地震規模頻率分布 (b值) 被視為是大地震前兆現象研究之一,台灣是年輕島弧造山帶,板塊碰撞活躍,已有不少規模大於6.0地震發生。中央氣象局地震觀測網 (CWBSN) 的地震偵測能力自2012年起大幅提升,台灣島內區域的最小完整規模 (MC) 下降至1.5,有益於初探台灣地區時間域b值在大地震發生時間點前後的連續變化。近年來全世界已有幾個大地震案例觀察到震前b值下降,本研究欲瞭解台灣地區大地震發生之前是否可觀察到類似現象,同時著重前人研究並未探討的台灣地區震後b值變化。由於b值計算結果高度取決於空間上地震分布特性,本研究採用CWBSN 2012年起重定位地震資料庫,以最大曲率法和最大似然法計算深度≤ 40公里地震的MC和b值特性,再以相同搜尋半徑和移動時間窗法針對大地震震央區分別計算時間域b值變化。移動時間窗法以每時間單位時間窗內固定地震事件數量方式,求取時間域b值穩定解並分析不同移動時間窗的重疊程度對b值解之影響。本研究結果顯示CWBSN地震偵測能力提升後,台灣島內區域b值主要介於0.7至1.1之間,與前人研究觀察相似,而ML ≥ 6.0地震幾乎位於b值低於0.9的區域。移動時間窗法結果顯示以震央取半徑20公里適合做為研究台灣地區ML ≥ 6.0地震於時間域b值變化分析的空間參數。時間窗內以大於MC以上地震數量為50筆事件為基準,時間窗的重疊程度取75%可提供穩定的時間域b值解析能力。本研究發現分析的17個ML ≥ 6.0地震中僅有2018 ML 6.3花蓮地震能觀察到明顯震前b值下降,而17個大地震的震後b值結果缺乏一致性時序變化,推測可能與餘震或震後滑移分布有關。本研究結果可作為未來台灣地區以b值研究大地震前兆現象的基礎。
b-value is regarded as a subject of precursor studies of large earthquakes. Taiwan island is a young orogenic belt where the plates collide actively and lead to numerous earthquakes with magnitude greater than ML 6.0. Earthquake detection capability of the Central Weather Bureau Seismic Network (CWBSN) has been largely improved since 2012. The magnitude completeness (MC) of the inland regions is around 1.5. This improvement can help us preliminarily survey the continuously temporal b-value variation before and after the large earthquakes in Taiwan. In recent years around the world, there are several pieces of evidence of the temporal b-value decrease before large earthquakes. This study aims to find out whether similar phenomenon can be observed in Taiwan. This study also puts the emphasis on temporal b-value variation after large earthquakes in Taiwan region, which is not discussed by previous studies. Since the b-value calculation highly depends on the characteristic of earthquake distribution, this study uses a relocated earthquake catalog with earthquakes recorded by CWBSN since 2012. We use the maximum curvature and maximum likelihood to calculate MC and b-value distribution, respectively, from the earthquakes with depths shallower than 40 km. We calculate temporal b-value variations in the epicentral areas of large earthquakes using moving windows with a constant radius. The moving windows require sufficient number of events in each window for the robust b-value estimation while overlapping with various time periods. The influence of the window’s overlapping on temporal b-value variation has also been analyzed. Our results show that the inland b-values are mainly between 0.7 to 1.1, which is similar to previous studies. The ML ≥ 6.0 earthquakes are almost located at the low b-value regions smaller than 0.9. The results of moving windows show that a 20-km radius from the epicenters is an appropriate spatial parameter for studying the temporal b-value variation pre- and post- ML ≥ 6.0 earthquakes in Taiwan region. The moving windows overlapping with 75% event numbers in which the number of earthquakes requires at least 50 with magnitude greater than MC can provide high-resolution temporal b-value variation. This study found that there’s a clear temporal b-value decrease before the 2018 ML 6.3 Hualien earthquake. The temporal b-value variarions after the ML ≥ 6.0 earthquakes do not vary consistently, which may be related to the localized aftershocks and afterslips. Our results can provide a fundamental framework of b-value earthquake precursors of the coming large earthquakes in Taiwan.