傳統地震災害潛勢評估皆應用衰減模式作為評估的依據，而局部場址特性對PGA 有相當大的影響。由於目前台灣測站的場址特性缺乏可靠的分類參考資料，因此，本研究將分析實測地震記錄測站之觀測值對參考衰減模式的系統誤差，建立台灣地區測站場址特性之修正函數。研究方法使用張毓文(2002)，採用簡文郁(2001) 使用59 個淺源地震分析所得出的衰減模式的係數做為本研究之參考之衰減式。在利用參考衰減模式得到各測站預測初步之計算結果與實測值相較，於第二階段分析時析得到其場址之修正函數。其關係式為ln(PGA Obs )i = C0 +C1 ×ln(Y att ) i，並由回歸分析可得到各測站的修正函數C0、C1 值。如此即可修正測站場址特性所造成之誤差，得到可靠的地動加速度極值。在利用新增之中央氣象局自由場強地動觀測網與中央研究院山區強震站的測站，補足缺乏測站資料的地區，減少因在呈現結果時因內插而造成之誤差，也藉以調查山區之場址特性是否皆為岩盤，於山區設置為岩盤場址之控制點是否為恰當的方法。 將二階段預測之方法並配合地震動潛勢評估所得到的資料，預測第一纇活動斷層有可能發生之地震災害。且每個地震之發生均有其獨特之震源特性，仍會造成預測之不準確。因此，當地震發生時，PGA 值藉由二階段計算得到後，若能結合速報系統的實測值，將可更準確地推估台灣地區地震發生後之地表PGA 分佈，作為早期地震災害潛勢評估。以331地震為例，在經由二階段場址係數修正後與測站實際觀測值之間的標準偏差只有0.62。在利用速報系統的資料做修正，可將標準偏差在降低至0.54，由此可看出此方法預測結果有相當的可靠性。 The main factors that affect the attenuation relationship are the existence of site effects. It is believed that the systematic bias mainly comes from the site effects. Therefore, it is necessary to take account of the site effects in applying the attenuation relationship of PGA, and the results in estimation of seismic hazard will be improved. Up to date, the sites for stations of the Taiwan Strong-motion Instrumentation Program (TSMIP) deployed by the Central Weather Bureau and Central Mountain Array (CMA) deployed by the Academia Sinica still not be classified explicitly. In this study, we used two steps to predict the best results of PGA shakemap. First step, we got the first PGA of each site by attenuation relationship of PGA following Jean (2001), which used more than 3000 seismic records from 59 earthquake events to study the attenuation relationship of PGA. Second step, the site correction of each station of the TSMIP and CMA can be simplified by the following law: ln(PGAobs)S=C0+C1×ln(Yatt)S, where (PGAobs)S is the observed PGA value, (Yatt)S is the predicted PGA value obtained by the attenuation relationship law. The selection criteria for earthquakes are ML > 4.0 and focal depth < 50 km. All seismograms are well recorded by the Taiwan Rapid Earthquake Information Release System (TREIRS) system, TSMIP system and CMA system. The results are reasonably agree with the surface geology from published maps. Site corrections are incorporated into the attenuation relationship to predict PGA values with two steps of process in this study. We have the following conclusions: (1) The PGA values are improved in prediction by taking into account the site effects. (2) The PGA values are improved in prediction by source correction of RTD. (3) For example of Chen-Kung earthquake occurred in 2002 off shore of Taitung, the standard deviation is reduced from 0.663 to 0.6238 with the two steps of data process. The standard deviation is reduced from 0.6238 to 0.5464 with the source correction of RTD of data process. (4) The sites of CMA do not display the characteristics of rock site and this may be due to the local site effect of CMA stations.
