近年來地震預警有著廣泛的發展，但是其地動預估準確度卻有待加強，特別是受到場址影響的地區，可能會有誤差較大的情形發生，然而臺灣擁有分布密集的地震觀測網，利用大量的地動紀錄，進行系統性的誤差分析，最終，有效的修正預估誤差，以利於評估地震災害造成生命及財產上的損失，而在地震預警上主要分成兩類，一種是區域型預警，而另一種則是現地型預警，區域型預警為利用數個測站組成觀測網進行觀測，當地震發生時迅速的利用觀測網決定地震規模、震源距等資料，除了會對觀測網內進行估算地震資訊，對較遠區域也會提供警訊。現地型預警則為利用單一測站收取資料，當地震發生時僅對該測站區域提供迅速警報。而本篇研究主要針對現地型地震預警進行討論，而現地型預警主要為使用兩種地動參數進行後續的估算，一個是τc為平均地表週期，另一個為Pd是地表最大位移值(peak displacement)。根據使用的參數不同，本研究將方法訂為τc- Pd法與Pd法。 本研究使用的資料為臺灣強地動觀測計畫(Taiwan Strong Motion Instrument Program, TSMIP)從1993~2014年的地震資料，篩選條件為芮氏規模(ML)五以上，在τc-Pd法方面，本研究主要是使用由許喬筑(2015)所提出建議使用的NCREE MODEL，利用參數Pd及τc來估算震源距及規模(Mw)，當估算完成後再對兩者進行測站修正以降低誤差，進而再用兩者的修正值套入地動預估式(Ground Motion Prediction Equation, GMPE)中並考慮各測站場址修正對最大地動加速度值(Peak Ground Acceleration, PGA)進行估算。而在Pd法上我們試著討論不同區域範圍之兩階段模型以進行估算，其一為使用全部研究區域之Pd及觀測PGA所建立，另一種則是利用各小區域的Pd及觀測PGA所建立出各區的模型，進行初步的PGA預估，之後對於各測站進行場址修正以降低預估誤差。本研究將會用NCREE MODEL及全部研究區域資料建立的模型和小區域資料建立的模型三者求得之PGA的修正量進行比較。經過全部測站平均比較，以及目標地震測試後，整體而已以T model在PGA預估上有較小的誤差值。 ;In recent years, there has been a wide development on the application of earthquake early warning system (EEW). But the accuracy, affected by the site still existed some estimated errors, has to be strengthen. In order to secure the life safety and property, improving the accuracy of EEW is important. If the error can be improved, accompanied with data information from the intense observation around Taiwan, the EEW can offer a better hazard prediction for different regions. The classification of EEW system is divided into two major approaches. One is the regional warning and the other is the onsite warning. In this study, we mainly use onsite EEW to discuss. There are two kinds of methods used in our discussion. One is τc -Pd method and the other is Pd method. In this study, we used the earthquake event (with magnitude over five) recorded by TSMIP from 1993 to 2014. For τc -Pd method, we performed NCREE model which Hsu proposed in 2015 by using parameter Pd and τc and station correction to estimate magnitude and hypo-distance. Once we got the estimated magnitude and hypo-distance after station correction, we could estimate the PGA through the attenuation equation (GMPE) and its site correction function. In Pd method, this study estimated the PGA with two kinds of models. One is the model established with Pd and observed PGA of whole regions in Taiwan (ALL model) and the other one is regional model established with regional Pd and PGA. We also do the site correction after getting initially estimated PGA. We will discuss value of estimated PGA by NCREE model, ALL model and regional model.
