EEW Disseminations - eBEAR System in CWB - Earthworm平台應用於臺灣地震速報預警研究

3 eBEAR System in CWB

3.5 EEW Disseminations

The eBEAR system has issued EEW warnings to about 3600 junior and senior high schools in Taiwan since January 2014. Those schools receive warnings from the CWB and transfer messages to their broadcast system using a user display software, shown in Figure 3-10. From January 2014 to September 2014, there are 28 earthquakes with magnitude greater than 4.5 and depth less than 40 km reported by the CWB. The eBEAR system has reported 20 events and missed 8 events. Figure 3-11 shows the epicenters distribution of the reported and missed events, as well as the reporting times of the

eBEAR system. All of the missed events are located on the offshore area. For the reported

events, the average location error is 4.7±2:9 km and the average magnitude error is 0.2±

0:1. The 21 May 2014 Hualien earthquake with local magnitude 6.0 is the largest event during this period. The eBEAR system issued the alert 15.4 s after the earthquake occurrence. It can provide about 25 s leading time for the Taipei area.

Since January 2014, there have been two false alerts issued by the system. Neither false alert was caused by false triggers. Instead, improper operation caused the false alerts to be generated and sent to the schools. The first false alarm was caused by performing an offline test; because the offline and the online systems run on the same computer, the result of the offline test was sent to the online reporting system and caused a false alert.

To avoid this kind of false alarm, we separated the offline and online systems. The second false alarm was caused by the Earthworm communication modules that provide a rapid message exchange facility between two Earthworm processing systems. When the earthquake occurred, the EEW1 determined the source parameters and sent them to the DCSN using the communication modules. However, the EEW message could not be sent (and instead was stored in the memory) because the connection between the communication modules was broken. When the system operator found the connection problem and restarted them several hours later, they were reconnected again. As a result, the source parameters were received by the DCSN. The alert was then sent to the schools, but it was delayed for several hours after the earthquake occurred. To solve the connection problem, we started to monitor heartbeat debug messages, which is a hand-shaking procedure between the communication modules. The system operator can figure out the connection problems and fix them before the system is triggered by

earthquakes.

Figure 3-10. Graphical output of the eBEAR system during a simulation of the ML 6.5 earthquake in central Taiwan. (top-left) The origin event time and the name of the target city. (center) The rectangle represents the target area. The black line represents the wave front of the P wave. The white line represents the wave front of the S wave. (center-right) The countdown timer for S-wave arrival. (top-right) The predicted intensity of the target area.

Figure 3-11. EEW disseminations of the eBEAR system. There are 28 events with magnitude greater than 4.5 and depth less than 40 km from January 2014 to September 2014. The eBEAR system reported 20 events of them indicated as open circles. The size of circles corresponds to the reporting time. Eight events did not reported by the eBEAR system (open triangles). During this period the largest event occurred in the Hualien area with local magnitude 6.0 and reported at 15.4 seconds after the earthquake occurrence.

Chapter 4 Low-cost Seismometer for EEW

4.1 Palert Seismic Network

The PSN, which consists of 543 low-cost accelerometers, transmits three-component real-time data streams, i.e., the x, y, and z axis data streams, back to the data processing center for regional EEW. The Palert device, shown in Figure 4-1, can sample earthquake shaking at a frequency of 100Hz. Sampled data are digitized with 16-bit resolution between -2g and +2g dynamic range, and time stamped by the Network Time Protocol (NTP) server through the Internet. Figure 4-2 shows the station distribution of the PSN.

Most of the devices are installed on the wall or pillar at elementary schools. Real-time data are packed by each one-second duration and transmitted via Internet. Each Palert accelerometer can transmit data to two servers located at the NTU and the Academia Sinica Grid Computing (ASGC) Centre.

Figure 4-1. Low-cost seismometer. (a)The Palert device. (b) The i-touch device.

Figure 4-2. The station distribution of the two seismic networks. (a)The station distribution of the Central Weather Bureau Seismic Network (CWBSN), (b) The station distribution of the Palert Seismic Network (PSN).

4.2 System Configuration

Figure 4-3 shows that the PSN and the CWBSN are integrated by the Earthworm platform. Although the seismic sensors of the CWBN are made by different manufacturers, corresponding modules can be found in the Earthworm for receiving data streams from the filed seismic stations or other data centers.

EEW systems aims to process real-time seismic data in order to determine the onset of the P-wave arrival, the amplitude of the triggered waveforms and then calculate the location and magnitude of the earthquake. Consequently, through a decision making procedure, warnings are issued to the target areas. In our EEW system, three Earthworm modules, including the PICK_EEW module for P-wave auto-picking, the TCPD module

for earthquake magnitude and location determination, and the DCSN module for warnings reporting are used (Chen et al., 2015). Figure 4-4 shows the configuration of our EEW system in which only the vertical component of the seismic waveforms are used. In Earthworm platform, each waveform packet is temporally stored in a shared memory, called WAVE_RINGs, which has a limited size and only keeps the latest data. The PICK_EEW module detects P-wave arrivals and obtains the peak amplitude in displacement (Pd) of the initial P waves within three-second time window. Then, the detected parameters are sent into another shared memory, called PICK_RING, in which the TCPD module uses the stored parameters for generating the EEW report including the earthquake origin time, location and magnitude. Finally, the DCSN module takes the EEW report from the EEW_RING for other applications such as generating the XML-formatted messages for clients running the EEW display and warning program provided by the CWB (Chen et al., 2015). The DCSN module will also pop up EEW messages on the corresponding CWB staff’s computers, insert EEW message into the MySQL database, and archive the triggered seismic waveforms.

When a large earthquake occurs and the seismic wave propagates away from the epicenter, the number of triggered seismic stations will increase with time. The EEW system will update the EEW report along the triggered seismic stations. However, in the early stage, the EEW report of the system may contain large uncertainties in location because only few stations are triggered. Thus, other metric should be used to ensure that the EEW report is reliable. The GAP is one of the key factors to determine if the earthquake location report is good enough when the earthquake is inside a seismic network (Wu et al., 1997, 1999, 2013). In the earthquake localization process, the

localization error can be reduced with a small value of the GAP. An EEW system normally updates its report along with the increase of triggered stations since the GAP value decreases. It is necessary to find a suitable criteria for obtaining an EEW report with relatively low GAP and low reporting time.

We analyzed the data set from the online CWBSN-EEW (Hsiao et al., 2011; Chen et al., 2012; Chen et al., 2015). There are 117 earthquakes detected by the system from January, 2014 to August, 2014. Figure 4-4(a) shows the relation between the order of the EEW reports and the number of triggered stations; Figure 4-4(b) shows the relation between the order of the EEW reports and the reporting time and the GAP. Generally, the GAP decreases along with the EEW reports but the reporting time increases. We found an intercept in Figure 4-4(b), which shows the fifth EEW report could be a good point for determining decent source parameters. Moreover, in order to obtain a specific proxy of the criteria, Figure 4-4(a) shows that the fifth report needs at least 13 triggered stations in average. Therefore, in this study, the CWBSN-EEW will issue reports when the number of triggered station is at least 13. In addition, for generating more stable results of the ISN-EEW, we chose EEW reports with GAP equal or less than the reports generated by the CWBSN-EEW.

Figure 4-3. A schematic diagram of the data processing of combined seismic network. The first part is the data source which provides real-time seismic data streams from different kind of seismic sensors and other institutions. The middle part is the procedure of data processing in the Earthworm system at the data center. The last part is the applications which receive information from the middle part and use the information.

Figure 4-4. Relationships between the EEW parameters of combined system. (a) Relationships between the order of the EEW report and the triggered stations; (b) Relationships between the order of the EEW report, and the reporting time, and the GAP.

4.3 Magnitude Estimation Using Palerts

For the EEW system, a reliable estimation of earthquake magnitude depends on two primary factors. One of them is a robust picker for precisely detect P wave onset time and intelligently avoiding the noise. The other is a statistically significant regression equation for predicting earthquake magnitudes using only the initial portion of P waves. Palerts installed in the buildings of elementary schools may affected by human activities and may be amplified the amplitude by building responses. Therefore, we adopted the Earthworm’s picker (Chen et al., 2015) and followed by new picking constraints for better determine the P wave onset time and preventing from false picks caused by noise. We also constructed a new regression equation to correct for the building response and aim to get better predictions of the earthquake magnitudes.

To ensure every P-wave picks from Palerts with high quality is crucial for the EEW system. We applied the P-wave picking algorithms from the Earthworm module, PICK_EEW, (Chen et al., 2015), and followed by new picking constraints with three parameters, XON, XP0 and XP1, for evaluating qualities of picks. XON is the first deference of filtered data at pick time; XP0 is the first maximum filtered data of the preceding half cycle; XP1 is the second maximum filtered data of the preceding half cycle.

All of them are normalized by the 1.6 times of the running mean absoluted value of filtered data. Each valid picks generated from Palerts should be satisfied by one of the following two criteria. One is that either XP0 or XP1 should larger than 13.0 and the XON should larger than 3.0. Another is that either XP0 or XP1 should larger than 20.0 and the XON should larger than 0.8. Figure 4-5(a) shows examples illustrating that picks corresponding to the criteria were considered as high quality; in contrast, Figure 4-5(b)

shows examples illustrating that picks were considered as poor quality. These criteria are quite useful to evaluate qualities of picks detected by the Palerts.

To correct that the seismograms recorded by the Palerts were amplified by the building response, we used 46 events, shown in Table 1, including 649 vertical-component records to determine Pd, which is the peak amplitude of the initial P-wave displacement, in 3-second time window. The seismograms recorded by the Palerts were integrated twice to obtain the displacement and then a 0.075 Hz high-pass recursive Butterworth filter was applied to remove the low-frequency drift after the numerical integration. Each P-wave arrivals was verified manually to ensure the quality is good for constructing an empirical formula between the Pd values and the earthquake magnitudes. We assume a linear relationship among the logarithmic Pd, the magnitude M, and the logarithmic hypocentral distance R:

Log Pd = A + B ╳ M + C ╳ log10 (R) (1)

where A, B and C are constants to be determined from the regression analysis using the P waves from the 46 events. In the regression analysis, we used R software (R Development Core Team, 2006) to detect and remove outliers within the data, and then fit the model to the data. The best-fitting attenuation relationship for log Pd is found to be

Log Pd = -2.797 + 0.404 ╳ M – 0.539 ╳ log10 (R) ± 0.33 (2)

The equation (2) was used for estimating earthquake magnitudes using vertical-component P waves recorded by the Palerts.

Figure 4-5. Examples of the automatic P-wave arrival detection. (a) High quality picks. The parameters XP0, XP1 and XON are over the criteria; (b) Poor quality picks. The parameters XP0, XP1 and XON are under than the criteria

4.4 Offline Test

To test the integrated system, ISN-EEW, in off-line mode, we collected seismic waveforms with magnitudes greater than 4.5, depths less than 40 km, and epicenters within 40 km of the coastline of Taiwan from 2013 to January 2015. Table 1 shows the dataset which consists of 46 events including three events with magnitudes between 6.0 and 6.5. The results of the off-line simulations are compared with those generated by the CWBSN-EEW.

In the off-line test, the ISN-EEW use the same Earthworm’s picker but different criteria to detect P wave arrivals for the P waves recorded by the CWBSN and the ISN.

Figure 4-6 shows the comparison of the source location errors between the CWBSN-EEW and the ISN-EEW. The difference of epicenter error of CWB-EEW and ISN-EEW are 0.3 km. For the depth error, the ISN-EEW is a little better than the CWB-EEW. It means the ISN-EEW can have stable results in earthquake location. The Pd

values from the CWBSN are used to estimate earthquake magnitudes (MPd) based on the empirical formula of Hsiao et al., (2011). However, for the Pd values from the PSN, the equation (2) was used for estimated earthquake magnitudes. Figure 4-7 shows the comparison for the estimated magnitudes. The estimated magnitudes from CWBSN-EEW and ISN-EEW are compared to the CWB catalog created by manual phase picking and locating. The CWBSN-EEW and the ISN-EEW have error of 0.28 and 0.25 unit, respectively. The ISN-EEW is able to provide robust estimations of earthquake magnitudes. The results implies that the amplified P waves caused by the building effects are correcting by the equation (2). Comparing Figure 4-8, the reporting time are 14.7 and 13.1 seconds for the CWBSN-EEW and ISN-EEW, respectively. Figure 4-9 shows the

comparisons of blind-zone areas distribution of each event. Some events located in the region with dense seismic stations may be reduced the blind zone area to 30 km.

Figure 4-6. The location error comparisons between CWBSN and ISN. (a) Comparison between the CWBSN-EEW and the he CWB catalog analyzed by manual phase picking; (b) Comparison between the ISN-EEW and the he CWB catalog analyzed by manual phase picking.

Table 2. Data for offline test in integration system

1

Date Latitude Longitude Depth ML CWBSN-EEW ISN-EEW CWBSN-EEW ISN-EEW

CWBSN-

40 02/19/13 120.60 22.91 16 4.7 14.1 9.8 158 82 5.3 1.4 13 11

41 02/20/13 121.39 23.23 20 4.5 15.2 15.4 156 148 3.4 2.1 16 3

42 03/04/13 121.33 23.00 24 4.6 14.8 11.1 177 173 4.0 5.4 16 4

43 03/07/13 121.46 24.30 6 5.9 14.1 12.1 69 69 1.3 1.1 16 4

44 03/07/13 121.45 24.34 6 4.6 13.7 13.6 69 69 1.6 1.6 20 3

45 03/20/13 121.95 24.45 12 4.6 18.9 19.4 149 145 0.5 2.9 60 15

46 03/27/13 121.05 23.90 19 6.2 9.8 9.8 102 82 1.5 0.8 21 16

Average 14.7 13.1 139.8 120.7 3.1 3.4 22.1 8.2

Figure 4-7. The magnitude error comparisons between CWBSN and ISN. (a) Comparison between the CWBSN-EEW and the he CWB catalog analyzed by manual phase picking; (b) Comparison between the ISN-EEW and the he CWB catalog analyzed by manual phase picking.

Figure 4-8. The reporting time comparisons between CWBSN and ISN. (a) Reporting time of the CWBSN-EEW; (b) Reporting time of the ISN-EEW. In average, the ISN-EEW has smaller reporting time than the CWBSN-EEW.

Figure 4-9. Blind zone radius comparisons between CWBSN and ISN. (a) Blind zone radius of the CWBSN-EEW; (b) Blind zone radius of the ISN-EEW.

4.5 Summary

Using low-cost seismometers to construct a regional seismic network is an attractive solution for EEW systems. In spite of the relative low signal-to-noise ratio and the impact of building responses on the amplitude of the seismic waveforms, the P-wave arrival time, detected by the Earthworm’s picker (Chen et al., 2015) and followed by the new picking constraints, is precise for large earthquake and the amplitude can be corrected by removing the building responses. Wu et al., (2013b) demonstrated that the regional seismic network based on the Palerts is good enough for determining earthquake location,

magnitude and intensity. We further integrate the PSN and the CWBSN to make a regional seismic network, ISN, with higher density in Taiwan. This is the first time to integrate a traditional seismic network with a low-cost seismic network. Because of the dense station coverage of the ISN, when inland earthquakes occurred, the EEW system based on the ISN is able to gather P-wave arrivals faster than that based on the CWBSN.

The results of the off-line test implies that the EEW system based on the ISN can reduce reporting time and estimate decent earthquake location and magnitude for the purpose of earthquake early warning.

EEW system updates the earthquake information along with the arrival of new data in the system. It is a challenge to decide when the accuracy of the updated result will be good enough. One possible metric is to use GAP. For earthquakes occurred inside the seismic network the lower the GAP, the higher the accuracy may be reachable for the estimated earthquake location as well as the magnitude. However, lower GAP usually needs more stations. For earthquake localization, it means that the calculation should wait until more stations are triggered and the reporting time of the system is increased. By studying the relationship between the GAP and the number of triggered stations, a specific criteria of 13 triggered stations is found for the specific condition in Taiwan to compromise the tradeoff between the speed and the accuracy in the EEW system.

Chapter 5 A Case Study for Mw7.6 Chi-Chi Earthquake

As a result of the large ground shaking of the 1999 Chi-Chi earthquake, several electrical power towers collapsed, which resulted in real-time data interruption. If the 1999 Chi-Chi earthquake were to happen again with limited workable stations and signal recording length, we wonder if the proposed EEW system would provide precise and reliable event information. It is a big challenge to the current EEW methods in magnitude and intensity estimations, because the data streams might be broken within the initial 10 seconds after the first P-wave arrival, as happened in the 1999 Chi-Chi earthquake. The purpose of this study is to offline test the new proposed EEW system (Hsiao et al. 2011) by feeding the raw records of the 1999 Chi-Chi earthquake into the system. Both τc and Pd

were used to estimate the magnitude. The results indicate that the first warning is available in about 12 seconds after the earthquake origin time and the magnitude estimated by theτc method (Mτc = 7.4) is better than that from using the Pd method (MPd = 6.3). Even with limited stations and data interruptions such as occurred during the 1999 Chi-Chi earthquake, the proposed EEW system still can provide quick and satisfying event information.

5.1 Signal Interruption

Before the 1999 Chi-Chi earthquake, there were 61 real-time strong-motion stations operated by the CWB with a 16-bit dynamic range and a 50-Hz sampling rate. To save communication expenses, some of the stations directly transmitted data to the processing center via 4.8-K phone line, while others first transmitted data to sub-centers, which are multiplex all data streams, and then transmitted them to the data processing center via a broadband dedicated line, named the T1 line. Unexpectedly, the Hualien T1 line, consisting of six stations, was interrupted five seconds before the Chi-Chi earthquake due to a mechanical problem. In addition, during the strong ground-shaking period the

在文檔中 Earthworm平台應用於臺灣地震速報預警研究 (頁 49-0)