URBAN EARTHQUAKE RAPID RESPONSE AND EARLY WARNING SYSTEMS

First European Conference on Earthquake Engineering and Seismology (a joint event of the 13 th ECEE & 30 th General Assembly of the ESC) Geneva, Switzerland, 3-8 September 2006 Paper Number: Keynote Address K4 URBAN EARTHQUAKE RAPID RESPONSE AND EARLY WARNING SYSTEMS Mustafa ERDIK 1 SUMMARY Technological advances in seismic instrumentation and telecommunication permit the implementation of real-time rapid response and early warning systems. During large earthquakes, such systems are capable of providing from a few seconds to a few tens of seconds of warning before the arrival of strong ground shaking and enable quick reports about the damage estimates to determine where emergency response is most needed. An earthquake early warning and rapid response system can provide the critical information needed to minimize loss of lives and property, and to direct rescue operations As part of the preparations for the future earthquake in Istanbul a Rapid Response and Early Warning system in the metropolitan area is in operation. For the Early Warning system ten strong motion stations were installed as close as possible to the fault zone. Continuous on-line data from these stations via digital radio modem provide early warning for potentially disastrous earthquakes. Considering the complexity of fault rupture and the short fault distances involved, a simple and robust Early Warning algorithm, based on the exceedance of specified threshold time domain amplitude levels is implemented. The encrypted early warning signals will be communicated to the respective end users through a service provider company. The users of the early warning signal will be power and gas companies, nuclear research facilities, critical chemical factories, subway system and several high-rise buildings. Depending on the location of the earthquake (initiation of fault rupture) and the recipient facility the alarm time can be as high as about 8s. For the rapid response system one hundred 18 bit-resolution strong motion accelerometers were placed in quasi-free field locations (basement of small buildings) in the populated areas of the city, within an area of approximately 50x30km, to constitute a network that will enable early damage assessment and rapid response information after a damaging earthquake. Early response information is achieved through fast acquisition and analysis of processed data obtained from the network. A shake map and damage distribution map (using aggregate building inventories and fragility curves) will be automatically generated using the algorithm developed for this purpose. The shake and damage maps will be conveyed to the governor s and mayor s offices, fire, police and army headquarters within 3 minutes using radio modem and GPRS communication. 1. BACKGROUND AND INTRODUCTION Technological advances in seismic instrumentation and telecommunication permit the implementation of realtime rapid response and early warning systems. During large earthquakes, such systems are capable of providing from a few seconds to a few tens of seconds of warning before the arrival of strong ground shaking and enable quick reports about the damage estimates to determine where emergency response is most needed. 1 Bogazici University, Kandilli Observatory and Earthquake Research Institute-Department of Earthquake Engineering, Cengelkoy 34684 Istanbul, TURKEY Email: erdik@boun.edu.tr 1

Earthquake Early Warning in urban and industrial areas allows for emergency shutdown of systems susceptible to damage such as power stations, transportation, computer centers and telephone systems, as well as critical facilities and industry. An earthquake early warning and rapid response system can provide the critical information needed to minimize loss of lives and property, and to direct rescue operations. [Kanamori et al., 1997; Teng et al, 1997; Allen and Kanamori, 2003; Lee and Espinosa-Aranda. 2003]. The implementation of urban earthquake early warning systems are proposed by United Nations - International Strategy for Disaster Reduction (UN-ISDR) (http://www.unisdr.org/unisdr/genwarning.htm) and by USGS- ANSS-Advanced National Seismic System (http://www.anss.org/). The basic idea of an earthquake early warning system was proposed more than one hundred years ago by Cooper (1868) for San Francisco, California. A modern approach of an earthquake early warning system for a seismic computerized alert network was proposed by Heaton (1985). In Japan, a method of earthquake early warning was employed by its bullet train operation, Nakamura (1989). Currently such systems are either implemented or in construction or in planning stage in Mexico, Romania, California, Japan, Taiwan, Italy, Turkey and Greece. Early warning systems, currently in operation in Bucharest (http://www.infp.ro/publications/ews.htm) and Mexico (http://www.gfz-potsdam.de/ewc98/abstract/espinosa.html), have the capability of issuing an earthquake early warning in advance, by relying on significant distances between the source and the populated urban regions. The close proximity to fault in Istanbul precludes such an approach. However, by using a more direct approach, few seconds warning prior to the onset of damaging ground motion is possible. Potential impact of large earthquakes on urban societies can be reduced by timely and correct action after a disastrous earthquake. Modern technology permits measurements of strong ground shaking in near real-time for urban areas exposed to earthquake risk. The assessments of the distribution of strong ground motion (Shake Maps), building damage and casualties can be made within a short time after an earthquake. Earthquake ground motion measurement and data processing systems designed to provide this information are called Rapid Response Systems. The rapid response systems implemented in California: USGS, Caltech and CDMG TriNet ShakeMapshttp://www.trinet.org/shake/; Caltech-USGS Broadcast of Earthquakes (CUBE) Systemhttp://www.gps.caltech.edu/~bryant/cube.html; UC Berkeley Seismological Laboratory and USGS Menlo Park (REDI)- www.seismo.berkeley.edu/seismo/redi; in Taiwan (Earthquake Rapid Reporting System of Taiwan Central Weather Bureau, CWB- http://www.earth.sinica.edu.tw/cdr/iaspei/data/cwb/rapid.html and in Japan (Real-time Earthquake Assessment Disaster System in Yokohama READY http://www.city.yokohama.jp/me/bousai/eq/index.html can be listed as example applications. The reduction of casualties in urban areas immediately following an earthquake can be improved if the location and severity of damages can be rapidly assessed by the information from Rapid Response Systems. Emergency management centers of both public and private sector with functions in the immediate post-earthquake period (i.e. SAR, fire and emergency medical deployments) can allocate and prioritize resources to minimize the loss of life. The emergency response capabilities can be significantly improved to reduce casualties and facilitate evacuations by permitting rapid and effective deployment of emergency operations. The Rapid Response data should possibly be linked with incident command and standard emergency management systems to increase effectiveness. Ground motion data related with power transmission facilities, gas and oil lines and rapid transportation systems also allows for a rapid preliminary assessment possible damages to avoid secondary risks. Water, wastewater and gas utilities can locate the sites of possible leakage of hazardous materials and broken pipes. The prevention of gas-related damage in the event of an earthquake requires understanding of damage to pipeline networks and prompt shut-off of gas supply in regions of serious damage. The SIGNAL system developed by Tokyo Gas uses 356 SI (instrumental intensity) sensors in Tokyo metropolitan area to monitor seismic motion to shut of the gas supply to damaged areas (http://www.tokyo-gas.co.jp/techno/stp/97c1_e.html). 2

2. EARTHQUAKE EARLY WARNING SYSTEMS An earthquake early warning system requires seismic stations close to the source of earthquakes and continuous communication between the seismic stations and a central processing station. The early warning systems utilize the fact that seismic waves propagate slower than electromagnetic waves that are used for ground motion data transmission and communications of the warning. The maximum pre-warning times in areas with well-defined fault zones can be as high as 60 to 80 seconds (Mexico City). In other areas, where the fault zones are close or the active faults are not known, the warning time may be less than 5 seconds. Regional Earthquake Early Warning Systems consist of regional ground motion detection (weak or strong) networks spaced to tens of kilometers and located close to the earthquake source zone(s). Ground motion data (online time histories or processed ground motion parameters) is transmitted to the central data processing facility where the source parameters and/or amplitude of ground shaking are computed and converted to information and early warning signals are communicated to end users. Earthquake Early Warning Systems in Mexico City, Japan, Taiwan and Istanbul can be considered Regional. Location (or Facility)-specific Earthquake Early Warning Systems are designed to serve to the needs of specific critical facilities, such as nuclear power plants or LNG storage units. The ground motion detection (weak or strong) units spaced within few kilometers and the early warning alarm is communicated on the basis of the exceedance of selected time domain ground motion parameters. An Early Warning System (EWS) forewarns an urban area (or facility) of the forthcoming strong shaking, normally within a few seconds to a few tens of seconds before the arrival of the destructive part of the strong ground motion. Even a small time window can provide opportunities to automatically trigger measures such as shutdown computers; remote electrical power; shutdown high precision facilities; shutdown airport operations; shutdown manufacturing facilities; shutdown high energy facilities; shutdown gas distribution; alert hospital operating rooms; open fire station doors; start emergency generators; stop elevators in a safe position; shutoff oil pipelines; issue audio alarms; shutdown refineries; shutdown nuclear power plans; shutoff water pipelines; maintain safe-state in nuclear facilities [Goltz, 2002 and Harben, 1991]. Every EWS consists of the following four components: 1. a monitoring system composed of various sensors, 2. a real-time communication link that transmits data from the sensors to a computer, 3. a processing facility that converts data to information, and 4. a system that issues and communicates the warning. For earthquake early warning systems, the output of the system can be an estimate of magnitude and location of the event but equally well a projection of the expected acceleration or intensity at specific sites. Earthquake Early Warning Systems can operate either on the objective of real-time assessment of source parameters (real-time seismology) or exceedance of selected ground motion parameter thresholds. 2.1 Rapid Assessment of Earthquake Magnitude and Location The main parameters of an earthquake (epicenter, time of origin, magnitude, focal mechanism, and amplitude of ground shaking) may take about one minute from the time an earthquake has been detected for reliability estimates [Wu, et al 1998]. The rapid location can be achieved in the ten seconds time window immediately following the first P-arrival. The rapid determination of the earthquake magnitude would be more difficult because the shear wave trains may not be recorded completely within this time window, and, more importantly, since a damaging event is large, a moment magnitude or its equivalent must be developed for damage potential assessment [Wu, et al 1998]. In literature, Many researchers (e.g., Nakamura, 1988; Grecksch and Kumpel, 1997; Tsai and Wu, 1997; Allen and Kanamori, 2003) have tried to estimate magnitude from the initial portion of accelerograms, but large uncertainties are a major problem. The ElarmS methodology [Olson and Allen, 2005] was designed with the goal of predicting the distribution of peak ground shaking across the region affected by an earthquake before the beginning of significant ground motion at the epicenter. The first few seconds of the P-wave at the station and stations closest to the epicenter is used to estimate the magnitude of the earthquake and attenuation relations provide the predicted distribution of ground shaking as a function of distance from the epicenter. 3

2.2 Ground Motion Parameter Threshold Exceedance For earthquake early warning and alarm systems there is usually insufficient time to compute the hypocenter, focal parameters and the magnitude of an earthquake, as this time is needed for the more complex alarm decision making process. The benefits of an early warning system increase with increasing pre-warning time. Considering the complexity of fault rupture and the short fault distances involved, a simple and robust early warning algorithm, based on the exceedance of specified threshold time domain amplitude levels needs to be implemented. The band-pass filtered peak ground accelerations (PGA) and the cumulative absolute velocity (CAV-time integral of the absolute acceleration) can be compared with specified threshold levels. To declare the first early warning alarm, a simple algorithm can be implemented. When any acceleration or CAV (on any channel) in a given station exceeds specific threshold values it is considered a vote. Whenever we have 2 or 3 (selectable) station votes within selectable time interval, after the first vote, the first alarm is declared [Erdik, et al, 2003b]. More complex algorithms based on artificial neural networks (ANN) can be used [Boese, et al, 2003]. ANN approach considers the problem of earthquake early-warning as a pattern recognition task. The seismic patterns can be defined by the shape and frequency content of the parts of accelerograms that are available at each time step. ANN can extract the engineering parameters PGA, CAV from these patterns, and map them to any location in the surrounded area. 3. EARTHQUAKE RAPID RESPONSE SYSTEMS Earthquake Rapid Response Systems provide rapid assessment of shake-maps, location and severity of damages and the distribution of casualties in urban areas immediately following an earthquake. The Earthquake Rapid Response Systems have the objective of providing: Reliable information for accurate, effective characterization of the shaking and damage by other rapid post-earthquake maps (Shake, Damage and Casualty maps) for rapid response; Recorded motion for post-earthquake performance analysis of structures; Empirical basis for long-term improvements in seismic microzonation, seismic provisions of building codes and construction guidelines; and to improve the understanding of earthquake generation at the source and seismic wave propagation. Current earthquake rapid response system methodologies have different approaches to measure and estimate the ground shaking of earthquake area, in order to estimate the intensity and damage maps. The first approach uses the seismic source parameters (hypocenter, magnitude, intensity) in order to compute the ground shaking and potential damage. The second approach use the direct engineering parameters such as peak ground acceleration (PGA), peak ground velocity (PGV), spectra displacements (SD) maps to compute the potential damage. For an earthquake rapid response system a large number of seismic stations (strong motion instruments) is needed, which are distributed uniformly over an urban area. The stations may not need a continuous communication with a central station. After triggering by earthquake, the stations will send electronic messages (SMS, E-mail, etc.) to the central station a few seconds after the end of an earthquake. The messages sent may contain information about the peak ground acceleration and the spectrum intensity, which will be the basis for the automatic preparation of damage maps. Both systems transmit this critical information electronically to emergency response agencies and other relevant governmental agencies. 3.1 Shake Maps ShakeMap is a representation of ground shaking produced by an earthquake. The information it presents is different from the earthquake magnitude and epicenter that are released after an earthquake because ShakeMap focuses on the ground shaking produced by the earthquake, rather than the parameters describing the earthquake source. So, while an earthquake has one magnitude and one epicenter, it produces a range of ground shaking levels at sites throughout the region depending on distance from the earthquake, the rock and soil conditions at sites, and variations in the propagation of seismic waves from the earthquake due to complexities in the structure of the Earth's crust (http://quake.wr.usgs.gov/recent/shaking.html). 4

Peak horizontal acceleration at each station is contoured in units of %g. The peak values of the vertical components are not used in the construction of the maps because the regression relationships used to fill in data gaps between stations are based on horizontal peak amplitudes. The contour interval varies greatly and is based on the maximum recorded value over the network for each event. For moderate to large events, the pattern of peak ground acceleration is typically quite complicated, with extreme variability over distances of a few km. This is attributed to the small scale geological differences near the sites that can significantly change the high-frequency acceleration amplitude and waveform character. An example of the peak ground acceleration map of Hector Mine Earthquake of Oct. 15, 1999 given in Figure 1. Figure 1. Peak Acceleration Shake Map for Hector Mine Earthquake of Oct. 15, 1999; http://quake.wr.usgs.gov Peak velocity values are contoured for the maximum horizontal velocity (in cm/sec) at each station. As with the acceleration maps, the vertical component amplitudes are disregarded for consistency with the regression relationships used to estimate values in gaps in the station distribution. Typically, for moderate to large events, the pattern of peak ground velocity reflects the pattern of the earthquake faulting geometry, with largest amplitudes in the near-source region, and in the direction of rupture (directivity). Differences between rock and soil sites are apparent, but the overall pattern is normally simpler than the peak acceleration pattern. Response spectra portray the response of a damped, single-degree-of-freedom oscillator to the recorded ground motions. This data representation is useful for engineers determining how a structure will react to ground motions. 5

ShakeMap spectral response maps are made for the response at three UBC reference periods: 0.3, 1.0, and 3.0 seconds. For each station, the value used is the peak horizontal value of 5% critically damped pseudoacceleration. 4. EXISTING EARTHQUAKE EARLY WARNING AND RAPID RESPONSE SYSTEMS 4.1 Urgent Earthquake Detection and Alarm Systems (UREDAS) in Japan In the mid 1950s Japanese National Railways started the installation of simple alarm seismometers along their railway lines. Later, with the operation of the bullet trains, the alarm system was improved by the installation of seismometers at every 20 km along the lines. The stations issue a warning whenever a preset level of horizontal ground acceleration (40 gal) is exceeded [Nakamura, 1989]. Trains in the vicinity to the respective alarm station are automatically slowed down or stopped in order to avoid derailments. The contemporary realtime earthquake disaster prevention, earthquake detection and alarm systems (UrEDAS) used for railways in Japan consist of 30 strong motion instruments. The main feature of the system is utilizing the information from P-wave data. Systems for different railways have been in operation since 1983 [Nakamura, 1989]. UrEDAS detects initial P-wave motions, estimates epicentre azimuth and magnitude, calculates epicentral distance and local depth within about 3 seconds. 4.2 Seismic Alert System of Mexico City Most of the large earthquakes, which are likely to cause damage in Mexico City have their source in the subduction zone of the Pacific coast at a distance of about 320 km [Espinosa-Aranda et al., 1995]. The Seismic Alert System (SAS) is a public earthquake early warning system developed with the sponsorship of the Mexico City government authorities. The SAS is operating since August 1991. The seismic alert system for Mexico City consists of four parts: the seismic detection system, a dual telecommunications system, a central control system and a radio warning system for users. The seismic detector system consists of 12 digital strong motion field stations located along a 300 km stretch of the Guerrero coast, arranged 25 kilometers apart (Figure 2). Each seismic continually processes local seismic activity which occurs within a 100 km radial coverage area around the station. The Central Control System continually receives information on the operational status of the field stations and communication relay stations, as well as the actual detection of an earthquake in progress. Information received from the stations is processed automatically to determine magnitude and is used in the decision to issue a public alert. Figure 2. Seismic Alert System for Mexico City 6

The system is capable of generating warning signals about 60 sec. average in advantage to the S waves first arrivals in Mexico City, when it detects strong earthquakes occurring 280 km faraway, in the Guerrero Gap. The SAS earthquake warning signal disseminates via local AM/FM commercial radio stations allowing the operation of the audio alerting mechanisms to residents of Mexico City, public schools, government agencies with emergency response functions, key utilities, public transit agencies and some industries. Public and private entities are equipped with specially designed radio receivers to obtain the SAS alert. Triggering prevention procedures, designed in each public school within the earthquake hazard reduction program applied by the Secretariat of Public Education in the valley of Mexico, since September, 1985 (Espinosa-Aranda et al., 1995). Up to June 2000, after nine years of continuous operation, the SAS has successfully detected more than 755 earthquakes with magnitude from M4 to M7.3 [Espinosa-Aranda et al., 2003]. Until now the greatest seismic event detected by the SAS was the September 14, 1995, M7.3 Copala earthquake. In a live test that checked the whole system, the SAS was activated and a general warning signal was issued in Mexico City, 72 sec. prior to the arrival of strong ground motion effects (http://www.cires.org.mx). 4.3 Istanbul Earthquake Rapid Response and the Early Warning System Istanbul faces a significant earthquake hazard and risk as illustrated by the recently developed earthquake risk scenario for Istanbul [Erdik, et al. 2003a]. The tectonic setting showing the location of the Main Marmara Fault and EMS98 intensity distribution that would result from a moment magnitude Mw=7.5 scenario earthquake is provided in Figure 3. The distribution of building damage (complete) and the life losses that would result from the same scenario earthquake are illustrated respectively in Figures 4 and 5. Figure 3. Scenario Earthquake-Based Intensity Distribution 7

Figure 4. Distribution of Building Damage (Complete Damage) Figure 5. Distribution of Human Life Losses To assist in the reduction of losses in a disastrous earthquake in Istanbul a dense strong motion network is established [Erdik, et al. 2003b]. To provide earthquake rapid response information one hundred strong motion accelerometers were placed in populated areas of Istanbul, within an area of approximately 50x30km, to constitute a network that will enable rapid shake map and damage assessment after a damaging earthquake (Figure 6). After triggered by an earthquake, each station will process the streaming strong motion to yield the spectral accelerations at specific periods and will send these parameters in the form of SMS messages to the main data center through available GSM network services. A shake map and damage distribution will be automatically generated. The shake and damage maps will be available on the Internet and will also be pushed to several end users. 8

Figure 2. Istanbul Rapid Response Stations For the generation of earthquake early warning information ten strong motion stations were located as close as possible to the Marmara Fault (Figure 7). The continuous on-line data from these stations will be used to provide near-real time warning for emerging potentially disastrous earthquakes. Figure 7. Istanbul Early Warning Stations 9

Furthermore about 40 strong motion recorder units were placed on critical engineering structures in addition to the already instrumented structures in Istanbul http://www.koeri.boun.edu.tr/depremmuh/stronmotion.htm). All together this network and its functions is called Istanbul Earthquake Rapid Response and Early Warning System (IERREWS). The system is designed and operated by Bogazici University with the logistical support of the Governorate of Istanbul, First Army Headquarters and Istanbul Metropolitan Municipality. The construction of the system is realized by the GeoSig and EWE (Switzerland) consortium. Data transmission for the Rapid Response System is provided by AVEA GSM service provider. 4.3.1 Istanbul Rapid Response System The Istanbul Rapid Response System has the objective of providing: Reliable information for accurate, effective characterization of the shaking and damage by other rapid post-earthquake maps (Shake, Damage and Casualty maps) for rapid response; Recorded motion for post-earthquake performance analysis of structures; Empirical basis for long-term improvements in seismic microzonation, seismic provisions of building codes and construction guidelines; and Seismological data to improve the understanding of earthquake generation at the source and seismic wave propagation. The Rapid Response System satisfies the COSMOS (The Consortium of Organizations for Strong-Motion Observation Systems) Urban Strong-Motion Reference Station Guidelines (www.cosmos-eq.org) for the location of instruments, instrument specifications and housing specifications. As such, for the location of instruments the results of deterministic earthquake hazard/risk assessment for Istanbul is used in consideration of Highest likelihood of shaking (Short and Long Period), High probability of damage (Damage Distribution Maps) and High probability of casualties (Casualty Distribution Maps) The relative instrument spacing is about 2-3km which corresponds to about 3 wavelengths in firm ground conditions and more than 10 wavelengths for soft soils for horizontally propagating 1s shear waves. Strongmotion instruments are generally located at grade level in small and medium-sized buildings, such that the motion recorded corresponds to that on the ground in the surrounding area site geology at stations has been characterized in general terms. Certain stations have borehole data. New borehole surveys for other stations are being planned. For communication of data from the rapid response stations to the data processing center and for instrument monitoring a reliable and redundant GSM communication system (backed up by dedicated landlines and a microwave system) is used, on the basis of a protocol agreement with the AVEA GSM Service provider. The strong motion accelerographs utilized in the IERREW System have the following basic specifications Overall recording range: The strong motion instrumentation utilized in the IERREW system can record an acceleration range of +/-2g Full Scale with industry accepted specifications. Recorder dynamic range: The instrumentation has 18-bit (dial-up stations) or 24-bit (on-line) resolution. The least significant bit (LSB) resolution is 0.015 mg. Noise floor: The noise level is less than 0.02 mg RMS in the frequency range of 0-40 Hz. The instruments provide on-site recording for 2 hours or more of strong motion recording. Timing accuracy: Within 1/10th of a sampling interval of GPS absolute time (UTC). Sample rate: 200 samples per second (5 ms sampling interval) with adequate antialias filtering (filter corner at 80% of the Nyquist frequency, and down by 100 db at the Nyquist). Triggering: Nominal trigger level is 1-5 mg within a pass band of 0.1 to 12 Hz. The actual trigger levels are established by site conditions. Once triggered, the recorder shall stay triggered for at least 30 seconds after the last occurrence of acceleration over 5 mg. All of the instruments were calibrated in the laboratory using a air-bed electro-magnetic shaker for calibration of the sensitivity constants of the sensors. Additional bi-directional tilt tests at site were conducted for confirmation. In normal times the rapid response stations are interrogated (for health monitoring and instrument monitoring) on regular basis. After triggered by an earthquake, each station processes the streaming three-channel strong motion data to yield the spectral accelerations at specific periods, 12Hz filtered peak ground acceleration and peak ground velocity and sends these parameters (in the form of SMS messages) at every 20s directly to the main data center through the GSM communication system. The main data processing center is located at the 10

Department of Earthquake Engineering - Kandilli Observatory and Earthquake Research Institute of Bogazici University (KOERI-BU). A secondary center located at the Seismological Laboratory of the same Institute serves as a redundant secondary center that can function in case of failure in the main center. Shake and damage distribution maps will be automatically generated at the data centers after the earthquake and communicated to the end users within 5 minutes. Full-recorded waveforms at each station can be retrieved using GSM and GPRS modems subsequent to an earthquake. For the generation of Rapid Response information two methodologies based on spectral displacements and instrumental intensities are used. These methodologies are coded into specific computer programs similar to HAZUS (http://www.fema.gov/hazus). Both of the methodologies essentially rely on the building inventory database, fragility curves and the methodology developments in the Istanbul Earthquake Risk Assessment Study conducted by the Department of Earthquake Engineering of Bogazici University [Erdik, et al. 2003a]. For the computation of input ground motion parameters, spectral displacements obtained from the SMS messages sent from stations will are interpolated to determine the spectral displacement values at the center of each geo-cell using two-dimensional splines. The earthquake demand at the center of each geo-cell is computed using these spectral displacements. The instrumental intensity at each the center of each geo-cell is computed as a function of short-period spectral acceleration. Using the response spectra and the instrumental intensities the building damage and the casualties are computed separately by using the spectral-displacement based and intensity based fragility curves. The computations are conducted at the centers of a 0.01 x 0.01 grid system comprised of geocells (1120 m x 830 m) size. The building inventories (in 24 groups) for each geocell together with their spectral displacement and intensity based fragility curves are incorporated in the software. Example of a building damage distribution map that results from a randomly simulated strong motion data set is provided in Figure 8. In addition a list of districts and sub-districts with number of completely damaged buildings above a selectable number is prepared. The building damage distribution map is communicated to the concerned emergency response centers (Istanbul Governorate, Istanbul Municipality and First Army Headquarters) through digital radio modem and GPRS communication systems are used (Figure 9). The data are also made available on the Internet after an earthquake Figure 8. Example of building damage map that results from a randomly simulated strong motion data 11

Figure 9. Communication of the Rapid Response Information 4.3.2 Istanbul Earthquake Early Warning System The Early Warning part of the IERREWS, ten strong motion stations were located as close as possible to the Great Marmara Fault in on-line mode. Continuous telemetry of data between these stations and the main data center is realized with digital spread spectrum radio modem system involving repeater stations selected in the region. Considering the complexity of fault rupture and the short fault distances involved, a direct (engineering) early warning algorithm based on the exceedance of specified threshold time domain amplitude levels is implemented. The early warning information (consisting three alarm levels) will be communicated to the appropriate servo shut-down systems of the recipient facilities, which will automatically decide proper action based on the alarm level. Depending on the location of the earthquake (initiation of fault rupture) and the recipient facility the alarm time can be as high as about 8s. Currently two types of early warning algorithms were implemented. The steps associated with each algorithm are as follows: Algorithm Based on Exceedance of Filtered PGA Treshold All online acceleration data from all stations will be low-pass filtered at selectable frequencies of 12 and 25 Hz. When any acceleration (on any channel) in a given station exceeds a selectable first threshold value (20 mg) it will be considered a vote Whenever we have 3 (selectable) station votes within a selectable time interval of (5s) after the first vote it will be declared the first alarm. After the first alarm, whenever we have 3 (selectable) votes for the second acceleration threshold value (50 mg) within selectable time interval of (5s) after the first vote it will be declared the second alarm. After the second alarm, whenever we have 3 (selectable) votes for the third acceleration threshold value (100 mg) within selectable time intervals of (5s) after the second vote it will be declared the third alarm. Algorithm Based on Cumulative Absolute Velocity (CAV) CAV (t) = Integral from 0 to t [abs (a).dt ] ( g-sec) (Figure 10) The CAV from acceleration data will be computed for only those 1s intervals where PGA is greater than 3mg. When any CAV (on any channel) in a given station exceeds a selectable first threshold CAV value (20 mg.s) it will be considered a vote. 12

Whenever we have 3 (selectable) votes for the first threshold CAV value within selectable time interval of (5s) after the first vote it will be declared the first alarm. After the first alarm, whenever we have 3 (selectable) votes for the second threshold CAV value (40 mg.s) within selectable time intervals of (5s) after the first vote it will be declared the second alarm. After the second alarm, whenever we have 3 (selectable) votes for the third CAV threshold value (70 mg.s) within selectable time intervals of (5s) after the second vote it will be declared the third alarm. Figure 10. Cumulative Absolute Velocity Early Warning algorithms based on Pattern Recognition (Neural Network) were developed by Dr. Maren Böse of University of Karlsruhe [Böse et.al, 2003]. In this methodology the seismic patterns are defined by the shape and frequency content of the parts of accelerograms that are available at each time step. From these, parameters relevant to seismic damage, such as peak ground acceleration (PGA), peak ground velocity (PGV) and response spectral amplitudes at certain periods are estimated using Artificial Neural Networks (ANN). The pattern recognition technique is combined with an additional rule-based system in order to detect inconsistencies between ground motion estimations and measurements. This combination provides a reliable and accurate system for early-warning that is demanded by its huge social and economic impact 4.4 Earthquake Rapid Reporting and Early Warning Systems in Taiwan Earthquake Rapid Reporting and Early Warning Systems in Taiwan uses a real-time strong-motion accelerograph network that currently consists of 82 telemetered strong-motion stations distributed across Taiwan, an area of 100 km x 300 km and monitored by Taiwan Central Weather Bureau (CWB). Each station has 3-component force-balanced accelerometers. The rapid reporting system can offer information about one minute after an earthquake occurrence. Information includes earthquake location, its magnitude and shaking maps of Taiwan area. Rapid damage assessment of the system is under development. By applying sub-network method approach, the early warning part of the system achieves earthquake reporting time about 20 sec. It will offer effective earthquake early warning for metropolitan areas located more than one hundred km from the epicenter [Tsai et al, 1997; Teng et al., 1997; Wu et al., 1998; Wu et al., 1999; Shin and Teng, 2001; Wu and Teng, 2002]. Rapid Reporting System (RRS) Digital signals (of 3-channel strong-motion and 3-channel weak-motion) are continuously telemetered to the headquarters of the CWB. Whenever the pre-specified trigger criteria were met, the digital waveforms are stored in the memory and are automatically analyzed by a series of programs [Wu et al., 1998]. The results were immediately disseminated to governmental emergency response agencies electronically. Results include earthquake location, local magnitude and intensities. Generally, the RRS of the CWB can offer information about one minute after an earthquake occurrence. 13

4.5 Early Warning System for Bucharest The Romanian capital Bucharest faces a significant earthquake hazard with a 50% chance for an event in excess of 7.6 moment magnitude every 50 years. An Early Warning System (EWS) based on the travel time differences between the primary P-wave and the destructive S-wave allows a warning time of about 25 s. Peculiarities of the Romanian intermediate depth seismicity such as the stationary epicenters and the stability of radiation patterns, and a line-of-sight connection between the epicentral area and the capital allow to design a simple and robust EWS. Early Warning System (EWS) for the capital city of Bucharest is designed and operated by a group of civil engineers and seismologists from the National Institute of Earth Physics (NIEP) in Romania and Karlsruhe University in Germany. Since the source mechanisms are extremely stable for larger and smaller events, a projection of the level of ground motion to be expected in Bucharest can be based on the amplitude of the epicentral P-wave [Wenzel, et al. 1999]. 4.6 Rapid Response and Disaster Management System in Yokohama, Japan Within Yokohama city 150 accelerographs are installed in free-field stations for ground shaking monitoring. Spacing between stations is about 2 km. A high precision digital accelerograph is used to record weak to very strong ground motion. In addition to the accelerographs at ground surface, borehole accelerometers are installed at 9 stations for liquefaction hazard assessment. All of these stations are connected to three observation centers, the disaster preparedness office of the city hall, the fire department office of the city and Yokohama City university, by the high-speed telephone lines. At 18 stations, the backup communication system by satellite is available. The full operation of the monitoring system started on May, 1997 [Midorikawa, 2004]. When an accelerograph is triggered by an earthquake, the station computes ground-motion parameters such as the instrumental seismic intensity, peak amplitudes, predominant frequency, total power, duration and response spectral amplitudes. These parameters are automatically reported to the centers. On receiving more than 10 reports in a prescribed time interval, the centers activate alert systems. The seismic intensity data is conveyed to the city officials by the pager, and the intensity map of the city is drawn within a few minutes after the earthquake. The map is immediately open to the public through the Internet (www.city.yokohama.jp/me/bousai/eq/index.html) and local cable TV. The map is utilized as the earliest information for disaster management. The ground motion data from the stations are used for the real-time seismic hazard and risk assessment. The assessed items are ground motion, liquefaction and building damage. Operation of the assessment system started on June, 1998. In the mapping of ground motion hazard, the city is divided into cells of 50 x 50 m size. The hazard and risk maps are created within twenty minutes after the earthquake, almost in real-time, so that the maps can be used for selecting strategy of emergency response activities [Midorikawa, 2004]. Acknowledgement This paper essentially represents a compilation of several studies jointly authored by Dr.Oğuz Özel, Dr.Yasin Fahjan, Prof.Nuray Aydınoglu, Prof.Eser Durukal, Eng. Hakan Alçık, Eng. Aydın Mert and Eng. Muzaffer Gül. Their contribution is gratefully acknowledged. 5. REFERENCES Allen, R. and Kanamori, H., (2003). The potential for earthquake early warning in South California, Science, 300, 786-789. Boese M., M. Erdik, F. Wenzel, (2003)Artificial Neural Networks for Earthquake Early-Warning, Proceedings AGU2003 Abstracts. Cooper, M.D., (1868). Editorial in San Francisco Daily Evening Bulletin, November 3. Erdik, M., Aydinoglu N., Fahjan Y., Sesetyan K., Demircioglu M., Siyahi B., Durukal E., Ozbey C., Biro Y., Akman H., and Yuzugullu O., (2003a). Earthquake Risk Assessment for Istanbul Metropolitan Area. Earthquake Engineering and Engineering Vibration, V.2, No.1, pp. 1-25. Erdik, M., Fahjan Y., Ozel O., Alcik H., Mert A., and Gul M., (2003b). Istanbul Earthquake Rapid Response and the Early Warning System. Bull. Of Earthquake Engineering, V.1, Issue 1, pp. 157-163. Espinosa-Aranda, J., Jiménez, A., Ibarrola, G., Alcantar, F., Aguilar, A., Inostroza, M., and Maldonado, S., (1995). Mexico City seismic alert system, Seism. Res. Lett. 66, 42-53. Espinoza-Aranda, J., and Rodriquez-Cayeros F.H., (2003). The seismic Alert System of Mexico City, In International handbook of Earthquake and Engineering Seismology, Ed. By W.H.K. Lee, H. Kanamori, P.C. Jennings and C. Kisslinger, Academic Press, pp. 1253-1260. 14

Goltz, J.D. (2002). Introducing earthquake early warning in California: A summary of social science and public policy issues, Caltech Seismological Laboratory, Disaster Assistance Division, A report to OES and the Operational Areas. Grecksch, G., and Kumpel, H. J., (1997). Statistical analysis of strong-motion accelerogram and it application to earthquake early-warning systems, Geophys. J. Int. 129, 113-123. Harben, P.E., (1991), Earthquake alert system feasibility study, Lawrence Livermore National Laboratory, Livermore, CA, UCRL-LR-109625. Heaton, T.H. (1985). A model for seismic computerized alert network, Science, 228, 987-990 Kanamori, H., (2003). Earthquake Prediction an Overview, In International handbook of Earthquake and Engineering Seismology, Ed. By W.H.K. Lee, H. Kanamori, P.C. Jennings and C. Kisslinger, Academic Press, pp. 1205-1216. Kanamori, H., Hauksson, E., and Heaton, T., (1997). Real-time seismology and earthquake hazard mitigation, Nature, 390, 461-464. Lee, W. H. K., (1995). A project implementation plan for an advanced earthquake monitoring system. Research Report of the Central Weather Bureau, Taipei, Taiwan, R.O.C., No. 448, 411 pp. Lee, W. H. K., and Espinosa-Aranda, J. M., (Conveners), (1996). Early warning and rapid response. Eleventh World Conference on Earthquake Engineering, Acapulco, Mexico, June 23-28, 1996, p 1437-1443. Lee, W. H. K., and Espinosa-Aranda, J. M., (2003). Earthquake early warning systems: Current status and perspectives, in Early Warning Systems for Natural Disaster Reduction, edited by J. Zschau and A. N. Kuppers, p. 409-423, Springer, Berlin. Midorikawa, S., (2004). Dense strong-motion array in Yokohama, Japan, and its use for disaster management. NATO Meeting, kusadasi,turkey. Nakamura, Y., (1988). On the urgent earthquake detection and alarm system (UrEDAS), Proc. of the 9th world conference on earthquake engineering, Tokyo-Kyoto, Japan. Nakamura, Y.: Earthquake alarm system for Japan Railways, Japanese Railway Engineering, 28,4, 3-7, (1989). Olson, E.L., and R.M. Allen (2005) The deterministic nature of earthquake rupture. Nature 438, 212-215, doi:10.1038/nature04214 Shin, T. C., and Teng, T. L., (2001). An overview of the 1999 Chi-Chi, Taiwan, earthquake, Bull. Seism. Soc. Am., 91, 895-913. Teng, T.L., Wu, Y. M., Shin, T.C., Tsai, Y.B., and Lee, W.H.K., (1997). One minute after: strong-motion map, effective epicenter, and effective magnitude, Bull. Seism. Soc. Am. 87, 1209-1219. Tsai, Y. B. and Y. M. Wu, (1997). Quick determination of magnitude and intensity for seismic early warning, 29th IASPEI meeting, Thessaloniki, Greece. Wenzel, F., Oncescu, M.C., Baur, M., Fiedrich, F., (1999). An early warning system for Bucharest, Seism. Res. Lett., 70, 2, 161-169. Wenzel, F., M. Baur, F. Fiedrich,C. Ionescu And M. C. Oncescu, (2001), Potential of Earthquake Early Warning Systems, Natural Hazards 23: 407 416, 2001. Wu, Y. M. Shin, T. C., and Tsai, Y. B., (1998). Quick and reliable determination of magnitude for seismic early warning, Bull. Seism. Soc. Am. 88, 1254-1259. Wu, Y. M., Chung, J. K., Shin, T. C., Hsiao, N. C., Tsai, Y. B., Lee, W. H. K., and Teng, T. L., (1999). Development of an integrated seismic early warning system in Taiwan - case for the Hualien area earthquakes, TAO 10, 719-736. Wu, Y. M., and Teng, T. L., (2002). A virtual sub-network approach to earthquake early warning. Bull. Seism. Soc. Am, 92, 2008-2018. 15