受惠於台灣擁有密集的地震測站且建立完整場址參數，本研究採用無參考站方法（Non-reference Site Approach， NSA），針對北台灣台北、宜蘭及桃園地區建立各測站之非遍歷性（Non-ergodic）線性場址放大函數，包括虛擬譜加速度（Pseudo-Spectral Acceleration， PSA）及傅氏振幅譜（Fourier Amplitude Spectrum， FAS）之結果。透過~\textcite{Kuo2023}~所定義之岩盤條件篩選參考測站並進行殘差計算，所得結果與遍歷性（Ergodic）模型之岩盤模擬高度吻合，驗證本方法在缺乏參考站地區之適用性。在空間分布方面，各區域主頻與放大倍率分布均與地質構造高度一致，台北盆地、宜蘭平原及桃園地區之場址特性差異明顯。本研究亦比較NSA兩種結果、地震水平垂直譜比法（Earthquake Horizontal-to-Vertical Spectral Ratio， eHVSR）與微地動水平垂直譜比法（Microtremor Horizontal-to-Vertical Spectral Ratio， mHVSR）等四種方法，四者在主頻空間分布上具良好一致性，惟部分測站存在系統性差異。依據PSA結果對台北盆地進行地震微分區，整體趨勢與現行耐震設計規範大致吻合，顯示本方法具有支援規範修訂之潛力。此外，以非線性程度指標（Degree of Nonlinearity， DNL）評估2024年4月3日$M_w$~7.3花蓮地震在台北與宜蘭地區之非線性場址效應，結果顯示整體非線性效應相對有限，顯示本研究所建立之線性場址放大函數在本次事件中仍具一定之適用性。整體而言，NSA方法所得之場址放大函數具有頻率相依性與空間代表性，可作為非遍歷性地動模型（Ground Motion Model， GMM）場址項建立之實證基礎，並為地震危害度評估與耐震設計規範修訂提供參考依據。;Benefiting from Taiwan′s dense seismic network and well-established site characterization database， this study applies a Non-reference Site Approach (NSA) to develop non-ergodic linear site amplification functions for stations in the Taipei， Yilan， and Taoyuan regions of northern Taiwan， in terms of both pseudo-spectral acceleration (PSA) and Fourier amplitude spectrum (FAS). Rock-site reference stations are selected based on the Taiwan Generic Rock model \parencite{Kuo2023}， which defines rock sites as having a time-averaged shear-wave velocity in the upper 30~m ($V_{S30}$) between 630 and 910~m/s， and site amplification functions are derived through residual analysis. The results show strong agreement with ergodic model predictions at rock sites， confirming the applicability of the proposed method in areas without reference stations. The spatial distributions of dominant frequency ($f_d$) and amplification factor are consistent with the geological structures of each region， revealing distinct site characteristics across the Taipei Basin， Yilan Plain， and Taoyuan area. This study also compares four methods — NSA (PSA and FAS)， earthquake horizontal-to-vertical spectral ratio (eHVSR)， and microtremor horizontal-to-vertical spectral ratio (mHVSR) — which show good overall agreement in the spatial distribution of $f_d$， although systematic differences are observed at some stations. A seismic microzonation of the Taipei Basin based on PSA results shows overall agreement with the current seismic design code， indicating the potential of the proposed method to support future code updates. In addition， the degree of nonlinearity (DNL) index is used together with direct comparison of site amplification functions to evaluate nonlinear site effects in the Taipei and Yilan regions during the $M_w$~7.3 Hualien earthquake on 3 April 2024. The results show that nonlinear site effects are generally limited， suggesting that the linear site amplification functions derived in this study remain applicable for this level of ground motion. Overall， the site amplification functions obtained by the NSA method， with clear frequency dependence and spatial representativeness， provide an empirical basis for developing non-ergodic ground motion models (GMM) and serve as a useful reference for seismic hazard assessment and seismic design code revision.
