The 2016 Fukushima earthquake (M 7.4) generated a moderate tsunami, which was recorded by the offshore pressure gauges of the Seafloor Observation Network for Earthquakes and Tsunamis (S‐net). We used 28 S‐net pressure gauge records for tsunami data assimilation and forecasted the tsunami waveforms at four tide gauges on the Sanriku coast. The S‐net raw records were processed using two different methods. In the first method, we removed the tidal components by polynomial fitting and applied a low‐pass filter. In the second method, we used a real‐time tsunami detection algorithm based on ensemble empirical mode decomposition to extract the tsunami signals, imitating real‐time operations for tsunami early warning. The forecast accuracy scores of the two detection methods are 60% and 74%, respectively, for a time window of 35 min, but they improve to 89% and 94% if we neglect the stations with imperfect modeling or insufficient offshore observations. Hence, the tsunami data assimilation approach can be put into practice with the help of the real‐time tsunami detection algorithm.

中文翻译：

---

2016 年福岛地震 S-Net 压力计记录的实时海啸数据同化

2016 年福岛地震 (M 7.4) 产生了中度海啸，由海底地震和海啸观测网络 (S-net) 的海上压力计记录到。我们使用 28 个 S-net 压力计记录进行海啸数据同化，并预测了三陆海岸四个潮汐计的海啸波形。S-net 原始记录使用两种不同的方法进行处理。在第一种方法中，我们通过多项式拟合去除潮汐成分并应用低通滤波器。在第二种方法中，我们使用基于集合经验模式分解的实时海啸检测算法来提取海啸信号，模拟海啸预警的实时操作。两种检测方法的预测准确率得分分别为 60% 和 74%，时间窗口为 35 分钟，但如果我们忽略建模不完善或近海观测不足的台站，它们会提高到 89% 和 94%。因此，可以借助实时海啸检测算法将海啸数据同化方法付诸实践。