自我校正之地震定位系統

Abstract

地震預警系統需要迅速的定位出發震位置來提供有效且即時性的警報來降低地震造成的傷害與損失。被廣泛應用之分層速度模型雖方便執行，卻難以精準描繪地震波實際傳導之情形，進而影響地震定位之準確度。 許多相關研究致力於加速地震定位之過程，提高定位之準確度，以及更精地去模擬地震波的傳遞過程。卻少有研究哲著墨於地震歷史資料之應用。 本文提出一個利用直接搜尋法的自我校正定位系統。該系統可根據測站的歷史資料校正其震波到達時間。校正系統對地震波在傳遞的過程中所經歷的系統性誤差建立模型，並根據歷史紀錄對於P波到達時間進行校時。為了進一步提高校正之精準度，本文提出了一個測站篩選流程，濾除不穩定之測站來提高定位的品質。本文所提出的校正系統會隨著其所參與的地震事件而演進，增加其校正之穩定性。 實驗結果指出，單純透過測站篩選可提高定位之準確度。P波到時之校正因需考慮面相較多而表現並不穩定。在部分結果中校正P波到達時間可大幅提高定位準確度，反之在其他結果中，卻有出現準確度降低之影響。

An earthquake early warning system has to rapidly detect the onset of an earthquake to reduce the damages done by the quake. However, the accuracy of earthquake location is often limited by the simplicity of the velocity model. Numerous researches have been done to shorten the computing times of hypocenter searching processes, raise prediction accuracy, and better simulate seismic wave propagation. However, few have utilized historical event data in practice. This thesis proposed an earthquake locator that calibrates station trigger times based on their historical behavior. The systematic uncertainties in seismic wave propagation are modeled in the form of trigger offsets. P wave arrival times are adjusted according to the offset obtained from historical records. A station selection method was introduced to further increase calibration accuracy. The proposed calibration scheme would develop with its experience to earthquake events, increasing the stability of its calibration accuracy. Experiment results indicate a constant improvement in predication accuracy when the station selection technique is solely applied. The effect of offset cancelling is not stable, it has proven to be effective in certain cases, but deteriorates prediction accuracy in others.