ADGEOAdvances in GeosciencesADGEOAdv. Geosci.1680-7359Copernicus GmbHGöttingen, Germany10.5194/adgeo-40-51-2015Trans-national earthquake early warning (EEW) in north-eastern Italy, Slovenia and Austria: first experience with PRESTo at the CE3RN networkPicozziM.matteo.picozzi@unina.itEliaL.PesaresiD.https://orcid.org/0000-0002-4411-7281ZolloA.https://orcid.org/0000-0002-8191-9566MucciarelliM.https://orcid.org/0000-0003-2398-9168GosarA.https://orcid.org/0000-0003-1511-1021LenhardtW.https://orcid.org/0000-0001-9031-3753ŽivčićM.RISSC, Università “Federico II” di Napoli – AMRA, Naples, ItalyCRS, OGS (Istituto Nazionale di Oceanografia e di Geofisica
Sperimentale), Trieste, ItalyARSO – Agencija Republike Slovenije za Okolje, Ljubljana, SloveniaZAMG – Zentralanstalt für Meteorologie und Geodynamik, Vienna, AustriaM. Picozzi (matteo.picozzi@unina.it)12May20154040516122December201423March20152April2015This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://adgeo.copernicus.org/articles/40/51/2015/adgeo-40-51-2015.htmlThe full text article is available as a PDF file from https://adgeo.copernicus.org/articles/40/51/2015/adgeo-40-51-2015.pdf
The region of central and eastern Europe
is an area characterised by a relatively high seismic risk. Since 2001, to
monitor the seismicity of this area, the OGS (Istituto Nazionale di
Oceanografia e di Geofisica Sperimentale) in Italy, the Agencija Republike
Slovenije za Okolje (ARSO) in Slovenia, the Zentralanstalt für
Meteorologie und Geodynamik (ZAMG) in Austria, and the Università di
Trieste (UniTS) have cooperated in real-time seismological data exchange. In
2014 OGS, ARSO, ZAMG and UniTS created a cooperative network named the
Central and Eastern European Earthquake Research Network (CE3RN), and
teamed up with the University of Naples Federico II, Italy, to implement an
earthquake early warning system based on the existing networks. Since
May 2014, the earthquake early warning system (EEWS) given by the integration
of the PRESTo (PRobability and Evolutionary early warning SysTem) alert
management platform and the CE3RN accelerometric stations has been under
real-time testing in order to assess the system's performance. This work
presents a preliminary analysis of the EEWS performance carried out by
playing back real strong motion recordings for the 1976 Friuli earthquake
(MW= 6.5). Then, the results of the first 6 months of
real-time testing of the EEWS are presented and discussed.
Introduction
With the aim of monitoring the seismic activity in the eastern sector of the
Alps, since 2001 OGS (Istituto Nazionale di Oceanografia e di Geofisica
Sperimentale) in Udine (Italy), the Agencija Republike Slovenije za Okolje
(ARSO) in Ljubljana (Slovenia), the Zentralanstalt für Meteorologie und
Geodynamik (ZAMG) in Vienna (Austria), and the University of Trieste (UniTS)
have been collecting, analysing,
archiving and exchanging seismic data in real time. The data exchange has
proven to be effective and very useful in the case of seismic events at the
borders between Italy, Austria and Slovenia, where the poor coverage of
individual national seismic networks precluded a precise earthquake location.
The usage of common data from the integrated networks improves significantly
the overall capability of real-time event detection and rapid
characterisation in this area. Furthermore, in 2014, OGS, ARSO, ZAMG and
UniTS signed a memorandum of understanding naming the cooperative network as
the Central and Eastern European Earthquake Research Network (CE3RN)
(Bragato et al., 2014).
Recently, in order to extend the seismic monitoring in north-eastern Italy,
Slovenia and southern Austria towards earthquake early warning applications,
OGS, ARSO and ZAMG teamed up with the RISSC-Lab group
(http://www.rissclab.unina.it) of the Department of Physics at the
University of Naples Federico II in Italy.
An earthquake early warning system (EEWS) is a real-time system integrating
seismic networks and software capable of performing real-time data telemetry
and analysis in order to issue alert messages within seconds from the origin
of an earthquake and before the destructive S-waves generated by the event
reach the users. When accompanied by appropriate training and preparedness of
the population, an EEWS is an effective and viable tool for reducing the
exposure of a population to seismic risk (e.g. Allen et al., 2009; Hoshiba,
2013; Picozzi et al., 2015a). The application of EEWS is nowadays increasing
and several countries around the world have already developed EEWS, or are on
the verge of doing so. Japan, Taiwan, Mexico, Romania and California, for
example, already have operational EEWSs (e.g. Horiuchi et al., 2005; Wu and
Zhao, 2006; Espinosa-Aranda et al., 2009; Böse et al., 2007, 2009). EEWSs
are also under development and testing in other regions of the world, such as
Italy, Turkey, Spain, and China (Satriano et al., 2010; Zollo et al., 2014;
Alcik et al., 2009; Peng et al., 2011; Picozzi et al., 2014, 2015b).
The collaboration among OGS, ARSO, ZAMG and RISSC-LAB focuses on testing the
EEW platform PRESTo (probabilistic and evolutionary early warning system:
http://www.prestoews.org) in north-eastern Italy, Slovenia and Austria
at the network CE3RN, and represents, to our knowledge, the first
worldwide attempt of implementing a trans-national EEWs. PRESTo is a
stand-alone software system that processes live accelerometric streams from a
seismic network to promptly provide probabilistic and evolutionary estimates
of location and magnitude of detected earthquakes while they are occurring,
as well as shaking prediction at the regional scale (Satriano et al., 2011).
Since 2014 PRESTo has run on OGS, ARSO and ZAMG data, by collecting and
analysing in real time the data streams from 20 stations (Fig. 1).
In the following, first, we briefly present the CE3RN project, and we
summarise the characteristics of EEWS and PRESTo. Then, we present the
results of a test carried out by playing back the waveforms of the strong
motion data of the MW= 6.5, 1976 Friuli earthquake, and,
finally, we report on the performance of the EEW system during this
preliminary testing phase.
CE3RN institutions involved in the EEW experiment (blue
squares), real-time accelerometric stations (yellow triangles).
The CE3RN project
The region of central and eastern Europe is an area characterised by a
relatively high seismicity. The active seismogenic structures and the related
potentially destructive events are located in the proximity of the political
boundaries between several countries existing in the area. An example is the
seismic region between north-eastern Italy (Friuli-Venezia Giulia,
Trentino-Alto Adige and Veneto), Austria (Tyrol, Carinthia) and Slovenia. So,
when a destructive earthquake occurs in the area, all three countries are
possibly affected. In the year 2001, the institutes OGS, ARSO, ZAMG, and
UniTS signed an agreement for real-time seismological data exchange in the
south-eastern Alps region. Soon after, the Interreg IIIa Italia-Austria
Trans-National Seismological Networks in the South-Eastern Alps and FASTLINK
projects started. The main goal of these projects was the creation of a
transfrontier network for the common seismic monitoring of the region for
scientific and civil defense purposes.
The OGS, ZAMG and ARSO seismic networks present many similarities. While
there is a variety of sensor typologies in use (i.e. from strong motion to
(very) broadband), all the stations are equipped with Quanterra data loggers
(Q6180, Q4120, Q730 and Q330), and similar strong motion
sensors are used almost at each seismic station of the single networks. As
shown by Stein and Reimiller (2014), the stations equipped with data logger
Q330 are capable of delivering data with a latency of less than 1 s, and
therefore are suitable for early warning applications. The use of similar
instrumentation facilitated a very important consequence of Interreg project
Trans-National Seismological Networks in the South-Eastern Alps, and the
adoption of common software suite Antelope from Boulder Real-Time
Technologies (BRTT), for seismic data real-time acquisition, archiving,
analysis and exchange. It is in fact straightforward, given that all the
involved institutions use the same data acquisition software, to extend the
single networks' seismic monitoring capabilities to the entire transfrontier
network, thus acting like an extended virtual network. All the involved
partners exchange waveforms and parametric data in real time through a
network of bi-directional data links, mainly via the Internet,
interconnecting all data centres.
Schematic representation of the regional approach for EEW (modified
from Satriano et al., 2011), and overview of the analyses carried out by the
PRESTo software system for the real-time event characterisation and ground
motion level at target site prediction.
During the recent past years, the high-quality data recorded by the
trans-national network have been used by the involved institutions for their
scientific research, for institutional activities and for civil defence
services. Several common international projects have been realised with
success. The instrumentation has been continuously upgraded and the
installations quality improved, as well as the data transmission efficiency.
In 2014, OGS, ARSO, ZAMG and UniTS signed a memorandum of understanding named
the Central and Eastern European Earthquake Research Network (CE3RN)
cooperative network (Bragato et al., 2014). CE3RN represents an
excellent example of international high-quality research infrastructure and
the starting point for the enlargement of the transfrontier network to all
countries and their seismological institutions of the central and eastern
Europe region. Furthermore, one of the main goals of the CE3RN is to
intensify the cooperation between these institutions through common research
activities and preparation of common international projects.
On 11 November 2014, the CE3RN partnership was enlarged to also include
the Croatian Seismological Survey (CSS) of the University of Zagreb in
Croatia.
Earthquake early warning systems and PRESTo
EEWS typically follows two basic approaches: “regional” (or network based),
and “on-site” (or a single station). Regional EEWS are based on the use of
a seismic network located near one or more expected epicentral areas, whose
aim is to detect and locate an earthquake, and to determine its magnitude
from the analysis of the first few seconds of the arriving P-waves at
multiple stations close to the epicentre (Satriano et al., 2011). On the
other hand, on-site EEWS are based on seismic sensors deployed directly at
the target site and exploit only the information carried by the faster early
P-waves to infer the larger shaking related to the incoming S and surface
waves.
One key parameter for an EEWS is the lead time, i.e. the time available to
perform safety measures at distant targets once an earthquake has been
promptly detected and characterised, and an alarm has been issued. The lead
time for regional EEWS is defined as the travel-time difference between the
arrival of the first S-waves at the target site and the P-waves recorded in
the source area, after accounting for the necessary computation and data
transmission times. In on-site EEWS, the lead time is equal to the difference
in S- and P-wave arrival times at the target itself.
Recently, Zollo et al. (2010) showed that the two approaches can be
profitably integrated within a unique system that allows the early estimation
of the potential damage zone (PDZ) associated with an event. Clearly, the
integration of regional and on-site approaches is particularly useful
whenever target sites are threatened by more than one seismic source area,
and the latter are located at variable distances from the target sites. An
exhaustive review of the concepts, methods, and physical basis of EEWS has
been presented by Satriano et al. (2010).
PRESTo is a free and open source, highly configurable and easily portable
platform for earthquake early warning (Iannaccone et al.,
2010). PRESTo processes the real-time accelerometric data
streams from the stations of a seismic network to promptly detect the P-wave
arrival, provide the probabilistic and evolutionary estimates of location and
magnitude of earthquakes while they are occurring, as well as the shaking
prediction on a regional scale (Fig. 2). Alarm messages containing the
continuously updated estimates of source and ground motion at target
parameters, and their associated uncertainties, are sent over the Internet,
and can thus also reach distant vulnerable infrastructures before the arrival
of destructive waves, enabling the activation of automatic safety procedures.
Following the idea proposed by Zollo et al. (2010), PRESTo implements both a
regional and an on-site approach.
In its regional configuration (Fig. 2), PRESTo uses (a) a phase detector and
picker algorithm, which is optimised for real-time seismic monitoring and EEW
(Lomax et al., 2012); (b) a location algorithm, RTLoc (Satriano et al.,
2008), which locates earthquakes using information from both triggered and
not-yet-triggered stations, and which provides a fully probabilistic
description of the hypocentre coordinates and origin time; (c) the RTMag
algorithm (Lancieri and Zollo, 2008), a Bayesian approach that uses the peak
displacement (Pd) measured on the first seconds of the high-pass-filtered
signal on short time windows of P-waves (i.e. 2 and 4 s) and S-waves
(i.e. 1 or 2 s), and empirical correlation laws between this latter
parameter and the final earthquake magnitude (M); and (d) finally, ground
motion prediction equations (GMPE) that allow one to predict the peak ground
motion at target sites and at seismic stations using EEW location and
magnitude estimates.
The regional approach to early warning is integrated with an on-site,
threshold-based method for the definition of independent local alert levels
at each station. To this aim, the dominant period, τc, and the peak
displacement in a short time window after the first P-arrival time, Pd, are
simultaneously measured at each station, independently of the rest of the
seismic network. As shown by Zollo et al. (2010), Pd can be correlated with
the final PGV and consequently with the modified Mercalli intensity
(IMM), which is a measure of the expected damage, while
τc can be correlated with the earthquake magnitude. These two
parameters are compared with threshold values that define a decisional table
with four alert levels, declaring the expected earthquake effects nearby the
station or at distant sites. The alert level can be used to initiate safety
measures at each site independently of the regional processing. At the same
time, on the regional scale, the local alert levels, as they become
available, can be combined with the estimated source parameters to define the
extent of the potential damage zone (PDZ), i.e. the area in which the highest
intensity levels are expected (Zollo et al., 2010).
Snapshot of the PRESTo system during the playback of the
MW= 6.5, 1976 Friuli earthquake, at the instant when three
stations have triggered and the first alert is issued.
Since 2009, PRESTo has been under real-time experimentation in southern Italy
on the data streams of the Irpinia Seismic Network (ISNet). Moreover, in
order to analyse its performance in different seismic hazard contexts and
seismic networks of varying extensions, PRESTo is also currently operating in
other seismological centres (e.g. at the Korean Institute of Geoscience and
Mineral Resources, KIGAM, in South Korea, at the Kandilli Observatory and
Earthquake Research Institute, KOERI, in Turkey, and at the National
Institute of Research and Development for Earth Physics, NIEP, in Romania).
In addition, the feasibility study of a nation-wide early warning system in
Italy using the National Accelerometric Network (RAN) and PRESTo is in
progress.
EEW analysis of the 1976 Friuli earthquake data
One of the first tests that we carried out was devoted to verifying what
could have been the performance of PRESTo in the case of the 1976,
MW= 6.5 Friuli earthquake in northern Italy (Carulli and
Sleiko, 2005). To this aim, we realised an off-line run of the algorithm
(i.e. a playback) of this earthquake using the historical recordings
downloaded by ITACA 2.0 (Luzi et al., 2008; Pacor et al., 2011). The playback
was run considering the network geometry of 1976, but assuming the existence
at the time of the hardware and the management software necessary for the
real-time data streaming to the OGS's seismological centre of Udine.
Figure 3 shows a snapshot of the first event detection and characterisation
provided by PRESTo at the instant when only three stations have triggered and
the first alert is issued. Although based on few initial P-measurements, the
early magnitude estimation with only two stations (ML= 6.8) is
close to the final value (i.e. 6.5) as inferred from authoritative
catalogues.
The blind zone is the region where S-waves arrive before the first alert is
issued, and it corresponds to the circular area where no lead time is
available and no safety actions can be undertaken. Given the station's
available density at that time, we observe that the blind zone has a radius
of 36 km. Despite the fact that, under such conditions, the municipalities
affected by the most severe damage level could not have been alerted, the
comparison with the macroseismic field estimated by Giorgetti (1976) shows
that some of the municipalities in the area of intensity VII and most of
those in the area of intensity VI could have potentially received an alert
(Fig. 3). For instance, at the city of Pordenone (falling within the area of
intensity VII and located about 65 km from the epicentre), we measure a lead
time of about 9 s. Furthermore, for the area included within isoseismal
level VI (i.e. where the perceived ground shaking level is strong), the lead
time could have been between about 15 and 20 s (e.g. 14 s for Trieste, and
21 s for Treviso, Fig. 3). Considering the network geometry that exists
nowadays, we estimated that, for an event with the same epicentre of the 1976
one, the blind zone radius may shrink to about 22 km. For instance, in the
case of Pordenone, the lead time might increase to about 13 s.
Figure 4 shows, still for the playback of the Friuli 1976 earthquake, the PDZ
obtained estimated from the Pd measurements at the instant when the first
four stations have triggered. Interestingly, despite the PDZ not showing the
complex shape of isoseismal level VII, this was somehow expected given the
few stations available for the analysis; in first approximation this early
estimation of the damage extension matches reasonably well with the size of
the observed damage zone by Giorgetti (1976). As shown by Colombelli et
al. (2012) on Japanese data examples, whenever a dense network of stations is
available, the PDZ maps can reproduce the extension of the damage area well
(i.e. the area for which the observed macroseismic intensity is larger
than VII).
PRESTo performance on CE3RN
Since the beginning of 2014, PRESTo (version 0.2.7) has been under
experimentation in the transnational area including north-eastern Italy,
Slovenia and Austria. During this preliminary test phase, in order to avoid
overloading the Antelope system managing the CE3RN, a dedicated SeisComP
server (SeisComP, 2009) has been set up at the OGS CRS data centre in Udine
with the aim of collecting and converting in SeedLink (Heinloo, 2000) the
data of 20 accelerometric stations from the Antelope system (Fig. 1), and
pushing them towards a dedicated PRESTo system at RISSC-Lab in Naples.
After an initial period during which we tested different set-ups of the
system parameters, since the end of March 2014 we have been experimenting
with the velocity model used for routine earthquake analysis and bulletin
production at OGS (OGS, 1995–2013); a minimum number of five stations
required to trigger within 12 s for event declaration; the coefficients of
the empirical correlation laws between the peak displacement (Pd) measured on
short time windows of P-waves and the earthquake magnitude (M) estimated by
Lancieri and Zollo (2008); and the Akkar and Bommer (2007) ground motion
prediction equation.
Same as Fig. 3 but showing the PDZ (pink area) corresponding to
real-time estimation of the area with macroseismic intensity equal to or
higher than VII.
Distribution of time of the first alert (a) and dimension
of the blind zone (b) for the grid of synthetic sources.
CE3RN stations (yellow). Location error within 10 km (green),
between 10 and 50 km (orange), and larger than 50 km (red).
Since the station distribution has a key role in determining the resilience
of a system, that is to say the network rapidity in issuing EEW alerts, we
estimated for the CE3RN network the time of the first alert and the
blind-zone extent when three stations have triggered (Fig. 5). The analysis
was carried out considering a grid of virtual seismic sources (i.e. a node
each of 0.05∘× 0.05∘ for a total of 9801 nodes) with
a fixed depth at 6.4 km.
Following Picozzi et al. (2015b), the time of the first alert is defined as
the time when P-waves reached the third station of the network. Furthermore,
the BZ is defined as the sum of three delays: (1) the time of the first
alert, (2) a fixed delay for the telemetry and computation equal to 2 s,
selected according to the value recorded with PRESTo at the ISNet
accelerometric network in southern Italy over a long period of testing
(Satriano et al., 2011), and (3) the constraint of having 2 s long P-wave
time windows at an N-1 station used by RTLoc, which is the needed information
for RTMag to estimate the magnitude. This latter constraint is due to the
fact that at the instant when RTLoc locates an event with N stations, RTMag
provides the first magnitude estimation using N-1 stations, under the
condition that they recorded at least 2 s of P-waves. Finally, the sum of
these three times is converted in the radius of BZ by multiplying it by the
S-wave velocity, assuming that this latter value is equal to 3 km s-1.
Figure 5a shows that the time of the first alert is less than or equal to
10 s for the central area of the network, which includes the Friuli 1976
earthquake's epicentre and the Italian–Slovenian boundary, where the
station's density is high. The first alert time, and the smallest as well as
the larger values, are in general elongated approximately in the east–west
direction, according to the network geometry. Also, the blind-zone map shows
a similar trend, having the smallest values (i.e. below 25 km) in the
Friuli 1976 earthquake's epicentre area, with larger values towards the
network boundaries (Fig. 5b).
Concerning the real-time testing of the EEWS, since the end of May 2014, that
is to say when a stable configuration of the EEWS was found, PRESTo
(version 0.2.7) detected in real time 23 earthquakes, while one event was
missed (i.e. event no. 21, Table 1).
Figure 6 shows that the performance of the system in locating the earthquakes
has in general been very good, with 18 events out of 23 located within 10 km
of the authoritative value. Only in one case is the discrepancy between
10 and 50 km, and in four cases it is larger than 50 km. Concerning the
depth estimation, it must be kept in mind that 90 % of the events in this
region are related to a seismogenic layer placed at a depth of about 8 km
(Gruppo di lavoro MPS, 2004). The peculiar distribution of events in depth,
together with the observation that, given the Pd vs. M relationship
adopted, location discrepancies of the order of 15 km determine a magnitude
error within 0.5 magnitude units, led the depth estimation to be, for the
moment, a parameter of minor importance in our experiment. Similarly to what
was already observed in the Irpinia region (Satriano et al., 2011), the
hypocentral locations for the events inside the network are generally well
constrained starting from the very first estimates. For the events outside
the network, the azimuth is well determined, but there is typically a larger
uncertainty in the distance.
Same as Fig. 1 but showing the correctly detected (green), missed
(red), and false (blue) events.
In order to quantitatively assess the EEWS performance, we compared the EW
magnitude (MEW) with the authoritative one (MBULL),
and we declared success (S) when MEW falls within a
±0.5 interval around MBULL, missed (M) when MEW
is lower than MBULL- 0.5 units, and false (F) when
MEW is higher than MBULL+ 0.5 units. Table 1
shows that the system had 17 successful detections (70.8 %), 3 false
detections (12.5 %), and 4 missed events (16.7 %), of which 2 were
detected but with underestimated magnitude, 1 was a MB,
5 occurred in Greece (event no. 20, http://cnt.rm.ingv.it), and 1 was
not detected (event no. 2). In general, we observed that both the
mis-detection and the wrong location and magnitude estimation occurred when
the events were located out of the network, or where the latter has a lower
station density (i.e. no. 16, no. 20, and no. 21, Table 1). On the contrary,
Fig. 7 shows that when the events occur in the area of higher station
density, which also corresponds to higher seismic risk areas, the estimation
of EEW magnitude is generally correct. Figure 8 shows, as an example, the
good detection of event no. 1 (Table 1) that occurred in Slovenia.
Earthquake detected by PRESTo at CE3RN during the period from
May to December 2014. The early warning (EW) estimates are compared with
those of the OGS-CRS bulletin (BULL; from OGS, 1995–2013). EEW performance:
success (S), missed (M), false (F).
IDDate (yyyy-MBULLLonBULLLatBULLMEWLonEWLatEWMEW-Time firstEEWmm-dd) and(±0.3)(∘)(∘)(∘)(∘)MBULLinfo loc.perf.time (UTC)& M (s)129 May 20143.813.86246.0983.513.851146.0967-0.310.0S07:24:18.6322 Jun 20142.012.91546.4142.012.966246.4153013.8S02:15:03.02319 Jun 20142.614.11446.1372.814.195545.45270.215.5S11:26:21.40424 Jun 20142.713.76246.2372.813.995246.63620.140.2S22:43:25.39529 Jun 20142.112.91646.4141.912.876246.3952-0.250.2S18:39:32.1565 Jul 20141.713.34246.4181.813.371946.44470.132.8S15:01:14.5775 Jul 20141.613.34446.4181.713.371946.45470.151.6S15:47:05.5087 Jul 20142.812.20646.0013.112.307145.98760.314.1S10:50:38.87920 Jul 20142.413.66846.4863.710.229345.06431.371.5F14:44:13.581025 Jul 20141.912.97246.3981.612.966246.3952-0.368.1S06:32:00.58118 Aug 20142.612.91746.3613.112.932546.35020.513.2S12:14:16.38121 Sep 2014–––2.313.964646.1859–23.9F00:50:52.701312 Sep 20142.213.40146.4552.013.428346.4495-0.221.2S15:50:52.851412 Sep 20142.013.40546.4552.013.473546.4694057.0S15:53:45.061518 Sep 20142.212.93746.3561.812.932546.3502-0.411.5S14:24:41.45165 Oct 20142.510.99744.6313.911.085844.62981.4180.5F07:09:23.001722 Nov 20141.913.65046.3162.013.674846.33820.1103.9S03:22:35.41185 Dec 20142.812.83546.4182.212.835746.4183-0.611.5M09:11:36.31197 Dec 20141.813.62046.1132.113.620846.11380.392.1S08:00:32.352011 Dec 20144.920.44438.4783.714.526644.6075-1.244.6M22:26:02.392112 Dec 20143.511.14644.866–––M07:01:25.002218 Jan 20152.912.89046.3352.712.853846.3351-0.212.0S14:42:23.982322 Jan 20151.712.83846.4081.312.831146.4151-0.410.2S15:34:35.272430 Jan 20154.113.14846.3913.513.146346.3751-0.68.4M00:45:48.51
Snapshot of the PRESTo system during the 29 May 2014
ML= 3.8 Slovenian earthquake (event no. 1, Table 1).
Recently, on the occasion of the M= 4.1 event that occurred nearby the
town of Udine, Italy (i.e. event no. 24 of 30 January 2015; Table 1), we
observed that PRESTo provided a correct location, but estimated an EW
magnitude 0.6 units less than the authoritative one (3.5 MEW,
4.1 MBULL). The location being accurate, we guessed that the
discrepancy between the early warning and the bulletin magnitude estimates
might be related to the parameters of the peak displacement (Pd) vs. M
relationships. We decided to investigate this case in more detail by using
the recordings of this event to run an off-line PRESTo playback. In
particular, the playback was run using new parameters of the Pd vs. M
relationship derived from the local magnitude law used by the INGV. Figure 9
shows that the new law provides potential MEW estimates in better
agreement with the MBULL (4.2 MEW,
4.1 MBULL). However, it is worth mentioning that the magnitude
law used by the INGV is the one computed for the southern California region,
which mostly adheres to actual Italian data for station–hypocentre distances
greater than 100 km, whereas it overestimates the local magnitude at closer
stations (M. Di Bona, personal communication, 2015; http://iside.rm.ingv.it). More tests on
this point are needed before drawing a conclusion.
Concerning the few wrong event characterisations, we guess that the low
magnitude of the events might have played a major role. Indeed, small
magnitudes lead to a low signal-to-noise ratio of the recordings, which in
turn makes the real-time analysis more difficult than in the case of moderate
size events. This issue can be overcome by considering velocity streams, a
feature that we included in the newest version of PRESTo (PRESTo 0.2.8;
http://www.prestoews.org) and that in the near future will also be
adopted at CE3RN.
Snapshot of the playback of the PRESTo system during the
30 January 2015 ML= 4.1 earthquake (event no. 24, Table 1)
using new parameters of the log(Pd) vs. M relationship.
CE3RN stations (yellow). Delays less than 15 s (blue) or
larger (red).
The time when the first EEW information on the location and magnitude of the
earthquake was available is also reported in Table 1, as the time after the
first P arrival detected at a CE3RN station (Fig. 10). We observe that,
in 10 cases, the EEW information is available within 15 s (the minimum value
of 8.4 s has been observed for event no. 24 of 30 January 2015), while in
13 cases the delay was larger than 15 s (the maximum vale was 180.5 s for
event no. 16 of 22 November 2014). The spatial distribution of the delays
(Fig. 10) highlights that, for EEW purposes, the reasons for the larger
telemetry delays of stations in the Slovenian sector should be better
investigated.
Conclusions
This work presents the preliminary results of a feasibility study carried out
with EEW platform PRESTo in the high seismic hazard region including
north-eastern Italy, Slovenia and Austria, where the 1976 Friuli earthquake
occurred.
Results from the offline analysis using the software platform for EW, PRESTo,
indicate that, despite the network geometry at that time being rather poor,
the EEWS could have been potentially very useful. Indeed, we estimated that
the blind-zone radius could have been of the order of 36 km, and that
municipalities located within the intensity VI and VII areas could have
potentially benefited from an alert. Of course, implementing an EEWS
requires, besides these scientific aspects, many further issues to be taken
into consideration. For instance: the definition of actions that end users
could effectively put into operation within the available lead time for the
reduction of their exposure to the seismic risk; cost–benefit analysis of
the aforementioned actions; the definition, test, and validation of the
procedures which allow the implementation of these mitigation actions; a
comprehensive campaign of information on what has to be done; and, finally, a
clear attribution of the responsibilities.
Interestingly, we also found that, in the case of a large event with a
similar epicentre to the 1976 Friuli earthquake, the performance of the EEWS
would improve, considering the actual CE3RN network configuration. In
particular, for such a scenario, we found that three major centres in the
region (i.e. Pordenone, Trieste, and Ljubljana) could fall within isoseismal
level VI (i.e. experiencing a strong ground shaking) but potentially benefit
from a lead time longer than 10 s. As discussed by Goltz (2002), when the
population is trained to rapidly respond and take protective measures
(e.g. duck and cover, turn off gas burners, move away from windows or
equipment, etc.), even fewer than 10 s can help to reduce the risk of injury
from an earthquake's secondary effects.
During the period May–December 2014, PRESTo detected in real time
23 earthquakes in the magnitude range 1.7 to 4.1, of which 14 were correctly
detected, while 4 and 3 events resulted in missed and false alerts,
respectively. Despite the testing period still being too short to come up
with definitive conclusions, it seems that the EEWS given by the integration
of PRESTo and CE3RN is efficient with respect to earthquakes that occur
nearby the area with higher station density. Nevertheless, more testing and an
improvement in the system are necessary to cope with events occurring out of
the network, and in general where it has a lower station density. With
respect to this last issue, we are evaluating to increase the network
density, including in the EEWS also stations with velocimetric sensors.
The testing period of the EEW system is carried out primarily with the goal
of highlighting the existence of weak points (i.e. in the hardware, network
management and analysis software with respect to the seismicity of the area).
In fact, besides the specific characteristics of an EEW algorithm, the
performance of an EEW system strongly depends also on technological issues,
like for example the efficiency of the data telemetry and the seismic noise
level at the stations. For this reason, especially these latter two aspects
will be studied in the next tests of the EEWS. Of course, the realisation of
the EEWS in the area monitored by CE3RN will be accompanied by an
extensive activity of communication and training, specifically tailored for
both the population and the different stakeholders.
Besides the standard application of EEW, the use of PRESTo in the area
surveyed by CE3RN will give a potential benefit to local civil
protection agencies. In the case of a very strong shock, the standard
monitoring network equipped with modern BB sensors has a saturation zone that
may hamper immediate response (e.g. see Fig. 2 from Faenza et al., 2011) in a
radius of the order of 100 km. This means that the epicentral location is
available when the strongest S-wave phase has already affected the area. On
the contrary, an EEW system may broadcast the information to civil protection
centres before the strong ground motion can cause potential failure or
hampering of the communication system. Hence, civil protection would have the
information necessary to act immediately, according to the severity of the
situation.
Acknowledgements
We would like to thank the Associate Editor J. Clinton, C. Cauzzi and an
anonymous reviewer for their comments and suggestions that allowed us to
significantly improve the manuscript's content and form.
This work has been partially supported by the REAKT-Strategies and tools for
Real Time Earthquake RisK ReducTion FP7 European project funded by the
European Community's Seventh Framework Programme (FP7/2007-2013) under grant
agreement no. 282862. Edited by: J. Clinton
Reviewed by: C. Cauzzi and two anonymous referees
References
Akkar, S. and Bommer, J. J.: Empirical prediction equations for peak ground
velocity derived from strong-motions records from Europe and the Middle East,
Bull. Seismol. Soc. Am., 97, 511–530, 2007.Alcik, H., Özel, O., Apaydin, N., and Erdik, M.: A study on
warning algorithms for Istanbul earthquake early warning system, Geophys.
Res. Lett., 36, L00B05, 10.1029/2008GL036659, 2009.
Allen, R. M., Gasparini, P., Kamigaichi, O., and Böse, M.: The status of
earthquake early warning around the world: an introductory overview, Seismol.
Res. Lett., 80, 682–693, 2009.Böse, M., Ionescu, C., and Wenzel, F.: Earthquake Early Warning for
Bucharest, Romania: Novel and revised scaling relations, Geophys. Res. Lett.,
34, L07302, 10.1029/2007GL029396, 2007.Böse, M., Hauksson, E., Solanki, K., Kanamori, H., and Heaton, T. H.:
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, L00B03, 10.1029/2008GL036366, 2009.Bragato, P. L., Costa, G., Gallo, A., Gosar, A., Horn, N., Lenhardt, W.,
Mucciarelli, M., Pesaresi, D., Steiner, R., Suhadolc, P., Tiberi, L.,
Živčić, M., and Zoppé, G.: The Central and Eastern
European Earthquake Research Network – CE3RN, EGU General Assembly 2014,
27 April–2 May 2014, Vienna, Austria, 2014.
Carulli, G. B. and Slejko, D.: The 1976 Friuli (NE Italy) earthquake,
Giornale di Geologia Applicata, 1, 147–156, 2005.Colombelli, S., Amoroso, O., Zollo, A., and Kanamori, H.: Test of a Threshold-Based Earthquake Early
Warning Using Japanese Data, Bull. Seismol. Soc. Am., 102, 1266–1275,
10.1785/0120110149, 2012..
Espinosa-Aranda, J. M., Cuellar, A., Garcia, A., Ibarrola, G., Islas, R.,
Maldonado, S., and Rodriguez, F. H.: Evolution of the Mexican Seismic Alert
System (SASMEX), Seismol. Res. Lett., 80, 694–706, 2009.Faenza, L., Lauciani, V., and Michelini, A.: Rapid determination of the
shakemaps for the L'Aquila main shock: a critical analysis, Bollettino di
Geofisica Teorica ed Applicata, 52, 407–425, 10.4430/bgta0020, 2011.
Giorgetti, F.: Isoseismal map of the May 6, 1976 Friuli earthquake,
Boll. Geofis. Teor. Appl., 19, 707–714, 1976.
Goltz, J. D. L.: Introducing earthquake early warning in California: A
summary of social science and public policy issues, technical report,
Governor's Off. of Emergency Serv., Pasadena, Calif., 2002.Gruppo di lavoro MPS: Redazione della mappa di pericolosita` sismica
prevista dall'Ordinanza PCM 3274 del 20 marzo 2003, Rapporto conclusivo per
il dipartimento di Protezione Civile, INGV, aprile 2004, Milano, Roma, 65 pp. + 5 appendici,
available at: http://zonesismiche.mi.ingv.it/elaborazioni/ (last access: 31 March 2015), 2004.Heinloo, A.: SeedLink design notes and configuration tips,
http://geofon.gfz-potsdam.de/geofon/seiscomp.de/geofon/seiscomp/seedlink.html
(last access: 31 March 2015), 2000.
Horiuchi, S., Negishi, H., Abe, K., Kamimura, A., and Fujinawa, Y.: An
automatic processing system for broadcasting system earthquake alarms, Bull.
Seismol. Soc. Am., 95 347–353, 2005.Hoshiba, M.: Real-time correction of frequency-dependent site amplification
factors for application to Earthquake Early Warning, Bull. Seismol. Soc. Am.,
103, 3179–3188, 10.1785/0120130060, 2013.Iannaccone, G., Zollo, A., Elia, L., Convertito, V., Satriano, C., Martino, C.,
Festa, G., Lancieri, M., Bobbio, A., Stabile, T. A., Vassallo, M., and Emolo, A.: A
prototype system for earthquake early-warning and alert management in
southern Italy, Bull. Earthq. Eng., 8, 1105–1129, 10.1007/s10518-009-9131-8, 2010.Lancieri, M. and Zollo, A.: Bayesian approach to the real-time
estimation of magnitude from the early P and S wave displacement peaks, J.
Geophys. Res., 113, B12302, 10.1029/2007JB005386, 2008.Lomax, A., Satriano, C., and Vassallo, M.: Automatic picker developments and
optimization: FilterPicker – a robust, broadband picker for real-time
seismic monitoring and earthquake early-warning, Seismol. Res. Lett., 83,
531–540, 10.1785/gssrl.83.3.531, 2012.
Luzi, L., Hailemikael, S., Bindi, D., Pacor, F., Mele, F., and Sabetta, F.:
ITACA (ITalian ACcelerometric Archive): a web portal for the dissemination of
Italian strong-motion data, Seismol. Res. Lett., 79, 716–722, 2008.
OGS: Bollettino della Rete Sismometrica del Friuli–Venezia
Giulia e del Veneto, OGS (Istituto Nazionale di Oceanografia e di Geofisica
Sperimentale), Centro di Ricerche Sismologiche, Udine, Italy, 1995–2013.Pacor, F., Paolucci, R., Ameri, G., Massa, M., and Puglia, R.: Italian
strong motion records in ITACA: overview and record processing, Bull.
Earthq. Eng., 9, 1723–1739, 10.1007/s10518-011-9327-6, 2011.
Peng, H. S., Wu, Z. L., Wu, Y. M., Yu, S. M., Zhang, D. N., and Huang, W. H.:
Developing a prototype earthquake early warning system in the Beijing
Capital Region, Seismol. Res. Lett., 82, 394–403, 2011.Picozzi, M., Colombelli, S., Zollo, A., Carranza, M., and Buforn, E.: A
Threshold-Based Earthquake Early-Warning System for Offshore Events in
Southern Iberia, Pure Appl. Geophys., 10.1007/s00024-014-1009-2, in press, 2014.Picozzi, M., Emolo, A., Martino, C., Zollo, A., Miranda, N., Verderame, G.,
Boxberger, T., and the REAKT Working Group: Earthquake Early Warning
System for Schools: A Feasibility Study in Southern Italy, Seismol.
Res. Lett., 86, 398–412, 10.1785/0220140194, 2015a.Picozzi, M., Zollo, A., Brondi, P., Colombelli, S., Elia, L., and Martino,
C.: Exploring the Feasibility of a Nation-Wide Earthquake Early Warning
System in Italy, J. Geophys. Res.-Solid, 10.1002/2014JB011669, in press, 2015b.Satriano, C., Lomax, A., and Zollo, A.: Real-Time Evolutionary
Earthquake Location for Seismic Early Warning, Bull. Seismol.
Soc. Am., 98, 1482–1494, 10.1785/0120060159, 2008.Satriano, C., Wu, Y.-M., Zollo, A., and Kanamori, H.: Earthquake early warning:
Concepts, methods and physical grounds, Soil Dynam. Earthq. Eng., 31, 106–108,
10.1016/j.soildyn.2010.07.007, 2010.
Satriano, C., Elia, L., Martino, C., Lancieri, M., Zollo, A., and Iannaccone,
G.: PRESTo, the earthquake early warning system for Southern Italy:
concepts, capabilities and future perspectives, Soil. Dyn. Earthq. Eng.,
31, 137–153, 10.1016/j.soildyn.2010.06.008, 2011.SeisComP: Seismological communication processor, 2009, http://geofon.gfz-potsdam.de/geofon/seiscompS,
last access: January 2010.Steim, J. M. and Reimiller, R. D.: Timeliness of Data Delivery from Q330 Systems,
Seismol. Res. Lett., 85, 844–851, 10.1785/0220120170, 2014.Wu, Y. M. and Zhao, L.: Magnitude estimation using the first three seconds
P-wave amplitude in earthquake early warning, Geophys. Res. Lett., 33,
L16312, 10.1029/2006GL026871, 2006.Zollo, A., Amoroso, O., Lancieri, M., Wu, Y. M., and Kanamori, H.: A
threshold-based earthquake early warning using dense accelerometer networks,
Geophys. J. Int., 183, 963–974, 10.1111/j.1365-246X.2010.04765.x, 2010.Zollo, A., Colombelli, S., Elia, L., Emolo, A., Festa, G., Iannaccone, G.,
Martino, C., and Gasparini, P.: An Integrated Regional and On-Site Earthquake
Early Warning System for Southern Italy: Concepts, Methodologies and
Performances, in: Early Warning for Geological Disasters, Advanced Technologies
in Earth Sciences, edited by: Wenzel, F. and Zschau, J., Springer,
Berlin, Heidelberg, 117 pp., 10.1007/978-3-642-12233-0_7, 2014.