Onsite Earthquake Early Warning Predictive Models for Acceleration

On-site Earthquake Early Warning: Predictive Models for Acceleration Response Spectra Considering Site-Effects* Iaccarino A. G. , Picozzi M. , Bindi D. & Spallarossa D. Antonio Giovanni Iaccarino University of Naples, Federico II Department of Physics RISSCLab May 06 th 2020 EGU 2020 *Recently accepted for publication on BSSA

On-Site Earthquake Early Warning Systems TARGET SEISMOGENIC ZONE Telecommunication (light velocity) P-wave (~6 km/s) S-wave (~3 km/s) Scale laws P-wave cinematic parameters at target Ground motion parameters at target Pd A. G. Iaccarino 12. 13. 2019 AGU 2

Dataset 1024 records from 58 earthquakes of the 2016 -17 Central Italy sequence with Mw between 3. 7 and 6. 5 at 100 stations at hypocentral distances smaller than 150 km. Stations, colored by EC 8 classification Events 2016/08/24, Mw 6. 0, Amatrice 2016/10/26, Mw 5. 9, Visso 2016/10/30, Mw 6. 5, Norcia A. G. Iaccarino 12. 13. 2019 AGU 3

Site effects in EEW laws M=5. 4 R=33 km Pd=0. 030 cm Iv 2=0. 0054 cm 2/s PGA=112 cm/s 2 Pd=0. 034 cm Iv 2=0. 0050 cm 2/s PGA=76. 7 cm/s 2 M=5. 4 R=22 km A. G. Iaccarino 12. 13. 2019 AGU 4

*(Pinheiro & Bates, 2000) Mixed-effect regression * δS 2 S 1 ϵ 1 i Ground motion parameter Residual ACT σSS σ T 1244 Group index δS 2 S 2 Corrected Residual ϵ 2 i Site-to-site coefficient A. G. Iaccarino 12. 13. 2019 AGU 5

EEWS laws calibration 9 periods, T= 0. 1, 0. 15, 0. 2, 0. 3, 0. 5, 0. 75, 1. 0, 1. 5, 2. 0 s Station grouping EC 8 grouping A. G. Iaccarino 12. 13. 2019 AGU 6

Study of residuals of “EC 8” relations It’s visible a general trend in the EC 8 residuals But all the classes residuals are consistent with zero. A. G. Iaccarino 12. 13. 2019 AGU 7

Study of residuals of “station” relations There is a great variations in mean normalized residuals through the stations Many of the stations, about 20%, present a normalized residual inconsistent with zero. A. G. Iaccarino 12. 13. 2019 AGU 8

γ residuals vs ϵ residuals A. G. Iaccarino 12. 13. 2019 AGU 9

δS 2 S vs H/V ratio Ground motion underestimated by classic model Ground motion overestimated by classic model A. G. Iaccarino 12. 13. 2019 AGU 10

Leave-one-out cross-validation Perform the procedure on the reduced dataset Take a data out from the dataset Pd-Station Iv 2 -Station When we group by station, we have a strong decrease in residuals variability Pd-EC 8 Compute residual from the retrieved models Iv 2 -EC 8 Grouping by EC 8 has not an appreciable impact on residuals variability A. G. Iaccarino 12. 13. 2019 AGU 11

EEWS simulation RSA Critical Event RSAO>=Alert threshold FA SA SNA MA We simulate the EWS on ACT and T 1244 using each periods as a single data EW Alert P(RSAE>=Alert threshold) >= Exceedance threshold Alert RSA Minson et al. , 2019 A. G. Iaccarino 12. 13. 2019 AGU Log(RSA) 12

EEWS simulation Mw 6. 5, Norcia case for the two stations Real data A. G. Iaccarino 12. 13. 2019 AGU 13

EEWS simulation Median model Mixed-effect model Considering station corrections improves very well the EW performance A. G. Iaccarino 12. 13. 2019 AGU 14

Conclusions • A. G. Iaccarino 12. 13. 2019 AGU 15

Thank you for the attention! A. G. Iaccarino 12. 13. 2019 AGU 16