The map shows the distribution of Earthquake Early Warning Systems around the world, with a color indicating the status of the system. In purple , the operative systems, which are providing warnings to public users. In black , the systems which are currently under real-time testing. Gray color is fi nally used for those countries where feasibility studies are currently being doing 

Context in source publication

Context 1

... early warning; Earthquake ground motion; Earthquake source observation; Real-time location; Real-time magnitude Earthquake Early Warning Systems (EEWS) are real-time, seismic monitoring infrastructures that are able to provide a rapid noti fi cation of the potential damaging effects of an impending earthquake. This objective is achieved through the fast telemetry and processing of data from dense instrument arrays deployed in the source region of the event of concern (regional EEWS) or surrounding/at the target infrastructure (front-detection or site-speci fi c EEWS). A regional EEWS is based on a dense sensor network covering a portion or the entire area that is threatened by earthquakes. The relevant source parameters (event location and magnitude) are estimated from the early portion of recorded signals (initial P-waves) and are used to predict, with a quanti fi ed con fi dence, a ground-motion intensity measure at a distant site where a target structure of interest is located. Site-speci fi c (or on-site) EEWS consist of a single sensor or an array of sensors deployed in the proximity of the target structure that is to be alerted and whose measurements of amplitude and predominant period on the initial P-wave motion are used to predict the ensuing peak ground motion (mainly related to the arrival of S- and surface waves) at the same site. Front- detection EEWS is essentially a variant of the on-site approach, where a barrier-shaped, accelerometric network is deployed between the source region and the target site to be protected. The alert is issued when two or more nodes of the array record a ground acceleration amplitude larger than a default threshold value. For typical regional distances, the peak acceleration at the barrier nodes is expected to be associated with the S-wave train, so that the distance between the network and the target is set to maximize the lead time (i.e., the time available for warning before the arrival of strong ground shaking at the target sites), which is, in this case, the travel time of S-waves from the barrier to the target site. EEWS have experienced a very rapid improvement and a wide diffusion in many active seismic regions of the world in the last three decades (Fig. 1). They are operating in Japan, Taiwan, Mexico, and California. Many other systems are under development and testing in other regions of the world such as in Italy, Turkey, Romania, and China. Most of these existing EEWS essentially operate in the two different con fi gurations described above, i.e., regional and on-site, depending on the source-to- site distance and on the geometry of the considered network with respect to the source area. The “ front-detection ” EEWS such as the barrier-type, Seismic Alert System (Espinosa-Aranda et al. 2011) for Mexico City can be particularly advantageous when the only potential seismic sources are at some distance from the strategic target to be protected. The regional EEWS approach is based on the detection of the initial P-wave signal at a number of near-source stations, typically 4 to 6. Several methodologies have been proposed for the real-time estimation of the earthquake location and magnitude and are now implemented in the EW algorithms, such as ElarmS (Allen et al. 2009), Virtual Seismologist (Cua et al. 2009), and PRESTo (Satriano et al. 2010) presently running in California, Switzerland, and Southern Italy, respectively. In the framework of EU REAKT (Strategies and Tools for Real-time Earthquake Risk Reduction, FP7:ENV2011.1.3.1-1) and international collaboration projects, testing of PRESTo early warning platform is performed in Romania, Greece, Turkey, Spain, and South Korea. The real-time magnitude estimation is generally inferred from the measurement of peak displacement amplitude and/or the predominant period measured in the fi rst few seconds of the recorded P-signal, typically 3 – 4 s. Although the saturation of the P-wave parameters has been observed for M > 6.5 – 7 earthquakes, several methodologies making use of longer time windows of the P-wave and/or the S-wave to update magnitude estimates have been shown to be ef fi cient in minimizing the problem of magnitude underestimation (Colombelli et al. 2012b). The source location and magnitude estimations, which are continuously updated by adding new station data, as the P-wave front propagates through the regional EW network, are then used to predict the severity of ground shaking at sites far away from the source, by using regional-speci fi c, ground-motion prediction equations. The on-site early warning approaches are generally aimed at estimating the expected peak ground shaking, associated with S- and surface waves, directly from the recorded early P-wave signal. This is achieved through the use of empirical regressions between measurements performed on the initial P-wave signal and the fi nal peak ground motion. Wu and Kanamori (2005) fi rst showed that the maximum amplitude of a high-pass fi ltered vertical displacement, measured on the initial 3 s of the P-wave (named P d ), can be used to estimate the peak ground velocity ( PGV ) at the same site, through a power-law relationship. The main advantage is that this relationship does not require an independent estimate of the magnitude as for regional EEWS. Although initially observed for near-source records (distances 30 km), further analyses on independent datasets have con rmed that log PGV vs. log P d scaling is still valid at relatively large distances (distances < 300 km) (Zollo et al. 2010; Colombelli et al. 2012a). Most of the currently operating on-site EEWS are threshold-based, alert methodologies: the alert is issued as the measured initial P-wave peak amplitude overcomes a given threshold which is arbitrarily set according to the predicted S-wave peak ground-motion amplitude. Since small magnitude earthquakes may have very large amplitudes driven by high-frequency spikes, such a basic threshold system can produce frequent false alarms. A more robust approach is to combine the P-wave peak (which scales with distance and magnitude) and P-wave predominant period (which scales with the magnitude), into a single proxy to be used for on-site warning (Wu and Kanamori 2005). Following this idea, Zollo et al. (2010) and Colombelli et al. (2012a) have proposed a threshold-based EW method based on the real-time measurement of the period ( t c ) and peak displacement ( P d ) parameters at stations located at increasing distances from the earthquake epicenter. The measured values of early warning parameters are compared to threshold values, which are set for a given minimum magnitude and instrumental intensity. At each recording site an alert level is assigned based on a decisional table with four levels de fi ned upon threshold values of the parameters P d and t c . Given a real-time, evolutionary estimation of earthquake location from fi rst P arrivals, the method also furnishes an estimation of the extent of potential damage zone as inferred from continuously updated averages of the period parameter and from mapping of the alert levels determined at the near-source accelerometer stations. P-wave-based, regional, and on-site EW methods can be integrated in a unique alert system (as actually done, e.g., in the new version of PRESTo, e.g., PRESTo Plus, Zollo et al. 2014), which can be used in the very fi rst seconds after a moderate-to-large earthquake to determine the earthquake location and magnitude and to map the most probable damaged zone, using data from receivers located at increasing distances from the source. Methodologies for regional earthquake early warning assume a point-source model of the earthquake source and isotropic wave amplitude attenuation. These assumptions may be inadequate to describe the earthquake source of large earthquakes and wave amplitude attenuation effects, and they can introduce signi fi cant biases in the real-time estimation of earthquake location and magnitude. This issue is critically related to the EEWS performances in terms of expected lead time and of uncertainties in predicting the peak ground motion at the site of interest. Within this context, new developments have been proposed, such as the strategy of expanding the P-wave time window for the real-time signal processing, the 2D mapping of the potential damage zone, and the use of continuous GPS measurements and methodologies to estimate fault rupture extent in real time by classifying stations into near source and far source. These innovative aspects of early warning will be discussed in the present review, with a speci fi c focus on methods for rapid and reliable source characterization for early warning applications. In EEWS the strong motion is generally synthesized by a single parameter (in most cases the peak ground velocity, PGV , the peak ground acceleration, PGA , or the peak ground displacement, PGD, Fig. 2), which can be directly related to the damage that a building or an infrastructure may undergo because of the earthquake. Two possible approaches can be explored for the prediction/estimation of ground-motion parameters at a given site. A rst possibility is to relate the ground shaking to simpli ed macroscopic description of the source, yielding ground-motion prediction equations. In such a case, indicating with PGX the selected ground-motion parameter, the simplest attenuation relationship relates the logarithm of PGX with the earthquake-to-site distance R and the magnitude M ...