| 摘要: | 震源尺度，即震源參數隨著地震大小變化之間的關係，從1970年代起直至目前仍持續地進行研究，除了探索震源參數隨地震大小是否存在自我相似尺度關係外，也建立了各種震源參數隨著地震大小變化之經驗關係式。一般而言，從地震工程的研究領域地震災害度分析上，不論定率法或是機率法，皆需要估算特定區域未來發生最大地震的潛力，而特地斷層所可能擁有的斷層參數大小則為最大地震災害度分析預估所必須採用之重要參數。此外，時間序列強地動的數值預估的研究上也需要一開始假設其可能之斷層尺度及滑移量。綜觀上述，震源尺度關係分析之研究成果不僅於地震學，且於工程地震學後續研究探討和應用皆為重要角色。 本研究專門除了收集並分析地震矩規模介於4.6至8.9地震之有限斷層錯動量分佈模型(slip model)，而包含斷層長(fault length)、斷層寬(fault width)及平均滑移量(mean slip)。這些震源參數可由波形逆推求得之有限斷層錯動量分佈模型上取得，接而進一步探討震源尺度關係及檢驗地震自我相似性。規模小於7以下地震主要為台灣碰撞造山帶地區地震，額外加入7個收集世界各地地震則讓整個尺度關係延伸至較大規模的地震。透過由強震波形資料或遠震波形資料所逆推得到的有限斷層錯動量分佈模型，由自我相關函數(autocorrelation function)推估其主要有效震源參數，包含有效斷層長，斷層寬及平均滑移量。本研究得到在震源參數對應與地震矩大小的尺度關係上，地震矩小於10^20Nm時有M0 ~ L^2的關係，而地震矩大於10^20Nm時則為M0 ~ L^3的關係，且這樣的關係在不同的震源機制得到類似結果。顯示小至中型地震不遵循假設應力降常數下自我相關尺度(self-similar scaling) 之震源尺度；反之，大地震卻仍遵循自我相關尺度。此外，進一步推估各個地震的應力降，顯示個別應力降存在明顯差異；然而，除了主要幾個中小型地震有著超過300 bar以上較大的應力降以外，其他地震之應力降主要介於10至100 bar之間。比較Shaw (2009)年所推導的地震規模與斷層面積關係式曲線，其中重要的孕震帶厚度(seismogenic thickness)因子假設35公里時，本研究所收集的資料落於10至100 bar的地震大致與預測曲線趨勢一致，而趨勢顯示第一個轉折於斷層面積1000平方公里，顯示在孕震厚度35公里之下，台灣地區最大可能的斷層破裂面積大小可達此值，而破裂超出這樣的斷層面積以上。 存在明顯較高應力降的地震發生位置主要皆為座落於台灣西部麓山帶的盲斷層(blind fault)，這些地震所紀錄到最大地表加速度(peak ground acceleration，PGA)於先前研究考慮之強地動衰減式所預估的預測值比較，明顯地存在有較高的PGA。透過經驗格林函數法推估強地動產生區域範圍(strong motion generation area，SMGA)重新估算應力降明顯較高之地震，得到其地震之強地動產生區域較小，亦即錯動量範圍較小。採用一個規模類似的地震同樣求取強地動產生區域則有著較大的面積，而不同的面積存在同樣的地震矩造成兩個地震應力降存在超過200 bar這樣一定的差異，間接證明了特定地震有著明顯較高應力降的情形的確存在。而儘管相對較小規模的地震，較高的PGA可能會對鄰近震源區域之建築及人民造成直接的損毀及災害，因此，本研究建議應力降的因子也應該被考量在強地動推估之研究上。 Source scaling, the relation between source parameters earthquake size, has been explored the self-similarity relation or developed a series of empirical relationships from the beginning of 1970s. In general, an important requirement for probabilistic and deterministic analyses in seismic hazard of engineering seismology is to estimate a future earthquake potential in a specific region. The fault parameters (e.g. fault length, width and mean slip) generated by a particular fault or earthquake source related to the size of the largest earthquakes is often necessary. In addition, pre-setting source dimensions and slip over the fault are also necessary for numerical prediction of time history ground motion. This kind of studies, thus, not only focuses on scientific purposes for seismology but provides the implications and applications for engineering seismology. Source parameters, including fault length, width and mean slip can be dug out from the finite-fault slip distribution inferred from waveform inversion. In this study, thus, we sorted and solved slip distribution models resolved from the period from 1993 to 2009 in Taiwan. We investigated the source scaling of earthquakes (Mw4.6~Mw7.7) from Taiwan orogenic belt, and made the global compilation of source parameter to discuss the scaling self-similarity. Finite-fault slip models (13 dip-slip and 7 strike-slip) using mainly from Taiwan dense strong motion and teleseismic data were utilized. Seven additional earthquakes (M>7) were included for further scaling discussion on large events. Considering the definitive effective length and width for the scaling study, we found M0~L^2 and M0~L^3 for the events less and larger than the seismic moment of 10^20 Nm, respectively, regardless the fault types, suggesting a non-self similar scaling for small to moderate events and a self-similar scaling for large events. Although the events showed the variation in stress drops, except three events with high stress drops, most of the events had the stress drops of 10-100 bars. The bilinear relation was well explained by the derived magnitude-area equation of Shaw (2009) while we considered only the events with the stress drops of 10-100 bars and the seismogenic thickness of 35 km. The bilinear feature of the regressed magnitude-area scaling appears at the ruptured area of about 1000 km^2, for our seismogenic thickness of 35 km. For the events having ruptured area larger than that, the amount of the average slip becomes proportional to the ruptured length. The distinct high stress drops events from blind faults in the western foothill of Taiwan yield local high Peak Ground Acceleration (PGA) as we made the comparison to the Next Generation Attenuation (NGA) model. Further, we implemented strong motion waveform modeling with the empirical Green’s function method to assess the area of strong-motion generation for inspecting the existence of fact that two earthquakes with similar magnitude have significantly distinct stress drop. Two earthquakes with similar magnitude have ~180 and ~610 bars, respectively, indeed, suggesting that significant difference of stress drop was proved. Regardless the relative small in magnitudes of these events, the high PGA of these events will give the high regional seismic hazard potential, and, thus, required special attention for seismic hazard mitigation. |
