本論文建立一套整合性架構，以改進臺北盆地的 VS30 製圖，本論文建立一套整合性架構，以改進臺北盆地的 VS30 製圖，並以提升高人口密度沉積盆地中地震場址特性評估之可靠性為整體目標。VS30 廣泛作為近地表場址條件的實用代理指標，但區域尺度的製圖常受限於直接剪力波速（VS）量測資料稀少，以及地下地質條件的不確定性。為了因應這些問題，本論文結合決定論與機率論方法，並由三項相互關聯的研究組成。首先，本研究利用基於孔隙比與有效應力所建立之新經驗式 VS 估算函數、將淺層（< 30 m）鑽孔外推至 30 m 深度，以及具有變動局部平均值之克利金法，建立改良後的決定論 VS30 製圖方法。此方法產製出解析度更高的臺北盆地 VS30 圖，並提升場址條件估算的空間涵蓋範圍。其次，本研究建立了一套以機率式馬可夫隨機場（Markov random field， MRF）為基礎的 VS30 製圖架構，用以將地層與地質材料性質的不確定性傳遞至最終的 VS30 預測結果中。此架構結合了基於 MRF 的地層模擬、受地層條件約束之地質材料性質隨機場、經驗式 VS 估算，以及以模型集合為基礎的 VS30 計算。與決定論製圖相比，此機率式架構可提供更佳的預測精度，以及在物理意義上更具合理性的不確定性量化結果，顯示最終 VS30 的變異性不僅受資料間距影響，也受到地質模糊性與地質材料性質變異的控制。第三，本論文利用以臺北盆地鑽孔資料為基礎所建構的合成地層模型，探討地層 MRF 建模中空間相關參數的尺度相依性。結果顯示，校正後的參數並非尺度不變，而是會隨取樣剖面長度與鑽孔密度而改變，突顯出在不同調查尺度下應用 MRF 建模時，進行明確參數校正與審慎解釋的必要性。綜合而言，這三項研究呈現出由決定論 VS30 製圖逐步推進至納入不確定性考量之機率式建模的完整脈絡。;This thesis develops an integrated framework for improving VS30 mapping in the Taipei Basin， with the broader aim of supporting more reliable seismic site characterisation in a densely populated sedimentary basin. VS30 is widely used as a practical proxy for near-surface site conditions， but regional mapping is often limited by sparse direct shear-wave velocity (VS) measurements and uncertainty in subsurface geological conditions. To address these issues， this thesis combines conventional and probabilistic approaches across three linked studies. First， improved conventional VS30 mapping was developed using new empirical VS estimation functions based on void ratio and effective stress， extrapolation of shallow (< 30 m) boreholes to 30 m depth， and kriging with varying local means. This produced a higherresolution VS30 map of the Taipei Basin and improved the spatial coverage of site-condition estimates. Second， a probabilistic Markov random field (MRF)-based VS30 mapping framework was developed to propagate stratigraphic and geo-property uncertainty into final VS30 predictions. This framework combines MRF-based stratigraphic simulations， stratigraphically conditioned geo-property random fields， empirical VS estimation， and ensemble-based VS30 calculation. Compared with conventional mapping， the probabilistic framework produced improved prediction accuracy and more physically meaningful uncertainty quantification， showing that final VS30 variability is influenced not only by data spacing， but also by geological ambiguity and geo-property variability. Third， the thesis investigated the scale dependence of the spatial correlation parameter used in stratigraphic MRF modelling using a synthetic stratigraphic model based on Taipei Basin borehole data. The results showed that the calibrated parameter is not scale-invariant， but varies with sampling profile length and borehole density， highlighting the need for explicit parameter calibration and careful interpretation when applying MRF modelling at different investigation scales. Together， these studies provide a coherent progression from conventional VS30 mapping to probabilistic uncertainty-aware modelling.
