Leveraging world's largest offshore observation networks and physics-based model integration with machine learning for real-time tsunami and storm surge forecasting along the Pacific coasts of Japan

拟议研究计划的第一部分在第一年实施。开发了一种利用机器学习和S网络观测系统观测数据实时预报海啸淹没的方法，并成功地应用于过去的几次海啸事件。这种新方法的主要优点是计算速度，这在海啸预警系统中是至关重要的。相比之下，常规方法通过数值模拟预测指定研究区域的海啸淹没大约需要30分钟。相比之下，在精度相当的情况下，所提出的方法只需不到一秒钟。这一成果已发表在包括国内和国际会议在内的一份科学期刊上。在拟议研究的第二部分，类似的方法将应用于风暴潮预报，但相对于第一部分有所改进。

The first part of the proposed research plan was implemented in the first year. A method to forecast tsunami inundation in real-time using machine learning and data observed at the S-net observing systems has been developed and successfully applied to several past tsunami events. The main advantage of this new method is related to the computational speed, which is crucial in the tsunami early warning system. In comparison, the conventional methods through a numerical simulation take approximately 30 minutes to predict tsunami inundation in the specified study area. In contrast, with comparable accuracy, the proposed method requires less than a second. The results have been published in a scientific journal, including domestic and international conferences. In the second part of the proposed research, similar methods but with improvements relative to the first part will be applied for storm surge forecasting.