SECOND EUROPEAN CONFERENCE ON EARTHQUAKE ENGINEERING AND SEISMOLOGYISTANBUL | Turkey | Aug. 25-29, 2014

Feasibility study of a nation-wide Early Warning System: the application of the EEW software PRESTo on the Italian Strong Motion Network (RAN)

Matteo Picozzi, Aldo Zollo, Luca Elia, Claudio Martino, Piero Brondi, Simona Colombelli, Antonio Emolo,

Gaetano Festa, and Sandro Marcucci

Worldwide EEWS

At the nation-wide scale, the Japanese system uses ~1,000 seismic instruments across Japan, 200 operated by JMA and 800 by NIED, and integrates methodologies developed by JMA and NIED

272 Free Field Stations:• Force Balance 3-C. Accelerometer• 18 bit digitizer• GSM modem• PGA via SMS

~ 500 Digital Strong Motion Stations:• Local Storage on PCMCIA disk• GSM / GPRS Modem to send waveforms cut between triggering and de-triggering Trigger at 0.1% g acceleration or on STA/LTA threshold (newer stations)• Message containing PGA within 5 min.

192 Stations in CabinsForce Balance 3-C. Accelerometer• 24 bit digitizer• GPRS router. No wait for coda to send data.• PGA via e-mail

• RAN• RAN + ISNet• Seismicity The Communication

Network Needs to beExpanded and

Improved!

The Italian Strong Motion Network (RAN)

3

Simulation analyses and scenario studies

CONCLUSION

Effect of network geometry

Offline analysis of recent M≥4.5 earthquakes

Feasibility of EW in Italy based on RAN

Working hypotheses:- RAN in its actual configuration is upgraded to operate in real-time

mode- Telemetry and data processing delays (1+1 sec) are those measured at

ISNET using PRESTo

PRESTo Probabilistic and evolutionaRy Early warning SysTem

P

An integrated software platform for real data processing andseismic alert notification

Automatic procedures for the probabilistic and evolutionary estimation of source parameters and prediction ofground motion shaking.

Automatic Picking

RT Earthquake Location

RT Magnitude Estimation

PGx Prediction at TargetsR E G I O N A L

On-Site Alerts

ON-SITE

Satriano & Elia (2010). PRESTo, the earthquake early warning system for Southern Italy: Concepts, capabilities and future perspectives. Soil Dyn Earthquake Eng

PLUS

http://www.prestoews.org/

Four Macro-Zones based on Mmax from PSHA :I) Seismic Zones (ZS) with Mmax ≥ 6.5

(1 st. in ~300 km2, aver. Inter-st. Dist. 17.6 km, ~ K-NET in Japan 19 km ) II) ZS with 6.5 > Mmax ≥ 6 (1 st. in ~ 540 km2; ~ 23 km)

III) ZS with 6 > Mmax > 5 (1 st. in ~ 620 km2; ~ 25 km)

IV) Outside ZS (1 st. in ~ 1100 km2; ~ 34 km) Mmax=5

Eff ect of network geometry

PSHA (37 Seismic Zones, 16921 nodes)

from INGV (http://esse1-gis.mi.ingv.it/

s1_en.php)

First Alert using 3 & 6 stati ons3 stations 6 stations

PSHA

Each of the 16921 nodes from PSHA is considered a source

Time of first alert between 4 and 12 sec

Time of first alert between 5 and 15 sec

Blind Zone using 3 & 6 stati ons

PSHA

3 stations 6 stationsBlind Zone radius between 30 and 49 km

Blind Zone Radius between 35 and 57 km

To compute the BZ radius: P-arrival time, P-wave time window, average telemetry and computation times at ISNET.

DM defined as the PGV+σ corresponding to the Instr.Int. VII class from Faccioli & Cauzzi (2006)

Effectivness of EW: Potential damage zone vs BZ zone

Example on the ‘80 Irpinia Mw 6.9 scenario

Municipalities

EWZ (PGV aver.)Lead-time 15.9 s

EWZ (PGV+1σ)Lead-time 33.6 s

EWZ (PGV-1σ)Lead-time 5.8 s

BZ

Offl ine analysis of recent earthquakes

RTLOC

RTMAG40 EQs, Mw≥4.5, 2002-2013 from ITACA 2.0 (http://itaca.mi.ingv.it; Luzi et al., 2008; Pacor et al., 2011)

Using 3 stations ΔM < 0.5

T 1st Alert T 1st Alert

Error on hypocenteral location

Real-ti me locati on perfomance

16921 hypothetical seismic sources (0.05x0.05°) spacing

For each node, the P-wave arrival times at 3 stations, are extracted assuming a gaussian reading error of 1 secondR T L O C

Average Performance at National scale over 10 runs

Real-Time magnitude esti mati on

Performance at the National scale using 3 stations

Percentage of successes (Mest Mtrue±0.5) at a national scale, using the first 3 stations.

At each node : 10 simulated sequences in 50 years with 5<MW>Mmax

This study does not include the EW operability, which asks for massive experimental testing and close involvement of end-users

The analysis of historical earthquake recordings and synthetics suggests that the integration of the RAN and PRESTo in an EEWS can provide, especially for the higher seismic hazard areas, reliable alert messages within about 5-10 seconds

Expected errors on location and magnitude estimation,although large, are acceptable for peak ground motion predictions.

The RAN seems to have the potential for a Nation-wide EEWS, but: The Communication Network Needs to be Expanded and

Improved A Blind Zone extent of 30km is not acceptable for M 6 eqks The

station density must be increased and onsite method should be used

Conclusions

Thanks for your attention

Real-Time magnitude esti mati on

For a given earthquake source and the closest RAN stations, the peak displacement (PD) is randomly extracted from the PD-M relationship.

Example for the 50 years EQ. sequences at the node of the 1980’ Irpinia event

PD values Input vs EEW M values

Average RTMag success, false, and missed rate (in %) for the four MZ in case three stations are used.

Success: Mest Mtrue±0.5False: Mest >Mtrue+ 0.5Missed: Mest <Mtrue- 0.5

Extensive tests on scenarios

INTRODUCTION Earthquake Early Warning The Italian Accelerometric Network (RAN)

CONCLUSION

Geometrical considerations

Playbacks onhistorical EQs

Outline

Real -T ime magnitude esti mati on

Example for the 50 years EQ. sequences at the node of the 80’ Irpinia event

For a given earthquake source and the closest RAN stations, the peak displacement (PD) is randomly extracted from the PD-M relationship.

PD values Input vs EEW M values

Performance at National scale (3 st.)

Average RTMag success, false, and missed rate (in %) for the four MZ in case three stations are used.

With 6 stations