DCSN Module

in which Wi is the weighting factor of each MPd value and Ri (in seconds) is the P-wave travel-time residual for each corresponding MPd value. Finally, a weighted average for waves propagate away from the epicenter. As a result, the TCPD module determines the earthquake message and continuously updates that message. We propose that the numbers of updating earthquake messages will increase quickly and will be significant for large and local earthquakes. In contrast, for small earthquakes or for noise, the number of updating earthquake messages will increase slowly and will be small. Therefore, if the EEW system determines a large number of updating earthquake messages for an ongoing earthquake, we consider the EEW information as a reliable warning. To prevent false alarms, the DCSN module always skips the first and second earthquake message generated from the TCPD module. The third earthquake message is the first EEW report

30

to users. The EEW report is written in an XMLformatted file for broadcasting. The EEW report is updated either when differences in the magnitude or the epicenter are larger than 0.5 or 20 km, respectively, as compared to the last EEW report. A user display pops up automatically when an XML-formatted message is received. The display estimates the seismic intensity, the wave fronts of P- and S-waves, and the remaining warning time (defined as the time between the reporting time and the arrival of the S wave to the target area). If the EEW report is updated, the user display directly changes the location of the epicenter and again re-estimates EEW-related parameters.

The DCSN module takes the EEW report from the HYPO_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.

31

StaFilt Sets the filter parameter (time constant) StaFilt used in the calculation of the short-term average (STA) of the characteristic function of the trace.

0.6

LtaFilt Sets the filter parameter (time constant) LtaFilt used in the calculation of the long-term average (LTA) of the characteristic function of the trace.

0.15

EventThresh Sets the STA/LTA event threshold. 5 RmavFilt The filter parameter (time constant) used to

calculate the running mean of the absolute value of the waveform data.

0.9961

DeadSta Sets the dead station threshold (counts). 1000000 MinPa (new) Defines the minimum value of peak

amplitude for acceleration (unit is cm/sec

)

0.01 MinPv (new) Defines the minimum value of peak

amplitude for velocity (unit is cm/sec)

0.0001

32 with one Geotech Smart24A seismometer that transmits real-time, strong-motion data to the CWB via 4800-baud leased telephone lines. Each telemetered signal is digitized at 50 samples per second using a 16-bit resolution. The current EEW system, VSN, operates within this seismic network. The second, the Central Weather Bureau Seismic Network (CWBSN), is an upgraded and integrated network that improves data quality, station coverage, and density by integrating various types of seismic stations and seismic networks from other institutes. The eBEAR system is operated under the CWBSN. The station distribution of the CWBSN, which integrates different types of seismic stations operated by the CWB and the Institute of Earth Sciences (IES) of Academia Sinica (which provides waveforms for 23 stations from the Broadband Array in Taiwan for Seismology), is shown in Figure 3-1. In addition, using a connection to buffer uniform data of the Incorporated Research Institutions of Seismology (IRIS), one Japanese station (YOJ) has been merged into the monitoring network and has improved station coverage within the eastern offshore region of Taiwan. Each real-time seismic signal, digitized at 24-bit resolution and obtained using time stamps from a Global Positioning System, is packed and transmitted to CWB headquarters in Taipei via Ethernet or Internet. With the

33

exception of IRIS data at 20 samples per second, digital signals are digitized at 100 samples per second.

Figure 3-1. The station distribution of the CWB Seismic Network.

The CWBSN consists of four types of seismic stations including six-channel seismic stations, broadband seismic stations, borehole seismic stations, and one cable-based ocean-bottom seismic station. Among seismic stations, some have been upgraded from older types, while others have been newly added. Six-channel seismic stations were upgraded and combined from original short-period and strong-motion instruments, digitized at 12- and 16-bit resolutions, respectively (Teng et al., 1997). Prior to station upgrades, two types of instruments were operated separately and transmitted data through telephone lines; signal time was stamped by the central station (Chang et al., 2012). Since

34

2007, using Geotech Smart24A accelerometers to replace the original instruments (Geotech A900A) and to connect Teledyne Geotech S13 short-period sensors, the CWB has combined these two types of seismic signals. As a result, 70 upgraded six-component stations have been constructed, each hosting three-component short-period velocity sensors and one three-component strong-motion sensor.

For EEW purposes, the data loggers located at broadband seismic stations were replaced using modern equipment capable of sending seismic waveforms with a 1 s packet length. The system consists of 23 stations that use one three component broadband seismograph. To prevent clipped waveforms caused by near-field strong shakings, most stations are equipped with an additional three-component strong-motion sensor. Such high-quality waveforms are also used to obtain centroid moment tensor (CMT) solutions (Shin et al., 2013).

In addition, 30 borehole seismic stations are operational. Each hosts a three-component strong-motion seismograph on the surface and a three-component strong-motion seismograph, as well as a broadband seismograph within boreholes at a depth of approximately 300 m from the surface. Seismic signals from borehole seismometers provide waveforms with a high signal-to-noise ratio, useful for improving the accuracy of phase picking. Since 2008, the number of borehole seismic stations has increased by approximately five stations each year. In the near future, the total number of borehole stations will increase to 70.

In November 2011, the first cable-based ocean bottom seismometer, the Marine Cable Hosted Observatory (MACHO), began operating in Taiwan. The MACHO has one seismic station located within the northeastern offshore area of Taiwan, with a cable line

35

length of 45 km, and hosts a three-component strong-motion accelerometer and a three component broadband seismograph (Hsiao et al., 2013). The MACHO is very expensive, and only one station is currently in operation. However, because the Philippine Sea plate subducts beneath the Eurasia plate of northeastern Taiwan and since many large earthquakes have occurred in this area in the past, the MACHO system is critical to the EEW system. The MACHO is capable of detecting seismic waves faster than inland stations.

All CWBSN waveforms are archived in CWB24 Format, shown in Appendix B. The CWB continuously records all seismic waveforms and archives into file every four minutes. Every day the CWB staff manually scan the continuous files and cut individual earthquake as a file. These files can be used for adjusting auto-picking parameters.

Figure 3-2 provides the system configuration of the CWBSN for a three-layer structure within the data-processing center used for the acquisition, integration, and application of real-time seismic signals. In the first layer, real-time seismic data streams are packaged and transmitted from field stations or external seismic networks then received by commercial software or Earthworm via various Internet Protocol (IP)-based networks. SMARTGeoHub and Scream software packages are used to receive real-time seismic waveforms from instruments made by Geotech and Güralp, respectively. Seismic waveforms from external seismic networks provided by IES and IRIS are received using the Earthworm modules IMPORT_ACK and SLINK2EW, respectively. In the second layer, an Earthworm cluster integrates seismic data streams from different seismic instruments and provides two types of seismic waveforms. One waveform type, WAVE_SERVERV, can store and provide seismic waveforms over a period of time and

36

is used for data displays and archives. The second waveform type, EXPORT_ACK, can provide data streams much faster than the previous one and is used for real-time data processing. For system backup, two computers running WAVE_SERVERV and three computers running EXPORT_ACK are operated in parallel. In the third layer, also called the application layer, several tasks are performed. These include EEW operation, the generation of products obtained from the earthquake catalog and the CMT, the maintenance of the seismic waveform data archive, and the display.

Via its modules and shared memory regions, the Earthworm system is designed for automatic seismic data processing (Johnson et al., 1995). Each module has specific tasks such as data acquisition, processing, and archiving. Adopting shared memory regions makes it convenient for each module to easily receive or broadcast messages such as waveform data, P-wave arrivals, hypocenter, and magnitude. Earthworm prepares seismic-related modules and is open source. Therefore, users can use existing modules or create new modules for specific purposes.

37

Figure 3-2. A schematic diagram of the data processing center. SP indicates short-period stations;

SM indicates strong-motion stations; BH indicates borehole stations; and OBS indicates cable-based ocean bottom seismic stations. “Ext” indicates stations operated by external institutions.

38

3.2 eBEAR System Configuration

Real-time data streams retrieved from seismic stations are integrated in Earthworm system. In order to process data effectively, three Earthworm modules (PICK_EEW, TCPD, and DCSN) were developed in this study. Appendix A. shows the configured files of the three modules. The configured files describe names of shared memories for data in and out. They also defined specific parameters and provide some detail information for the three modules. First, the PICK_EEW module is in charge of detecting onsets of P-wave arrival and estimating Pd andτc values. Thus, the configured file of PICK_EEW provides station information including location, gain factor and specific auto-picking parameters for each channel. Second, the TCPD module is in charge of locating earthquake and estimating magnitude. Thus, the configured file of TCPD provides parameters for associating P-wave arrivals, P-wave velocity model, and other related parameters. Third, the DCSN module is in charge of decision making and delivering EEW information. Thus, the configured file of DCSN provides criteria for alarm release, information of MySQL database, directories for storing XML-formatted file.

The velocity model used in the TCPD module is a one-dimension continuous velocity model. The equation can be shown by:

V(D) = G0 + G ╳ D

where V represented velocity is a function of depth D, G0 and G are constants. The unit of V and D are km/s and km, respectively. In this study, an averaged one-dimension velocity model was obtained from averaging three-dimension velocity model (Wu et al., 2009).

For depth shallower than 40 km:

V(D) = 5.103 + 0.067 ╳ D

39

For depth deeper than 40 km:

V(D) = 7.805 + 0.005 ╳ D

For magnitude estimation, the regression equations are represented as follows:

For BroadBand Sensor:

Earthquake magnitude is estimated by obtaining an average for each MPd value from each seismic station, following section 2.4.2 in this dissertation. Figure 3-3 provides the hardware configuration of the eBEAR system. For system backup, we designed two parallel EEW units, EEW1 and EEW2 that run the same procedure and data for generating earthquake messages. When an earthquake occurs, both EEW1 and EEW2 send earthquake messages to the system running the DCSN module. Only the first system sending the earthquake message is activated within the DCSN module. After receiving an earthquake message, the DCSN module writes an XML-formatted file onto the EEW server used to broadcast EEW reports to end users; then, to warn the end user, a display program pops up on the computer screen.

40

Figure 3-3. A flowchart of the algorithms designed for the PICK_EEW module. The rectangle represents different computers.

3.3 Offline Test

To calibrate the eBEAR system, an offline test was implemented in this study. From 2012 to 2013, we collected recorded seismic waveforms with magnitudes greater than 4.0, depths less than 40 km, and epicenters within 40 km of the coastline of Taiwan based on the upgraded CWBSN. A total of 154 seismic events, including four events with magnitudes between 6.0 and 6.5, were used in the test. The results, including earthquake locations and magnitudes, were compared to the CWB earthquake catalog. The reporting time of the offline test (defined as the time the EEW report is issued following the event origin time) does not include a telemetry delay of within 2 s. Figure 3-4 provides the offline performance of the eBEAR system in comparison with the results from the CWB catalog. The average errors for epicenter and focal depth locations are 4.2 and 5.3 km, respectively. The standard deviation of the local magnitude is 0.3 units. The average reporting time is 14.7 s. Some events located in southwestern Taiwan with relatively higher station density and coverage may be reported within 10 s. The offline results are acceptable for EEW purposes and suggest three points. First, the two-step method for determination of the epicenter and focal depth is suitable for a complicated tectonic environment such as Taiwan. Second, using a Pd value within a 3 s time window

41

following P-wave arrival is useful for measuring the size of moderate-sized earthquakes with magnitudes ranging from 4.0 to 6.5. Third, when an earthquake occurs in an area with a relatively higher station density and coverage, the number of updating earthquake messages quickly increases within the eBEAR system. For this type of event, the system is able to obtain a third earthquake message (an EEW report) within a short period of time.

For further discussion of the reporting time, Figure 3-5 provides the relationship between the reporting time and the station coverage gap. For most inland events with a station coverage gap generally less than 150°, reporting can occur within 15 s. On the other hand, for offshore events the reporting times may take more than 20 s when the station coverage gap is greater than 200°. The results indicate that currently the station coverage gap is a key factor for controlling the reporting time of the eBEAR system.

42

Figure 3-4. A comparison between the offline test and the CWB published catalog, as follows: (a) the epicenters, (b) the magnitudes, (c) the focal depths, and (d) the reporting time of the offline test. Open circles represent earthquake locations obtained from the published CWB catalog.

Solid circles represent earthquake locations from the eBEAR system.

43

Figure 3-5. The relationship between reporting time and station coverage gap.

3.4 Online Performance

For an online system comparison between the VSN and eBEAR systems, we collected online operating performance data from January to March of 2014. Figure 3-6 indicates that the eBEAR system had no missed events and that determinations of location were better than for the VSN system. For inland earthquakes, both systems had location errors less than 10 km. On the other hand, for offshore earthquakes, the VSN system missed two events and displayed larger location errors of approximately 50–100 km. On average, the epicenter errors of the eBEAR and VSN systems are 10.0 and 16.2 km, respectively. When considering depth determinations, the VSN displayed better results than the eBEAR system because the VSN system used both P- and S-wave arrivals, whereas the eBEAR system only used P-wave arrivals. For magnitude determinations, the

44

eBEAR system yielded a smaller standard deviation (0.2) compared to the VSN system,

with a standard deviation of 0.5, shown in Figure 3-7. The solid circles represent the events detected by both systems; the open circles represent the events only detected by the eBEAR system. If we only compare the solid circles, it also shows the eBEAR system has

better magnitude estimations than the VSN system. In the comparison of reporting times, Figure 3-8 indicates that almost every earthquake processed by the eBEAR system displayed an earlier reporting time. On average, the eBEAR system shortens reporting times by 3.2 and 5.5 s, compared to the VSN system for inland and offshore earthquakes, respectively. Because the eBEAR system contains 149 seismic stations distributed in a smaller station coverage gap and because station locations are denser than the VSN without considering the influence of the station distribution. Figure 3-9 provides warning times to target areas in metropolitan Taipei. Warning time is defined as the time between the reporting time and the arrival of the S wave. The eBEAR system provides a longer warning time than the VSN system. For the eastern offshore area of Taiwan, the eBEAR system can provide a warning time that is 5 s longer, on average, than the VSN system.

The major reason is that by adding the MACHO system and the YOJ station into the seismic network, the eBEAR system has a smaller station coverage gap. In addition, for events with the approximate locations of the 1999 Mw 7.6 Chi-Chi earthquake and the

45

2002 Mw 7.1 eastern Taiwan offshore earthquake, the Taipei metropolitan area would have had a warning time of 26 and 15 s, respectively.

Figure 3-6. Location estimations as compared to online performance between the eBEAR and VSN systems, as follows: (a) the epicenter distribution of the CWB catalog and events of the EEW alarms & missed alarms from the eBEAR system, and (b) the epicenter distribution of the CWB catalog and events of the EEW alarms & missed alarms from the VSN system. Open circles represent earthquake locations from the published CWB catalog. Solid circles represent earthquake locations from the EEW system. Open triangles represent missing reports from the EEW system.

46

Figure 3-7. Magnitude estimations as compared to online performance between the eBEAR and VSN systems, as follows: (a) results from the eBEAR system and (b) results from the VSN system. The solid circle represents the events detected by the eBEAR and VSN systems. The open circle represents the events only detected by the eBEAR system. The solid line represents the 1:1 line. Dashed lines represent one standard deviation.

Figure 3-8. The reporting time comparison for online performance between the eBEAR and VSN systems, as follows: (a) results from the eBEAR system using 149 stations within the CWBSN, and (b) results from the VSN system using 109 RTD stations.

47

Figure 3-9. Warning time comparisons for online performance between the eBEAR and VSN systems, as follows: (a) results from the eBEAR system using 149 stations, and (b) results from the VSN system using 109 RTD stations. The solid square represents the target area for obtaining warnings. Open circles represent epicenters. The number over the open circle is the warning time, defined as the time between the reporting time and the arrival of the S-wave. If the warning time value is negative, the target area has no warning time.

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

48

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

To avoid this kind of false alarm, we separated the offline and online systems. The second