本論文採用中位數基準方法，分析1999~2024年期間，台灣與日本地區規模M≥5.0之地震引發全球電離層圖(Global Ionosphere Map， GIM)總電子含量(Total Electron Content， TEC)異常變化。透過單樣本Z檢定(one-sample Z-test)探討震前電離層前兆(Seismo-Ionospheric Precursors， SIPs)特徵，包括異常之正或負極性變化、提前天數、持續時段以及前兆發生的當地時間等。進一步利用接收者操作特徵曲線(Receiver Operating Characteristic， ROC)分析，驗證前述各項特徵之統計顯著性。結果顯示，臺灣地區規模M≥6.0地震發生之前1–2天TEC會異常減少，而日本地區M≥6.0地震發生之前1–4天TEC則會異常增加，在統計檢定支持下該時段之異常前兆與地震發生具有顯著關聯。計算SIPs出現與不同地震規模與震源深度之勝算(odds)，以探討其發生機率與地下構造特性之關聯性。研究結果發現勝算變化趨勢與區域性地質背景以及電性構造存在密切相關。;This study employs the median-base to statistically investigate seismo-ionospheric precursors (SIPs) in the Global Ionosphere Map (GIM) of the total electron content (TEC) associated with M ≥5.0 earthquakes occurring in Taiwan and Japan during 1999–2024. The one-sample Z-test is applied to identify the characteristics， including the polarity of positive or negative， lead days， duration， and the local time of SIPs. Receiver Operating Characteristic (ROC) curve is further used to confirm the statistical significance of the identified SIPs. Results show that， for M ≥ 6.0 earthquakes， SIPs are significant decreases (increases) in GIM TEC 1–2 (1–4) days before the earthquakes in Taiwan (Japan). Furthermore， odds are calculated to find the relationship between SIP occurrence and different earthquake magnitudes as well as focal depths， thereby exploring the dependence of SIP occurrence probability on subsurface structural characteristics. Results show that the trends in odds variations are closely associated with regional geological settings and subsurface electrical structures.
