In order to clarify ultra-low-frequency (ULF) seismomagnetic phenomena, a sensitive geomagnetic network was installed in Kanto, Japan since 2000. In previous studies, we have verified the correlation between ULF magnetic anomalies and local sizeable earthquakes. In this study, we use Molchan’s error diagram to evaluate the potential earthquake precursory information in the magnetic data recorded in Kanto, Japan during 2000–2010. We introduce the probability gain (PG′) and the probability difference (D′) to quantify the forecasting performance and to explore the optimal prediction parameters for a given ULF magnetic station. The results show that the earthquake predictions based on magnetic anomalies are significantly better than random guesses, indicating the magnetic data contain potential useful precursory information. Further investigations suggest that the prediction performance depends on the choices of the distance (R) and size of the target earthquake events (Es). Optimal R and Es are about (100 km, 108.75) and (180 km, 108.75) for Seikoshi (SKS) station in Izu and Kiyosumi (KYS) station in Boso, respectively.

中文翻译：

---

评估 2000-2010 年日本关东记录的超低频磁场数据中潜在的地震前兆信息：距离和震级相关性

为了澄清超低频（ULF）地震磁现象，日本关东自 2000 年起安装了敏感地磁网络。在之前的研究中，我们已经验证了 ULF 磁异常与当地大地震之间的相关性。在这项研究中，我们使用 Molchan 的误差图来评估 2000-2010 年日本关东记录的磁数据中潜在的地震前兆信息。我们引入概率增益（PG'）和概率差（D'）来量化预测性能并探索给定超低频磁站的最佳预测参数。结果表明，基于磁异常的地震预测显着优于随机猜测，表明磁数据包含潜在的有用前兆信息。进一步的调查表明，预测性能取决于目标地震事件的距离（R）和大小（Es）的选择。伊豆的 Seikoshi (SKS) 站和房总的 Kiyosumi (KYS) 站的最佳 R 和 Es 分别约为 (100 km, 108.75) 和 (180 km, 108.75)。