水平，后者的范围在5到15分钟之间，甚至更长。两个关键的障碍仍然需要克服：首先，从可用的波态测量数据中，必须为后续的预测过程确定初始或初始边界条件。目前，这一识别过程是过程链中最薄弱的环节。其次，如果要实现实时能力，预测过程本身需要达到更高的质量和数值效率。目前，对流氓海浪物理特性的认识、对流氓海浪短期发生的数值模拟预测、对流氓海浪数据驱动方法的研究等三个相关研究领域在很大程度上是脱节的。我们建议整合这三个领域的知识和方法，开发新的混合方法。其目标是在特定混沌海况的理论可预测性范围内预测海况演变和异常浪事件的发生。该提案旨在研究上述三个领域的优化组合，以推动人们如何实现更接近理论上可能的可预测性的预测边界。我们努力通过开发波浪物理，先进的数值模拟和数据驱动方法的结合来克服目前的局限性。我们的目标是达到5到10分钟的预测时间。这意味着在期望的时间范围内，预测的幅度和相位误差具有足够的精度。目标是在预测期内保持幅度和相位误差在+/- 10%的误差范围内。我们希望这种方法能够在以后的许多领域得到广泛的应用，如海况保持、船舶路线、船舶或平台危险区域撤离预警、海上风力发电厂建设、海洋能源装置等。

Achieving short-term predictions of rogue waves on the ocean is widely considered an important objective for offshore operations and shipping. At present none of the approaches and methods published or used today can be considered satisfying. Prediction times at present are on the order of a minute or less. Probably the most important reason for this is the inability to integrate the wave amplitude dependent dispersion effects properly into a computating and simulation environment that is numerically efficient enough to have real-time capability. Furthermore, also small errors in the initial conditions extracted from the data sources grow quickly during the prediction, which is unavoidable with a view to the usually chaotic or even weakly turbulent wave dynamics under consideration. The present state-of-the-art is thus far below what is needed, and it is far below what our knowledge on nonlinear wave physics suggests is possible, which ranges in the order of five to fifteen minutes, or even beyond. Two key obstacles still need to be overcome: First, from the available wave-state measurement data, initial or initial-boundary conditions have to be identified for the subsequent prediction process. At present this identification process forms the weakest part in the process chain. Second, the prediction processes themselves need to reach a substantially higher quality and numerical efficiency, if real-time capability is intended. At present three related research fields of (i) understanding the physics of rogue ocean waves, (ii) predicting the short-term occurrence of rogue ocean waves by numerical simulation, and (iii) employing data driven methods in the context of rogue ocean waves, are largely disconnected. We propose to integrate the three fields of knowledge and methods to develop novel hybrid approaches. The vision is to predict the sea state evolution and the occurrence of rogue wave events up to the limit given by the theoretical horizon of predictability of the specific chaotic sea state. This proposal aims at studying optimised combinations of all three fields mentioned to push the boundary of how one can achieve predictions closer to the theoretically possible horizon of predictability. We strive to overcome the present limitations by a combination of exploiting wave physics, advanced numerical simulation, and data driven methods. It is our aim to reach prediction times of 5 to 10 minutes. This implies a sufficient accuracy in terms of amplitude and phase error of the prediction within the desired time frame. The objective is to keep both amplitude and phase errors within an error bound of +/- 10 % within the prediction period. We expect such a method to subsequently find wide applicability in many fields, like sea-keeping, ship routing, early warning to evacuate dangerous zones on ships or platforms, construction of offshore wind plants, ocean energy devices, etc.