TECHNOLOGIES FOR RISK MONITORING AND EMERGENCY MANAGEMENT DEVELOPMENT OF TECHNOLOGIES FOR THE MONITORING AND SEISMIC RISK MANAGEMENT

Transcription

1 G. Manfredi, M. Dolce (eds), The state of Earthquake Engineering Research in Italy: the ReLUIS-DPC Project, , doi: /r101309, 2015 Doppiavoce, Napoli, Italy TECHNOLOGIES FOR RISK MONITORING AND EMERGENCY MANAGEMENT DEVELOPMENT OF TECHNOLOGIES FOR THE MONITORING AND SEISMIC RISK MANAGEMENT TASK AT MOBILE SYSTEM FOR EARTHQUAKE EARLY WARNING Aldo Zollo University of Naples, Naples, Italy, 1 INTRODUCTION The concept of Earthquake Early Warning System (EEWS) today is becoming more and more popular in the seismological and engineering communities, especially in the most active seismic regions of the world, as one of the most effective strategies for the realtime mitigation of earthquake risk. EEW means the rapid detection of an ongoing earthquake and the prompt broadcasting of a warning message in a target area, before the arrival of the destructive waves. During the last three decades, EEWS have experienced a sudden improvement and a wide diffusion in many active seismic regions worldwide. They are actually operational in Japan [Nakamura 1984, Allen and Kanamori 2003, Odaka et al. 2003, Horiuchi et al. 2005], Taiwan [Wu and Teng 2002, Wu and Zhao 2006], and Mexico [Espinosa-Aranda et al. 2009]. Many other systems are under development and testing in other regions of the world such as in California [Allen and Kanamori 2003, Allen et al. 2009a, Allen et al. 2009b, Bose et al. 2007], Turkey [Alcik et al. 2009], Romania [Bose et al. 2009], and China [Peng et al. 2011]. In Italy, the early warning system PRESTo (PRobabilistic and Evolutionary early warning SysTem) [Satriano et al. 2010] is under testing in southern Apennines, since December It is currently used to monitor the Apenninic fault system and to detect small-to-moderate size events along the fault zone where the M W 6.9, 1980 Irpinia earthquake has occurred [Zollo et al. 2009a, Zollo et al. 2009b, Iannaccone et al. 2010]. Most of existing EEWS essentially operate in two different configurations, the regional (or network-based) and the on-site (or standalone station-based), depending on the source-to-site distance and on the geometry of the considered network with respect to the source area. The regional configuration is generally adopted when the network is deployed in the source area, while the targets to be protected are far away from it. In this approach, the early portion of recorded signals is used to rapidly evaluate the source parameters (essentially, event location and magnitude) and to predict a ground-motion intensity measure (e.g., Peak Ground Velocity, PGV, and/or Peak Ground Acceleration, PGA) at distant sites, through empirical Ground Motion Prediction Equations (GMPE). As data are progressively acquired by the network, the initial estimations are updated, providing a continuously refined information on the earthquake parameters and providing the ground

2 354 A. Zollo shaking prediction at the target sites. Given the source-to-site distance, the "lead-time" (i.e., the time between the alert issue and the arrival of damaging waves at the target site) can be relatively long in a regional configuration, although the prediction of the shaking at distant sites may be affected by large uncertainties due to the use of empirical predictive relationships and errors in location and magnitude estimates. The on-site approach, instead, is generally used when the sites to be protected are close to the source area or the source area is not accessible (e.g. off-shore). In this configuration the early portion of recorded P-wave signals is used to predict the ensuing peak groundmotion at the same site and to provide a local alert level, based on the combination of early warning observed quantities (such as P-wave peak displacement and/or predominant period). The main advantage of such an approach is that the alert for an impending earthquake at the target site is issued based on a local measurement of P-wave ground motion amplitude, avoiding the use of empirical predictive laws and bypassing the estimation of earthquake location and magnitude, which might be affected by large uncertainties in a real-time analysis. The new idea for EEWS developed in this RELUIS II project is the integration both from the technological and methodological points of views, of the two approaches, which allows to get accurate estimations of earthquake parameters, reliable prediction of the expected ground motion and quite large lead times. The integrated approach, proposed by Zollo et al. 2010, is essentially based on three key-elements: i) the definition of a local alert level from the combination of the initial Peak Displacement (P d ) and the average dominant period (τ c ); ii) the use of the initial peak displacement as a proxy for the Peak Ground Velocity; iii) the real-time mapping of a Potential Damage Zone (PDZ). The integrated approach has been off-line tested for the 2009, M W 6.3 L Aquila (Central Italy) earthquake and ten Japanese large earthquakes [Colombelli et al. 2012]. Recently, the method has been also implemented in the PRESTo software platform, and is currently under testing in Southern Italy using data streaming of small-to-moderate events from the Irpinia Seismic Network (ISNet). In the RELUIS II project, one additional objective was to design and develop a mobile early warning system to install and use during seismic crisis occurring before (foreshocks) and after a main event (aftershocks) to complement an existing permanent alert system. Such a mobile system has the aim to support the Department of Civil Protection for the real-time monitoring of ongoing seismic activity and provide with early warnings to be used for emergency management. We aimed at developing the technologies for a single node of the network (EW box) which included the PRESTo software platform implementation, focusing on the methodologies and algorithms for the real-time notification of the seismic alert and on-site threshold based alert strategy. In addition to the aforementioned objectives, a new line of research, structural damage monitoring from the analysis of seismic noise and seismic interferometry techniques, was developed in collaboration with researchers of the Polytechnic of Turin. The goal was to use ambient noise and shot records from an hammer to retrieve the impulse response of a bridge from cross-correlation analysis. During the acquisition, the bridge was damaged to mimic the interaction between the basement of the bridge and the water flow in a river and the damage quantified as a percentage variation in average wave velocity in the structure. This method can complement an early warning system, because it monitors the structural behavior of a structure without invasive methods and may indicate when it approaches to severe damage.

3 Mobile System for Earthquake Early Warning BACKGROUND AND MOTIVATION In the framework of RELUIS II project the threshold-based earthquake early warning methodology [Zollo et al. 2010] has been verified and applied to a set of Japanese earthquake (M>6) occurred during the last decade. This approach represents an integration of regional/on-site early warning method and can be used in the very first seconds after a moderate-to-large earthquake to map the most Probable Damaged Zones (PDZ). The key element of the method is the real-time simultaneous measurement of initial peak displacement (P d ) and period parameter (τ c ) in a 3-second window after the first P-arrival time at accelerometer stations located at increasing distances from the epicenter. As for the on-site approach the recorded values of P d and τ c are compared to threshold values, which are set for a minimum magnitude M = 6 and instrumental intensity I MM VII, according to empirical regression analysis of strong motion data from different seismic regions. At each recording site, an alert level is assigned on the basis of a decisional table with four entries defined by critical values of the parameters P d and τ c. The alert levels refer to the possible occurrence of a damaging earthquake nearby or far-away the recording site. A regional network of accelerometers provides the event location and transmits the information about the alert levels recorded at near-source stations to more distant sites, before the arrival of the most destructive S phase. A set of off-line performance tests of this method has been carried out by using ten, M > 6 Japanese earthquakes showing that the methodology is very robust for mapping the PDZ in the first seconds after a moderate-to-large earthquake (Fig.1). The results displayed a very good matching between the rapidly predicted and observed damage zone, the latter being mapped a few days after the event from detailed macro-seismic surveys. A second aspect of earthquake early warning considered in the framework of the project concerned the development of a Mobile Early Warning Network (EW MobNet). The EW MobNet represents an innovative system able to integrate both regional and on-site approaches (represented in Fig. 2) based on the real-time measurement of peak displacement and predominant period on early P-wave signals. Using the threshold-based method [Zollo et al. 2010] the system can provide a rapid estimation of the PDZ, by-passing the magnitude estimation for ground motion prediction. Moreover it can be easily integrated with recently developed regional methods, e.g. PRESTo [Satriano et al. 2010]. The most critical issue for an EW MobNet is the choice of the data transmission system. It is well known that satellite transmission may be slow to be effective for early warning (due to several seconds of time-delays between transmitters and receivers), expensive and difficult to be implemented. On the other hand, radio links can be not fully suitable for streaming due to actual band limitations. Proprietary WI-FI and/or GPRS/UMTS are our preferred solutions which enable to reduce the time transmission and to increase the lead-time (i.e., the time between the alert notification and the arrival time of potentially destructive waves at a given target site), available for risk mitigation actions. The choice of the optimal sensor and Analogto-Digital Converter (ADC) is also critical for the realization of an EW MobNet, in relation to the target of detecting and issuing an alert for moderate to strong intensity events (Instrumental intensity larger than V-VI). Our research group is developing a prototype node for an EW MobNet based on Micro-Electro-Mechanical Systems (MEMS) sensor technology. Three different axial accelerometers are being tested in combination with several ADC (12, 14, 16, 18, 24 bit) to define the optimum hardware configuration in terms of cost, energy consumption, portability and dynamic range.

4 356 A. Zollo Figure 1. Example of output maps of a threshold-based EW system. The figure refers to the 2000, M = 7.3 Tottori earthquake and shows the results obtained about 12 seconds after the origin time. The real-time intensity map (top): the epicenter is identified by a black star; stations for which 3 seconds of data are available are represented by a black triangle. The color scale represents the predicted intensity value, while the real intensity value is identified by the red number close to each station. The operative alert level map (bottom): white triangles represent stations triggered by the earthquake, while red and blue triangles suggest the alert level value at each station, 3 and 1 respectively. The red line delimits the true PDZ, while the maximum PDZ is identified by a black dashed line. Figure 2. Example of Regional and On-Site Early Warning System. The scenario with the target outside the epicentral zone (left): during earthquake each node (red point) acquire and analyse data in real-time and send a coarse alert signal to the target. Simultaneously they send traces to the control room of the regional early warning system. The control room send a second alarm to the target with more information. The scenario with the target inside the epicentral zone (right): during earthquake each node (red point) acquire and analyse data in real-time and send an alert signal to the target. Simultaneously they send traces to the control room of the regional early warning system.

5 Mobile System for Earthquake Early Warning 357 During the project, the structural damage of a structure was monitored using interferometric methods. Several authors have theoretically and experimentally demonstrated that cross-correlation of a random isotropic wave-field computed between two recording points A and B results in a waveform that differs only by an amplitude factor from the Green function G AB (t) between the receivers [Sanchez-Sesma & Campillo, 2006; Shapiro et al., 2005]. In the last decade this relation has been widely used in seismology to obtain information about the Earth s crust, and generally about the wave propagation media, from the seismic ambient noise. Methods based on similar approaches are recently applied also for structural monitoring and for studies on structural material changes [Larose et al., 2006; Stähler et al. 2011]. In order to test these methods for seismic engineering applications, we used them to evaluate the variation of elastic properties of a 1:2 scaled model of a masonry arc bridge undergone to an increasing level of controlled damage. The bridge was constructed in the laboratory of the Department of Structural and Geotechnical engineering of the Polytechnic of Turin (Italy) to study the evolution of damage mechanisms related to the application of foundation movements. Different excitation sources were applied to the bridge model: ambient vibrations, impact hammer and a shaker; then acceleration signals where acquired by 18 mono-axial accelerometers, distributed in different points of the bridge and recording the signal at a sampling rate of 400 Hz. For our analysis we used 3 dataset composed by 3 minutes of signals acquired with the bridge excited by ambient noise vibration at three different level of damage. Following the process proposed to obtain reliable Green Functions from seismic ambient noise measurements [Bensen et al., 2007], we organized the single station data in 18 window of 10 s length that have been equalized in both time and frequency domains by whitening and one-bit normalizations. The records from couple of stations have been cross-correlated to build up a database of Green s Functions. Finally, in order to relate the variations of Green s Functions to the damage level of the bridge, the functions at the same receivers couple for different damage level are cross-correlated and a statistical analysis was performed computing the mean of correlation and the standard deviation. The final correlation values are significantly lower when they are computed between receivers couples in undamaged-damaged cases respect the values of undamaged undamaged or damageddamaged cases. Figure 3. (a) the scaled madison bridge. Mean cross-correlations and standard deviation for each receivers couple. The correlations values and related standard deviations are computed by Green s Functions between undamaged-undamaged (b) and undamaged-damaged cases (c). 3 RESEARCH STRUCTURE The work in the ReLUIS II projectwas organized in activities and sub-activities as follows: 1) PRESTo software platform

6 358 A. Zollo Methods of signal analysis and algorithms for the real-time notification of the seismic alert PRESTo PLUS software platform implementation with on-site threshold based alert strategy for Early Warning Box (EW-Box) 2) Early Warning Box as a node of EW MobNet Technical specifications of the system Prototype installation at a ISNET seismic station Development of a P-wave based threshold EW algorithm running on a single station 3) Structural damage from the analysis of seismic noise and seismic interferometry techniques Cross-correlation analysis aimed at detecting the structural damage of bridges Temporal variations of the elastic properties of a structure 4 MAIN RESULTS 4.1 Methods of signal analysis and algorithms for the real-time notification of the seismic alert. On-site Threshold based Alert Strategy: methodological development and testing In the line with the threshold based approach best suited for dense regional and national seismic networks, we developed a P-wave threshold-based alert strategy for a stand-alone seismic station (onsite approach) to be finally implemented in the EW node of the EW Mobnet. The methodology is based on two observed correlations between EW parameters and ground motion quantities. The first one is the empirical correlation between the Peak Displacement (Pd) (measured in the first few seconds of P-wave signal) and the Peak Ground Velocity (PGV) (measured on the entire record) [Zollo et al. 2010]. The second one is the correlation between the PGV and the Instrumental Intensity (Mercalli-Modified Intensity, MMI) at each site. Given this empirical scaling, and assuming a threshold Intensity value, we defined a threshold value for the observed PGV. This value can be converted into a corresponding threshold value for the Early Warning parameter, Pd. The strategy for the Onsite warning consists in the continuous measurement of the Peak Displacement, starting from the P-wave trigger time, and progressively expanding the observation time window, following the approach of Colombelli et al., 2012a. As soon as Pd overcomes the established threshold, an alert level is issued at the considered station. The threshold-based alert strategy has been tested off-line using a database of moderate-tostrong Japanese earthquakes (4<M<9), including the last destructive Magnitude , Tohoku-Oki earthquake. We performed a massive statistical analysis using a total number of about accelerometer waveforms, corresponding to 68 earthquakes, in the distance range between 0 and 500 km. Following the Alert table as defined by Colombelli et al., 2012b, for each available record, we compared the observed PGV value with the one predicted from the Pd vs. PGV equation. Based on the combination of the observed Pd and PGV at each site (Fig. 4a), we defined successful, missed and false alarms (see Table 1) and counted their total number.

7 Mobile System for Earthquake Early Warning 359 Table 1. Criteria adopted to distinguish among successful, missed and false alarms in testing the performance of the PRESTo system in the regional configuration. Successful alarm Missed alarm False alarm Pd Pd* and PGV PGV* Pd > Pd* and PGV > PGV* Pd Pd* and PGV > PGV* Pd > Pd* and PGV PGV* The results of the cumulative statistics is shown, in the form of histograms, in Figure 4b, for the Intensity threshold value MMI = 6 and in Figure 4c, for the Intensity threshold value MMI = 7. In both cases we found an excellent percentage of successful alarms (> 90%), a small percentage of false alarms (< 6%) and a percentage of missed alarms smaller than 1%. Figure 4. Panel a) Threshold-based scheme: definition of successful, missed and false alarms based on Pd and PGV correlation. Cumulative statistics on the entire database for an Intensity threshold of 6 (panel b) and 7 (panel c). 4.2 Defining the architecture and components of the system with the technical specifications of the hardware A new prototype of low-cost sensor named ASTERISK was produced to represent the main component of the EW node of MobNet. The prototype is based on MEMS technology as regards the accelerometric sensors, and a processor ATMEL 400Mhz for the part of the processing and management of information (Fig. 5). The performances of the new sensor have been defined by the measurement of its electronic noise level (Fig. 6), as well as by the tilt test. The sensitivity of the MEMS sensor resulted equal to 0.061x10-3 g (6.1x10-3 gal), while the level of electronic noise was 13.8x10-3 g (1.38 gal).

8 360 A. Zollo Furthermore, keeping into consideration the nominal and effective sensor performances, that is to say including the sensor electronic noise except for the environmental seismic noise, the threshold levels for the recordable ground motion in terms of P- and S- waves have been defined considering different magnitudes and distances (Fig. 7). For example, Figure 7 shows that in principle ASTERISK should be able to trigger on the P-waves a magnitude 2 event at the epicentral distance of 10 km. Finally, in order to perform a field assessment of the sensor performance, a prototype has been recently installed at the ISNET seismic station named Rocca San Felice (AV), in proximity to high quality sensors (Fig. 8). Figure 5. ASTERISK hardware components and main characteristics. Figure 6. PSD spectra of the ASTERISK s electronic noise (left) and its histogram (right).

9 Mobile System for Earthquake Early Warning 361 Figure 7. Nominal (red) and effective (green) ASTERISK s sensitivity with respect to the theoretical ground motion for different magnitude and distance ranges (black). Figure 8. Example of ASTERISK installation at the ISNET station of Rocca San Felice (AV). 4.3 Structural damage from the analysis of seismic noise and seismic interferometry techniques This activity is carried out in collaboration with the Department of Structural and Geotechnical Engineering of Polytechnic of Turin. Several authors have shown that the cross-correlation of a wave field isotropically diffused inside a medium and recorded at two receivers approximates the Green function between these two points [Sanchez-Sesma & -Campillo, 2006; Shapiro et al., 2005]. This relationship allows to obtain information about the velocity model and its temporal variation from the analysis of the ambient seismic noise and the earthquakes codas. Recently, similar methods have been also applied for the monitoring of the structures [Larose et al., 2006; Stähler et al. 2011]. In this activity, we tested these methods to estimate the variation of the elastic properties of a masonry arch bridge, built in laboratory at a scale 1:2, subjected to an increasing level of controlled damage. The bridge was excited with different types of sources: ambient

10 362 A. Zollo vibrations, hammer hits and shaking table. The signals were acquired by 18 mono-axial accelerometers, distributed in different points along the bridge and with a sampling frequency of 400 Hz. For this analysis we used three datasets of 3 minute-long time records of ambient noise for the three levels of damage. Analogously, we also analyzed the records in correspondence of the hammer hits. We arranged the noise records at each station in 18 time windows of 10s, equalized both in the time domain and the frequency domain through spectral whitening and one-bit normalization. Similarly we extracted hammer related events and selected their codas as the portion of the signal that does not contain the ballistic arrivals and which exponentially decays in time. The noise and the codas were separately crosscorrelated for couples of stations to build a database of Green's functions. Finally, these were compared with each other for the three levels of damage. In the case of the ambient noise main differences were observed in the Green's functions during the three phases. This difference was quantified by the cross-correlation of the Green's functions. Anyhow, at this level we were not able to distinguish the effect of the variability of the propagation medium from the variability of noise sources, which may be different in the three phases of the experiment. Analyzing the codas of the signals, instead, we found that in a broad range of frequencies (30-70 Hz), the Green's functions for the damaged bridge were delayed as compared to the Green s functions of the intact bridge (Fig. 9). On a single section, with a selection of the cross-correlations with a signal to noise ratio above 5, we measured the variation in the arrival of the phases ΔT and the ratio ΔT/T related to the relative variation of speed ΔV/V. This was estimated to be of the order of 3%, uniformly in the analyzed frequency band. Assuming a S wave speed of 3km/s, the speed change is of the order of 100 m/s. Figure 9. Cross correlated of event codas for the un-damaged and the heavily damaged bridge. 5 DISCUSSION The research activities were fully compliant in the timing and objectives of the program originally proposed. Our research unit closely collaborated on Early Warning developments with the Department of Structural Engineering (University of Naples Federico II). We have jointly participated to the project Early Warning Sismico: sviluppo e test di un sistema prototipo per l allerta sismica preventiva in un edificio pubblico in the frame of the F.A.R.O. Programme (funding programme managed by the Polo delle Scienze e delle Tecnologie dell Università di Napoli Federico II ).

11 Mobile System for Earthquake Early Warning 363 As a follow-up of the scientific interaction with research groups working at the same research line, we established a new collaboration with the researchers from the Department of Structural and Geotechnical Engineering of Polytechnic of Turin on the topic of structural damage monitoring from the analysis of seismic noise and seismic interferometry techniques. The results obtained in the project by our research unit allowed to make a step forward in the design and implementation of reliable early warning systems for engineering and societal applications. The regional, network-based method and on-site warning have been fully integrated and made operational in the software platform PRESTo, which is now delivered freely under GUI license through the site The software is actually running in testing mode on the seismic networks of South Korea, Romania and INOGS in Italy. A prototype of the Early Warning Box has been designed and realized in laboratory. The technical characteristics of the node HW components (data logger, sensor, data transmission, local storage and computing capabilities) have been specifically designed to be integrated in a EW mobile network for aftershocks, although no network testing has been performed due to limited financial resources. A new prototype of low-cost sensor named ASTERISK was produced based on MEMS technology as regards the accelerometric sensors, and a processor ATMEL for the part of the processing and management of information. The performance of the sensor has been tested by comparing the noise and signal records at the ISNET seismic station named Rocca San Felice (AV), in proximity to high quality sensors (Fig. 8). We found that the use of ambient noise for isolated experiments does not allow to distinguish the effect of the variability of the propagation medium from the variability of noise sources, which might have been different in the three phases of the experiment. Analyzing the codas of the signals, instead, we found that in a broad range of frequencies (30-70 Hz), the Green's functions for the damaged bridge were delayed as compared to the Green s functions of the intact bridge (Fig.9). The relative time delay is a proxy for the relative variation of the wave speed ΔV/V. This was estimated to be of the order of 3%, with a speed change of the order of 100 m/s, one order of magnitude larger than the value reported for fault preparing to large earthquakes. 6 VISIONS AND DEVELOPMENTS The RELUIS II project has supported and co-financed the research activity of our group during these years aimed at the development and implementation of advanced methodologies for early warning. The new generation of EWS that our group has developed integrates the concepts of regional and on-site approaches providing an evolutionary and probabilistic estimation of the source parameters (location and magnitude) but also a reliable estimation of the ground shaking amplitude at the site to protect, trying to maximize the lead-time and minimize the uncertainties on the estimated parameters. The project RELUIS II allowed to develop the new version (PRESTo PLUS ) of the previously existing EW platform which now can operate both as a regional or onsite, threshold-based system, providing an alert level at each recording site, in addition to the estimation of earthquake location and magnitude, and the predicted peak ground motion through the GMPE. Furthermore, the measure of P-wave peak amplitude and characteristic period allows to map in real-time the Potential Damage Zone, e.g. the near source region where ground shaking is expected to produce significant damage during a destructive earthquake.

12 364 A. Zollo The software platform PRESTo PLUS is now distributed and a number of researchers managing local and regional seismic networks are testing the performances in other seismic area of the world. An experimentation is ongoing also in Italy, in collaboration with INOGS of Trieste. One of the main products of our research is the design of an early warning box whose prototype has been tested in laboratory and in the field. The hardware components of the node are specifically designed to operate as a stand-alone instruments providing a local alert or being integrated in a larger network by sharing the acquired information with other nodes of the network. The node has the software capability to run the algorithms that have been developed for the platform PRESTo and adapted to run in stand-alone mode. After the first exploratory phase and the realization of the prototype, done in this project, we are currently investigating the feasibility of a joint venture with a specialized company to produce a commercial version of the EW node. Interferometric techniques are passive, non-invasive methods that allow to retrieve changes in the structure without modifying it. However, to constrain the velocity change in the structure a complete characterization of the noise in repeated experiments is required. Specifically, to avoid deconvolution for noise sources, ambient noise should be stationary. To deeply explore the changes, moreover, higher frequency records are required. Due to the relatively large speeds in the medium, anomalies may be detected if khz records are available. The same results can be obtained by the use of shots (realized by hammers); again, shots require the same source time function to avoid source deconvolution. 7 MAIN REFERENCES Alcik H, Ozel O, Apaydin N, Erdik M (2009). A study on warning algorithms for Istanbul earthquake early warning system, Geophys Res Lett 36 L00B05 doi: /2008gl Allen RM, Kanamori H (2003). The potential for earthquake early warning in Southern California. Science 300: Allen RM, Brown H, Hellweg M, Khainovski O, Lombard P, Neuhauser D (2009a). Realtime earthquake detection and hazard assessment by ElarmS across California. Geophys Res Lett 36 L00B08 doi: /2008gl Allen RM, Gasparini P, Kamigaichi O, Böse M (2009b). The status of earthquake early warning around the world: an introductory overview. Seism Res Lett 80: Bensen G. D., Ritzwoller M. H., Barmin M. P., Levshin A. L., Lin F., Moschetti M. P., Shapiro N. M. and Yang Y. (2007). Processing seismic ambient noise data to obtain reliable broad-band surface wave dispersion measurements, Geophys. J. Int., 169, Böse M, Ionescu C, Wenzel F (2007) Earthquake early warning for Bucharest, Romania: novel and revised scaling relations, Geophys Res Lett 34 doi: /2007gl Böse M, Hauksson E, Solanki K, Kanamori H, Heaton TH (2009). Real-time testing of the on-site warning algorithm in Southern California and its performance during the July 29, 2008 Mw 5.4 Chino Hills earthquake. Geophys Res Lett 36 doi: /2008gl Colombelli S, Amoroso O, Zollo A, Kanamori H (2012). Test of a threshold-based earthquake early-warning method using Japanese data. Bull Seism Soc Am 102 doi: / Espinosa-Aranda JM, Cuellar A, Garcia A, Ibarrola G, Islas R, Maldonado S, Rodriguez FH (2009). Evolution of the Mexican Seismic Alert System (SASMEX). Seism Res Lett 80:

13 Mobile System for Earthquake Early Warning 365 Horiuchi S, Negishi N, Abe K, Kamimura K, Fujinawa Y (2005). An automatic processing system for broadcasting system earthquake alarms. Bull Seism Soc Am 95: Iannaccone G, Zollo A, Elia L, Convertito V, Satriano C, Festa G, Martino C, Lancieri M, Bobbio A, Stabile TA, Vassallo M, Emolo A (2010). A prototype system for earthquake early-warning and alert management in Southern Italy. Bull Earthquake Eng 8: doi: /s Larose E., de Rosny J., Margerin L., Anache D., Gouedard P., Campillo M., van Tiggelen B: (2006). Observation of multiple scattering of khz vibrations in a concrete structure and application to monitoring weak changes, Phys. Rev. Lett., 73 (1), Nakamura Y (1984). Development of earthquake early-warning system for the Shinkansen, some recent earthquake engineering research and practical in Japan. The Japanese National Committee of the International Association for Earthquake Engineering Nakamura Y (1988). On the urgent earthquake detection and alarm system (UrEDAS) In Proceedings 9th World Conf. Earthquake Eng. 7: Odaka T, Ashiya K, Tsukada S, Sato S, Ohtake K, Nozaka D (2003). A new method of quickly estimating epicentral distance and magnitude from a single seismic record. Bull Seism Soc Am 93: Peng HS, Wu ZL, Wu YM, Yu SM, Zhang DN Huang WH (2011). Developing a prototype earthquake early warning system in the Beijing Capital Region. Seism Res Lett 82: Sánchez-Sesma F. J. and Campillo M. (2006). Retrieval of the Green s Function from Cross Correlation: The Canonical Elastic Problem, Bull Seism Soc Am 96: , doi: / Satriano C, Elia L, Martino C, Lancieri M, Zollo A, Iannaccone G (2010). PRESTo, the earthquake early warning system for southern Italy: concepts, capabilities and future perspectives. Soil Dyn Earthq Eng doi: /j.soildyn Shapiro, N.M, Campillo M., Stehly L., and Ritzwoller M.H. (2005). High resolution surface wave tomography from ambient seismic noise, Science. 307, Stähler S. C., Sens-Schönfelder C., & Niederleithinger E. (2011). Monitoring stress changes in a concrete bridge with coda wave interferometry. The Journal of the Acoustical Society of America, 129, 4, Wu YM, Teng LT (2002) A virtual sub-network approach to earthquake early warning. Bull Seism Soc Am 92: Wu YM, Zhao L (2006) Magnitude estimation using the first three seconds P-wave amplitude in earthquake early warning. Geophys Res Lett 33 L16312 doi: /2006gl Zollo A, Amoroso O, Lancieri M, Wu YM, Kanamori H (2010). A threshold-based earthquake early warning using dense accelerometer networks. Geophys J Int 183: Zollo A, Iannaccone G, Convertito V, Elia L, Iervolino I, Lancieri M, Lomax A, Martino C, Satriano C, Weber E, Gasparini P (2009a). The earthquake early warning system in southern Italy. In Encyclopedia of Complexity and System Science 5: doi: / Zollo A, Iannaccone G, Lancieri M, Cantore L, Convertito V, Emolo A, Festa G, Gallovič F, Vassallo M, Martino C, Satriano C, Gasparini P (2009b). Earthquake early warning system in southern Italy: methodologies and performance evaluation. Geophys Res Lett 36 L00B07 doi: /2008gl03668

14 366 A. Zollo 8 RELUIS REFERENCES Colombelli S., Amoroso O., Zollo A. and Kanamori H. (2012). Test of a Threshold-Based Earthquake Early-Warning Method Using Japanese Data Bulletin of the Seismological Society of America, Vol. 102, No. 3, pp , doi: / Colombelli S., Zollo A., Festa G. Kanamori H. (2012). Early magnitude and potential damage zone estimates for the great Mw 9 Tohoku-Oki earthquake, Geophysical Research Letters, Vol. 39, L22306,doi: /2012GL Zollo A., Festa G., Emolo A. and Colombelli S., Source Characterization for Earthquake Early Warning (2014). Encyclopedia of Earthquake Engineering, Springer, Eds C. Galasso and F. Jalayer, in press A. Zollo, S. Colombelli, L. Elia, A. Emolo, G. Festa, G. Iannaccone, C. Martino, and P. Gasparini (2013). An integrated regional and on-site Earthquake Early Warning System for Southern Italy: concepts, methodologies and performances. Early Warning for Geological Disasters: Scientific Methods and Current Practice, Springer-Verlag Berlin Heidelberg, Eds F.Wenzel and J. Zschau