| 摘要: | 九二一大地震後，經檢視全省磁力連續觀測資料發現，位於車籠埔斷層兩端的磁力站—鯉魚潭站及灣丘站，都觀測到磁力異常訊號，這些異常訊號與地震的發生有時間上的巧合。一般認為，當岩層受到應力作用而產生裂隙，地層內所含的帶磁礦物會散發在空間中，造成地磁場的變化。本計畫除了協助更新三分量地磁連續觀測網的觀測儀器外，同時也分析新建置的全省三分量地磁連續觀測網資料，以了解地磁場強度改變與地震活動的關聯性。 台灣有密集的地磁觀測網與相對頻發的大地震，透過全島地磁場觀測網的緊密監測，我們期望有機會為臺灣再造一個全球地震預測研究的獨特角色，提升臺灣在國際地震預報、預測之地位。然而，儀器升級/變更導致的物理量型態及數量之改變、乃至於測站位置之變更，為難以避免之情事。例如正在進行的以三分量地磁場紀錄儀替換舊有的全磁場紀錄儀計畫將陸續新增測站以及升級舊有儀器，以提供更高品質的地磁場資料。在傳統的TIP (Time of Increased Probability)預報方法上，輸入變數數量與來源型式須固定否則參數模型需重新訓練。也就是說，三分量地磁場必須被轉換為全磁場才能延續舊有模式進行預報，而這麼做卻使更豐富的資訊被浪費。另一方面，若要改用三分量地磁場進行模型最佳化，則需等待額外一段訓練期(例如7年)才能開始進行預報。地下動力系統的變化涵蓋之時間尺度通常是巨大的，因此長時間累積的觀測資料具有寶貴的價值。為了能充分發揮新硬體設備所提供之內涵豐富的紀錄，同時不浪費珍貴的舊測站資料，我們預計發展多變量的TIP預報模式。本年度計畫將發展多變量的MagTIP系統，預計將支援三分量地磁紀錄輸入、全場－三分量混合訓練與預報、多重日統計量之計算，且容許更多時間序列(例如地磁導出量或地磁以外的物理量)參與訓練與預報。此系統預計將為通用多變量TIP系統之原型，為未來發展整合地磁與其他可能與地震相關的時間序列用於TIP預報模式之第一步。 本計畫旨在利用臺灣地區電離層全電子含量(Total Electron Content, TEC)分析電離層地震前兆訊號。過去統計研究發現，在大地震(規模大於等於5以上)發生之前一個星期之內，電離層電漿濃度有出現異常減少/增加的情況，因此根據這種特性，我們有機會利用電離層異常出現來作為將來地震發生可能之參考。然而，電離層電漿濃度的變化受到很多因素影響，主要有來自於太陽活動(磁暴、日冕噴發物質)影響電離層高層，以及大氣潮汐作用影響電離層低層。在進行電離層異常天判定時，如果能夠去掉上述主要影響電離層電漿變化的因素，將有很大的機會補捉到來自於地震前地殼作用的影響，以增加電離層地震前兆的判定準確度。本計畫將利用上一期計畫中所建立之臺灣上空長期間(共25年)電離層全電子含量資料與電離層異常天之判定結果，並利用太陽活動性指標(Dst index)以及大氣潮汐分解方法去除電離層高層與低層的影響，嘗試尋找電離層地震前兆。 在過去數十年的歲月中，台灣的地震學家及相關學者，致力於收集這五項前兆的資料。資料。對於這些觀測資料，蔡義本先生和他的共同研究者曾發表了四篇論文(Tsai et al., 1983, 2004, 2006, 2018)。第一篇文章僅回顧中央研究院地球科學研究所成員在1983年以前所做的初步前兆研究成果。第二、三篇文章則以中央大學所執行的教育部和國科會共同支持之iSETP前兆研究計劃的成果為主要內容。第四篇文章除了前述的結果外，並包含台灣大學所執行之地震化學前兆的成果。雖然如此，仍然有許多其他單位學者的前兆研究成果並未包含在這四篇文章中。此外，對於個別前兆的研究，也有幾位作者完成回顧性文章:劉等人(Liu et al., 2000)的規模大於6之台灣地震的電離層前兆研究；劉等人(Liu et al., 2004)的規模大於5之台灣地震的電離層之foF2前兆研究；劉等人(Liu et al., 2006)的規模大於5的台灣地震之地磁異常前兆研究；陳等人(Chen et al., 2004)之台灣地震的地磁異常前兆的研究；陳等人(Chen et al., 2013)的規模大於6之台灣地震的地下水水位異常前兆的研究；傅和李(Fu and Lee, 2018)的台灣地震前之化學前兆研究。除了蔡等人針對一九九九年集集大地震有較多項前兆的整合性研究外，其他的文章則不涉及單一地震的不同前兆之整合性研究，而且也較不討論這些可能前兆的物理和化學的原理及可靠性。要將這些研究成果用於實際的預測地震上則較為困難。因此，必須將這些研究（包含不列在這些回顧或多個地震單項前兆之研究文章)進行整合性的分析，根據已有的物理和化學原理，檢驗各項前兆之可靠性，並比較不同前兆的相關性及在震前的時間發生序列。資料本身是不會預測地震，而根據可靠的資料，建立合理可靠的物理模型，然後再利用模型來預測地震。 ;After Chi-Chi earthquake, numerous magnetic anomaly signals were observed in the northern and southern parts of the Chelungpu fault, and it is coincided with the time of Chi-Chi earthquake occurred. It is generally believed that the magnetic minerals contained in the formation will be dispersed in space, causing changes in the geomagnetic field. This project includes not only the re-establishment of the geomagnetic field observation network, and the replacement of observation instruments, but also analyzing geomagnetic observations for earthquake occurrences and changes in tectonic stress. Taiwan locates on a region where earthquakes frequently occur. With the dense geomagnetic observation network covering the entire island, we hope to recast Taiwan a unique role in the study of earthquake prediction. However, the changes in the type or number of the observed physical quantities due to instrumental upgrades are inevitable. For example, in the undergoing project to replace the old one-component geomagnetic sensors by the three-component ones, locations of some station may be altered, new stations will be added, and the number of channels in recording is going to be tripled. In the traditional TIP (Time of Increased Probability) forecasting, the type and the number of input data have to be uniquely formatted otherwise re-training according to the new format is required. As a result, either the conversion for fitting new data to old format (where some information must lose), or an extra waiting period (e.g. 7 years) for re-training models with new data, is required. The time scale of the evolution of the underground dynamical system is generally thought to be large, hence data collected from old stations are precious. For being able to fully utilizing both the data from the most modern instruments and older sensors, the project of this year aims for the multivariate MagTIP forecasting system. We aim for a multivariate MagTIP forecasting system that not only supports three-component and one-component geomagnetic signals simultaneously, but also allows additional earthquake-relevant time series to be involved in calculating TIP. The main purpose of this study is to analyze the seismic precursor signals by employing the ionospheric total electron content (TEC) around the Taiwan area. Previous statistical studies found that before the bigger earthquakes (magnitude larger than 5), the ionospheric anomaly usually occur by the decrease/increase of plasma density. This might be seen as an index of the seismic precursor. However, the variations of ionospheric plasma density will be caused by many effects, such as the solar activity (geomagnetic storm and corona mass ejection) from the upper ionosphere and the atmospheric tidal effect from the lower ionosphere. If these effects can be canceled in the data, it is possible let us to catch the seismic activity before the earthquake and determine the ionospheric anomaly day in more precisely. In this study, the long-term, totally 25-year, ionospheric TEC observations and the results of ionospheric anomalous days developed by the last term project, as well as the solar activity index (Dst index) data and the atmospheric tidal decomposition method will be employed to find the seismic precursor in the ionosphere. In past few decades, Taiwan’s earth scientists collected many data of different earthquake precursors. Later Prof. Y.B. Tsai and his co-authors published four review articles (Tsai et al., 1983, 2004, 2006, 2018). In the first paper, they reviewed the preliminary studies of precursors done by the colleagues of Institute of Earth Science, Academia Sinica before 1983. In the second and third papers, they reviewed the studies of precursors done by colleagues of National Central University under the iSTEP Research Project sponsored by Ministry of Education and National Science Council. In addition to the studies shown in the second and third ones, enclosed in the fourth paper are the studies of chemical precursors done by colleagues of National Taiwan University. In addition, there are a few articles for the studies of a single precursor for numerous earthquakes. Liu et al. (2000) studied the anomalies of ionospheric foF2 for M?5 earthquakes; Liu et al. (2004) studied the anomalies of ionospheric total electron content for M?6 earthquakes; Liu et al. (2006) studied seismo-geomagnetic anomalies for M?5 earthquakes; Chen et al. (2004) studied geomagnetic anomalies for M?5 earthquakes; Chen et al. (2013) studied groundwater level changes for M?6 earthquakes; Fu and Lee (2018) reviewed geochemical precursors. Except for the articles done by Tsai and his co-authors, other articles are not the multidisciplinary studies of a single event and do not consider the correlations of a precursor used in an article to other precursors. Moreover, the physical and chemical mechanism to generate the precursors are not taken into account in these articles. In practice, it is difficult to directly apply those given studies of precursors to predict an impending earthquake. First, it is necessary to analyze given results (including those not shown in the above-mentioned articles) to test reliability of individual precursor based on some known physical and chemical models. Secondly, we must find out the correlations among the precursors. Data of precursors themselves cannot be used to predict an earthquake. We must construct a prediction model for Taiwan’s earthquakes based on these data. |
