Earthquake Reaearch in China  2017, Vol. 31 Issue (1): 125-135
A Study on the Early Warning Time of Strong Earthquakes in Taiwan
Chen Huifang1, Kang Lanchi1,2, Jin Xing1,2, Shao Pingrong1, Cao Yi1
1 Earthquake Administration of Fujian Province, Fuzhou 350003, China;
2 Institute of Engineering Mechanics, China Earthquake Administration, Harbin 150080, China
Abstract: The paper collects the records by the Fujian Digital Seismic Network of 40 shallow earthquakes in Taiwan with MS ≥ 5.0 from 1999 to 2013, analyzes the seismic phase (Pn, Sn phase) characteristics and travel-time rules, determines travel-time models and develops a seismic phase travel-time equation based on the two-step fitting algorithm. With the deduction of processing time and network delay time, this method can provide an accurate estimation of early warning time of Taiwan earthquakes for the Fujian region, and has been officially employed in the earthquake early warning system of Fujian Province.
Key words: Strong earthquakes of Taiwan     Characteristics of seismic phases     Travel-time rule     Early warning time

INTRODUCTION

Earthquake early warning is the rapidly estimation of earthquake parameters and prediction of the influence of earthquakes on surrounding areas based on early seismic wave information observed by nearby seismic stations where the earthquake occurred, and to issue early-warning information about earthquake motion intensity and hitting time in various regions before a destructive seismic wave arrives at the surrounding areas of the epicenter, accounting to the fact that electromagnetic waves propagate much faster than seismic waves, and seismic initial P-waves propagate much faster than subsequent destructive seismic S-wave or surface waves, so enterprises and the public can take measures for earthquake emergencies as soon as possible, thus reducing casualties (Andrew, 2005; Zhao Jidong et al., 2009). Although the warning time is very short, usually only a few seconds or more than ten seconds, such a short time can still save many lives and greatly reduce economic losses (Yang Maling et al., 2004; Li Shanyou et al., 2004; Chen Yuntai, 2007; Yuan Zhixiang et al., 2007). When a great earthquake with MW9.0 struck eastern Japan on March 11, 2011, the Japanese earthquake early warning system issued warning to 37 cities within a minute, which provided valuable escape time for a large amount of the population. Without doubt, regions closer to the epicenter gain a shorter early warning time. With respect to Tokyo, where a 1-minute early warning time was achieved, Iwate which is closer to the epicenter, Ishinomaki, only received 12 seconds. For this earthquake, almost all the land had an early warning time of more than 10 seconds, saving many lives. At that time, 23 Shinkansen trains carrying passengers were in operation, and trains in Tokyo applied brakes immediately after receiving the warning information. The first tremor started 9-12 seconds after braking, and the most intense shaking appeared after 70 seconds, so there were no train derailment accidents occurring at that time. The early warning system in Mexico also had successful experience in early warning information release. On September 14, 1995, an MS7.3 earthquake occurred at Guerrero in central Mexico, and the system issued a public warning 72 seconds before the S-wave arrived. The subway stopped 50 seconds before the S-wave arrived, gaining a 60 second early warning time, and schools made emergency responses of evacuations (Wyss, 1991; Espinosa et al., 1995; Xing et al., 2008a; Jin Xing et al., 2008b; Espinosa et al., 2009; Jin Xing et al., 2010; Zhang Hongcai et al., 2012).

Taiwan is located on the boundary of the Eurasian plate and Pacific plate, with intense seismic activity, where strong earthquakes with MS≥6.0 occur at least twice per year on average and earthquakes with MS≥7.0 occur once every 3 years on average (Chen et al., 2012; Chuang et al., 2012). Each earthquake had great influence on Fujian Province, thus implementation of existing early warning system in Fujian to realize early warning for earthquakes in Taiwan have important practical significance (Liao Xu et al., 2002; Ma Qiang, 2008; Zhang Hongcai et al., 2011; Robert et al., 2012; Serdar et al., 2014). Early warning time is one of the important services for an early warning system, which is also an important parameter reflecting the disaster reduction effectiveness of an early warning system. For early warning time for arrival of strong earthquakes in Taiwan to the Fujian area, the original early warning system uses the direct longitudinal wave Pg and Sg to calculate the early warning time, while actual main seismic phases from Taiwan earthquakes to Fujian area are head waves (Pn, Sn) (Liu et al., 2009; Lin et al., 2011), which directly affect the estimate of early warning time. In this paper, research is done based on strong earthquake seismic phases in Taiwan, and according to travel time of a certain seismic phase, with the deduction of processing time and network delay time, helps realize accurate estimation of early warning time for strong earthquakes in Taiwan.

1 DATA

Since 1998, the Digital Seismic Network in Fujian Province has undergone rapid development during periods of the ninth "Five-year Plan", the tenth "Five-year Plan" and the eleventh "Five-year Plan" (Table 1, Fig. 1), which is currently composed of 85 seismic stations. The Digital Seismic Network in Fujian Province uses a EDAS-24IP data collector and 7 types of seismometers provided by the Beijing Gangzhen Mechanical & Technology Co., LTD (broadband CMG-3ESPC-60, broadband CMG-3ESPC-120, broadband CTS-1, broadband BBVS-60, broadband FBS-3B, short period FSS-3B and short period FSS-3M), with sampling rate of 100Hz. All stations are built on bedrock stylobate, with average spacing of about 31km, and observed data is transmitted in real-time using telecommunication lines (Jin Xing et al., 2007).

Table 1 The number of stations for the Fujian Seismic Network at different stages

 Fig. 1 Observation stations of the Fujian Digital Seismic Network Pink triangles are stations built during the period of the ninth "Five-year Plan"; green triangles are stations built during the period of the tenth "Five-year Plan"; blue triangles denote stations built during the period of the eleventh "Five-year Plan"; red circles represent epicenters of strong earthquakes in Taiwan

In this paper, the selection of seismic data follows the following principles: Firstly, earthquakes have large magnitude, which has certain influence on the Fujian area. Secondly, seismic phases are clear and can be accurately identified. Thirdly, the epicenters are distributed all over Taiwan island as far as possible. Fourthly, seismic phases of earthquakes in a certain area have almost the same travel time rule, and to make processing a bit easier, some typical earthquakes are selected in this paper. Finally, deep-focus earthquakes often occur in Taipei and its surrounding waters in northeastern Taiwan. Such earthquakes have little impact on the Fujian area, and cannot possibly trigger the early warning system, thus they will not be selected. Based on the above basic principles, 40 shallow-focus earthquakes with magnitude over 5.0 in Taiwan recorded by the Earthquake Administration of Fujian Province during 1999-2013 are selected in this study (Table 2). For the Fujian area, earthquakes in Taiwan are all outside-network earthquakes, and the existing technology of the Fujian Digital Seismic Network can't accurately determine epicenter location and origin time of earthquakes, therefore, all seismic information used in this paper is subject to earthquake reports generally released by the website of the "Central Weather Bureau" of Taiwan (http://www.cwb.gov.tw).

Table 2 Earthquake catalogue for strong earthquakes in Taiwan

Seismic waves spread across the Taiwan Straits from Taiwan to Fujian. Due to their special propagation path and special geological structure, for all Taiwan earthquakes recorded by the Fujian network, only seismic phases Pn and Sn are identified. The head wave (Pn) arrives as the first seismic phase (Li Ke et al., 2007), while seismic phases Pg and Sg are not developed (Fig. 2), and altogether 893 records of clear seismic phase Pn and 716 records of seismic phase Sn are obtained by manual processing (Table 2, Fig. 3).

 Fig. 2 The Nantou MS6.2 earthquake in Taiwan on March 27, 2013 (a) Observation data from Tancheng station in Pingtan, Fujian. (b) Observation data from Jingfeng station in Huian, Fujian

 Fig. 3 Distribution map of epicenters for seismic phase Pn and Sn (a) Distribution map of epicenters for seismic phase Pn used in this paper. (b) Distribution map of epicenters for seismic phase Sn used in this paper; Red solid circles denote epicenters for strong earthquakes in Taiwan during 1999-2013
2 METHODS

Calculation of the early warning time is required to meet the basic conditions of efficiency and accuracy. As mentioned, the seismic phase Pn and Sn of strong earthquakes in Taiwan can be observed by the Fujian Digital Seismic Network, therefore, a one-layer one-dimensional model is adopted in this paper to calculate the travel time of the seismic phase, and in the one-layer one-dimensional model, travel time equations for the head wave are as follows:

 $t = \frac{{2H - h}}{{{v_1}}}\cos {i_0} + \frac{\mathit{\Delta }}{{{v_2}}}$ (1)
 $sin{i_0} = \frac{{{v_1}}}{{{v_2}}}$ (2)

Where, t denotes travel time for Pn or Sn, H crustal thickness, h focal depth, v1 and v2 represent respectively P-wave velocity or S-wave velocity of the earth's crust or mantle, and Δ is the epicentral distance. It is obvious that formula (1) is composed of two terms. The former term stands for travel time of the seismic wave propagating from the earth crust to mantle and then to the ground after gliding through the mantle, which includes the impact of focal depth, and the second term is the travel time for the slide wave on Moho discontinuity. Strong earthquakes in Taiwan adopted in this paper have epicenter distances of over 120km, therefore the second term is the main control factor in formula (1), and the former term is simplified as a constant T. Formula (1) can be simplified as follows:

 $t = T + \frac{\mathit{\Delta }}{{{v_2}}}$ (3)
 $T = \frac{{2H - h}}{{{v_1}}}\cos {i_0}$ (4)

Two process fitting is done to formula (3). Step one, fitting is conducted on the jth earthquake, and we get vj and Tj. By combined analysis we get average v2; step two, to put v2 into formula (3), T is calculated.

Owing to differences in earthquake depths, the influences on T are also different. Therefore, the adoption of the first step in the two-step method can effectively deduct the influences from earthquake depths, avoid influence on the fitting of v2 caused by different focal depths and obtain mantle wave velocity with high accuracy and precision. This term is the main control factor in formula (1), which ensures accurate results from our study.

3 RESULTS AND ANALYSIS

Fitting is conducted on 40 selected Taiwan earthquakes by the two step method, and fitting results are formula (5) and (6),

 ${t_{{\rm{Pn}}}} = 6.28 + \frac{\mathit{\Delta }}{{8.00}},\;\;\left( {\sigma = 0.76} \right)$ (5)
 ${t_{{\rm{Sn}}}} = 10.21 + \frac{\mathit{\Delta }}{{4.57}},\;\;\left( {\sigma = 0.86} \right)$ (6)

All parameters are shown in Table 3, and results are shown in Figs. 4 and 5. It can be seen from Figs. 4 and 5 that fitting results from our study more accurately reflect the rule of travel time for seismic phase Pn and Sn, in mantle vPn=8.00km/s, vSn=4.57km/s. In Table 1, the differences between them are all less than 1s, and as shown in Fig. 7, there is a 98.2% probability that the difference between observed Pn-wave travel time and speculated travel time is within 1s and 95% probability that statistical travel time difference is less than 0.738s, and there is a 87.1% probability that the difference between observed Sn-wave travel time and speculated travel time is within 1s and 95% probability that statistical travel time difference is less than 1.989s. Fitting results for the Pn-wave are better than that of the Sn-wave, and the above analysis demonstrates that fitting results in this paper are stable and reliable.

Table 3 Fitting results

 Fig. 4 Fitting results for Pn wave travel time

 Fig. 5 Fitting results for Sn wave travel time

 Fig. 6 Distribution probability for fitting results (a) Pn distribution probability, (b) Sn distribution probability

 Fig. 7 Comparisons between our research results and early warning time algorithm of original early warning system The red solid line is the Pn travel time for our research; blue solid line is the Sn travel time for our research; red dotted line is Pn travel time given by current early warning system; blue dotted line represents Sn travel time given by the current early warning system
4 EXAMPLES

Results in this paper are compared with the early warning time algorithm of original early warning system. As shown in Fig. 7 and Table 4, the early warning time algorithm of original early warning system obviously overestimated the warning time for Taiwan earthquakes, and error of early warning time also increases with the increase of distance. At the distance of 400km, travel time errors for both Pn-wave and Sn-wave are more than 10s. In this paper, we take the earthquake early warning in the year of 2013 as an example illustrating the practicability of our research results. Because positioning results will have a great influence on calculation of early warning time, for this earthquake, positioning results from the Taiwan network and early warning positioning results are very close to each other. The early warning system successfully processed information of this earthquake and issued S-wave early warning results within the Earthquake Administration of Fujian Province. Four cities greatly influenced by this earthquake are selected in this paper, and results released from their early warning system are shown in Fig. 8. Comparisons are shown in Table 5. Results from our research are apparently better than that of the original early warning system, which is closer to actual fact.

Table 4 Comparisons between our research results and early warning time algorithm of the original early warning system

 Fig. 8 Early warning time interface released by original early warning system

Table 5 Comparisons of early warning time from our research, original early warning system and actual observations
5 CONCLUSIONS

For the purpose of improving the practical value of early warning and developing the early warning system, from the perspective of the most important service product-the early warning time, following the principle of efficient and accurate estimate of early warning time, and starting with observation data analysis, a way to estimate early warning time based on different seismic phases of strong earthquakes in Taiwan is proposed in this paper, and the following conclusions are reached:

(1) By the comparison between early warning time given by original early warning system and actual observation data, we believe that the accuracy of the early warning time estimated by the early warning system from our research is greatly improved, indicating that the suggested method in this paper can provide technical support for further improvement of early warning systems in future.

(2) By the analysis of abundant observation data of strong earthquakes in Taiwan, this study shows that seismic waves from strong earthquakes in Taiwan that propagate to Fujian are seismic phase Pn and Sn. The one-layer one-dimensional crustal model is selected and two-step fitting method is used for statistical regression. The method is simple and clear, using a clear physical concept, with short-time calculation, and the accuracy that meets early warning requirements.

(3) With the large-scale construction of an early warning system in China and promotion of early warning techniques, the research in this paper can be of reference for other regions and the suggested research method can be applied to other regions.

(4) Seismic waves from Taiwan earthquakes pass through the Taiwan Straits and then arrive at Fujian. The path is complex and the crustal structure has not been verified, resulting in rather complicated seismic phases from Taiwan earthquakes. Research in this paper can provide reference for analysis of seismic phase in network catalogues.

At present, results in this paper have been used for the earthquake early warning system in Fujian Province, and have been officially employed to estimate early warning times in Fujian Province for Taiwan earthquakes.

REFERENCES
 Chen Yuntai. Earthquake prediction:Progress, difficulties and prospect Ⅱ[J]. Seismological and Geomagnetic Observation and Research, 2007, 28(2): 1–24. Jin Xing, Kang Lanchi, Ou Yiping. Ground motion attenuation relation for small to moderate earthquakes in Fujian region, China[J]. Acta Seismologica Sinica, 2008b, 30(3): 279–291. Jin Xing, Liao Shirong, Chen Feiwen. Study on real-time fast report system in seismic network[J]. Seismological and Geomagnetic Observation and Research, 2007, 28(2): 51–54. Jin Xing, Zhang Hongcai, Wei Yongxiang, et al. Preliminary study of near real-time interpolation calculation of ShakeMap based on seismic monitoring network's information[J]. Journal of Disaster Prevention and Reduction, 2010, 26(1): 1–11. Jin Xing, Zhang Hongcai, Wei Yongxiang. Shaking intensity criterion based on rapid release information by seismic monitoring network and its application[J]. Recent Developments in World Seismology, 2008a, 10(1): 20–27. Li Ke, et al. Seismology and Earthquake Observation[M]. Beijing: Seismological Press, 2007: 169-174. Li Shanyou, Jin Xing, Ma Qiang, et al. Study on earthquake early warning system and earthquake intelligent emergency controlling system[J]. World Information of Earthquake Engineering, 2004, 20(4): 21–26. Liao Xu, Huang He. Study on earthquake alarm system for enterprises[J]. Earthquake Engineering and Engineering Vibration, 2002, 22(6): 142–149. Ma Qiang. Study and Application on Earthquake Early Warning System[D]. Ph D. thesis. Harbin: Institute of Engineering Mechanics, China Earthquake Administration, 2008 (in Chinese with English abstract). Yang Maling, Shen Fanluan, Wang Zhengshang. Forecast of warning level-A new proclamation form of earthquake prediction and thoughts on its countermeasures[J]. Recent Developments in World Seismology, 2004, 8(2): 1–12. Yuan Zhixiang, Shan Xiuzheng, Xu Shifang, et al. An overview of earthquake early warning technology[J]. Journal of Natural Disasters, 2007, 16(6): 216–223.

1 福建省地震局, 福州市华鸿路7号 350003;
2 中国地震局工程力学研究所, 哈尔滨 150040