| 摘要: | 本研究利用非負解平滑化阻尼最小平方時間域解迴旋有系統地分析自1996年至2018年14個台灣地區中、大規模地震(Mw 5.5~7.5)的震源時間函數，並藉由消去法拆解震源時間函數探討其多重破裂特徵，藉此評估這些地震的震源參數，如地震矩(M0)、震源輻射能量(ES)、震源歷時(T)、靜應力降(s)等等。首先，為快速得到震源時間函數，本研究只選用垂直破裂方向的遠場P波，並在主震處製造無震源歷時的合成P波當成經驗格林函數，以避免主震處無參考地震的問題；為加速時間域解迴旋的運算，經由測試後，發現資料點數在400點內，阻尼値100時，可以使震源時間函數不至太過平滑，又能保留一些細節；為能得到最佳的震源時間函數，本研究也加入震源深度測試，以獲得地震的最佳深度，也可降低震源時間函數和震源深度的消長關係；加入AICc統計說明震源歷時時間長度的選取。結果顯示透過多重破裂分析所得的ES，較能找回較多的高頻能量；所估算的平均s也較由單一震源的評估來得合理；ES/M0比值與s呈現正相關，不僅存在每個地震的多重破裂過程，也存在於不同地震之間。台灣中、大規模地震的ES/M0比值在1-510-5，s在3-130 bars，平均約35-50 bars，這都符合一般全球的觀測。M0和T也呈現冪次關係，即M0~T3，其M0/T3比值接近全球的平均觀測，但比在墨西哥地區淺源地震觀測的結果為低，表示不同孕震地帶有著不同的M0/T3比值。整體而言，本研究的分析結果與前人研究或全球觀測大致吻合，顯示此方法在求解臺灣地區中、大規模地震的震源時間函數是穩定的。未來此方法將應用在區域的地震資料，以評估規模更小地震的震源時間函數。
In this study, we used teleseismic P-waves and damped least-squares time-domain deconvolution with smoothing and non-negative constraints to analyze systemically source time functions (STFs) for 14 moderate- and large-sized earthquakes (Mw 5.5-7.5) from 1996 to 2018. Afterwards, using elimination method to decompose the STF can further infer multiple-event features for each earthquake studied. From sub-events decomposed from the STF, source parameters, including seismic moment (M0), radiated seismic energy (ES), source duration (T), and static stress drop (S), were obtained. First, for rapidly estimate the STF, only stations perpendicular to the rupture direction were selected to complete the non-negative time-domain convolution with empirical Green’s functions (EGFs), which were synthetic teleseismic P-waves without source duration created at the position of main shock. Secondly, for accelerating the non-negative time-domain deconvolution, we used 400 data points and damping value of 100 to the deconvolution by testing. This made the STF not too smooth and retained some details in the STF. Thirdly, a depth testing was done to gain the optimal focal depth for each earthquake. This would decrease the trade-off between focal depth and source duration. Finally, we used a statistical method, AICc, to reasonable judge the time-length of source duration. Results showed high-frequency ES to be caught from the multiple-event analysis. In addition, the average S can be also calculated using multiple sources instead of a single source. The ratio ES/M0 was proportional to S not only among earthquakes but also during each earthquake rupture. For earthquakes in Taiwan, the ratio ES/M0 was 1-510-5 and S was 3-130 bars with the average of 35-50 bars, which were consistent with global observations. The relationship between M0 and T also correspond with a power law, M0~T3. The ratio M0/T3 was close to a value for global observations, but lower than that for earthquakes in Mexico. As a whole, the results from this study were consistent with previous studies or global observations. This also showed that the non-negative time-domain deconvolution developed from this study could stably estimate the STFs for moderate- and large-sized earthquakes in Taiwan. In future, we will use the method to regional data to estimate the STFs for small-sized earthquakes. |
