To continuously monitor crustal behavior associated with earthquakes, magmatic activities and other environmental effects (e.g., tides and rain precipitation), we have developed a continuous monitoring system of seismic velocity of the Japanese Islands. The system includes four main processing procedures to obtain spatio-temporal velocity changes: (1) preparing ambient-noise data; (2) creating virtual seismograms between pairs of seismometer stations by applying seismic interferometry; (3) estimating temporal velocity variations from virtual seismograms by stretching interpolation approach, and (4) mapping spatio-temporal velocity variations. We developed a data-processing scheme that removes unstable stretching interpolation results by using the median absolute deviation technique and a median filter. To map velocity changes with high stability and high temporal resolution during long-term (i.e., longer-term monitoring), we proposed the “sliding reference method”. We also developed evaluation method to select the optimum parameters related to stability and temporal resolution. To reduce computation time for continuous monitoring, we applied parallel computation methods, such as shared memory and hybrid distributed memory parallelization. Using our efficient and stable monitoring system, we succeeded to continuously monitor the spatio-temporal velocity variation of the whole Japanese Islands using ambient-noise data from 767 seismometers. Finally, we developed a web application that displays spatio-temporal velocity changes. In the monitoring results that we open through the website, we identified velocity variation (e.g., pore pressure variation) that could be related to earthquake, aftershock, magmatic activities and environmental effects in a stable manner.

中文翻译：

---

基于时空地震速度变化的日本列岛地壳实时监测系统

为了持续监测与地震、岩浆活动和其他环境影响（例如潮汐和降雨）相关的地壳行为，我们开发了日本列岛地震速度的连续监测系统。该系统包括四个主要处理程序以获得时空速度变化：（1）准备环境噪声数据；(2) 应用地震干涉技术在地震台对之间创建虚拟地震图；(3) 通过拉伸插值法从虚拟地震图估计时间速度变化，以及 (4) 映射时空速度变化。我们开发了一种数据处理方案，通过使用中值绝对偏差技术和中值滤波器去除不稳定的拉伸插值结果。为了在长期（即长期监测）期间以高稳定性和高时间分辨率绘制速度变化图，我们提出了“滑动参考法”。我们还开发了评估方法来选择与稳定性和时间分辨率相关的最佳参数。为了减少连续监控的计算时间，我们应用了并行计算方法，例如共享内存和混合分布式内存并行化。使用我们高效稳定的监测系统，我们成功地使用来自 767 个地震仪的环境噪声数据连续监测了整个日本列岛的时空速度变化。最后，我们开发了一个显示时空速度变化的 Web 应用程序。在我们通过网站打开的监测结果中，我们确定了速度变化（例如，