ShakeMap® Implementation in Italy Istituto Nazionale di Geofisica e Vulcanologia: Michelini A., Faenza L., Lauciani V. (Email: [email protected] [email protected] [email protected]) Observations: PGA, PGV and PSA response at 0.3, 1 e 3s with 5% damping all on horizontal components ~ 260 Italian Stations total including short period but not used for shakemaps ~50 Accelerometer ~140 Broad-band Other stations V s30 Site Corrections: classification based on a detailed geology map of Italy [ . The geology units have been gathered into five classes, according to the EuroCode8 provision. References: Ambraseys et al. (1996), Earthquake eng. Struct. Dyn, 25,371-400; Bommer et al. (2000), Proceedings of the 12th World Conference on Earthquake Engineering, paper no. 207 Carta Geologica d’Italia, scale 1:100000; Morasca et al. (2006), Journal of Seismology, 10, 351-333; Malagnini et al. (2002), Bull. Seismo. Soc. Am., 92, 6, 2186-2204; Malagnini et al. (2000), Bull. Seismo. Soc. Am. 90, 1062-1108; MPS Working Group(2004), http://zonesismiche.mi.ingv.it; Wald et al. (2006),http://pubs.usgs.ov/tm/2005/12A01/pdf/508TM12A1.pdf; Wald and Allen (2007), Bull. Seism. Soc. Am., 97, 1279-1395. Acknowledgements: This research has been supported by the EU-SAFER project (contract n. 036935) and the INGV-DPC (2005-2007) project S4. The authors thank the DPC for the data of the 1st of March Mugelllo sequence. Ground Motion Predictive Equations (GMPEs): a) Regionalization of the attenuation relations for M <5.5 events; b) PGA for M ≥ 5.5 events adopts the relation of Ambraseys et al. (1996) (light blue). c) PGV for M ≥ 5.5 events adopts the relation of Bommer et al (2000) (light blue). Colors in b) and c) match the zones in a). b) c) Discussion and Conclusions Since October 2006, shakemaps have been published for188 M3+ events (http://earthquake. rm . ingv .it ). 1. Data Acquisition and Processing: Currently, shakemaps are available within 15-30 minutes after an event, using a revised location In the near future, almost real-time shakemap will be generated automatically after an event (within 5/6 min). 2. Testing the importance of the data: The shakemaps have been found to be generally reliable. The results shown in Box A confirm an overall adequacy of the adopted GMPEs and local geology corrections when attempting to represent the true ground shaking. The inclusion of the near field data (BOX B: INGV+RAN data) improved the detail of the shakemaps and the BIAS compensates for possible erroneous estimations of magnitude. The test in BOX C, where no BIAS correction is applied, shows that the current GMPEs adopted tend to overestimate the PGM near the epicenter. The tests show that availability of dense strong motion station data are of much relevance when determining shakemaps and that the BIAS correction is important to compensate for incorrect initial magnitude estimates. Published shakemaps for M 3.0+ (188 evens since October 2006) IMPLEMENTATION The March 1 st , 2008, Mugello sequence of medium size events which occurred in Northern Apennines, featured three events with magnitude 4+ and many smaller ones. In this example we report the results obtained from the analysis for the first event, with M 4.2 (2008:03:01 08:43 MST). BOX A: Comparison between the shakemaps processed with no observed data and with INGV stations only (red and purple triangles). The %Diff and the Δ PGX comparison indicators are small at the stations. The BIAS (a factor within ShakeMap® designed to minimize the difference between the data values at the seismic stations and those predicted by the GMPEs and expressed in log scale) indicates a good agreement (BIAS ≈1.0). PROBLEM: the INGV stations are not in proximity of the epicenter. BOX B: Inclusion of the RAN data, the strong motion network operated by DPC (yellow triangles in the INGV+RAN data panels) shows relevant differences in the near field with INGV+RAN data shakemaps having values as much as 30% higher than those with INGV data only. The BIAS in this case is also larger (1.37) and a possible explanation is given by an underestimation of magnitude. This could in fact be the case since moment magnitudes, Mw, determined in various ways, have always indicated values as large as 4.5 for this event. BOX C: shakemaps with M 4.5 imposing a BIAS=1( i.e., no BIAS calculation). %Diff and the Δ PGX show differences in the areas west and in proximity of the epicenter. For the amplifications to the west, this could be attributed to site effects not captured by the adopted amplification correction. For the near-field differences, we observe that the ground-motion predictive relationships tend to over-estimate the PGA, compared to the real shaking. This would imply that the adopted GMPEs while describing properly the attenuation at larger distances have difficulties to account accurately for the near-field variability. No Data - Scenario Based on INGV data No Data Based on INGV data Based on INGV data Bias 0.98 "PGX = (PGXINGV # PGXNoData) %Diff = "PGX PGXINGV $100 " PGX = (PGX INGV # PGX NoData ) BOX A Based on INGV+RAN data Based on INGV data Based on INGV+RAN data Based on INGV data " PGX = (PGX INGV+RAN # PGX INGV ) Based on INGV data Bias 0.98 Based on INGV+RAN data Bias 1.37 "PGX = (PGXINGV+RAN # PGXINGV) %Diff = "PGX PGXINGV+RAN $100 B O X B Based on INGV+RAN data Based on INGV data Based on INGV+RAN data Based on INGV data " PGX = (PGX INGV+RAN # PGX INGV ) "PGX = (PGXINGV+RAN # PGXINGV) %Diff = "PGX PGXINGV+RAN $100 BOX C NH5.4-A-07207 DATA ACQUISITION The two panels on the left show the new data feeding that will be implemented at INGV 5 to 6 minuets are necessary to process an event shakemap. The procedure is fully automatic and based on the EarthWorm processing packages. As more information (data and source parameters) become available, updated shakamaps are generated automatically, Similarly, the “manual” procedure allows for off-line shakemap revision. Shakemap generation at INGV : The parametric PGM data are determined directly from the waveforms assembled for each event in SAC format after Backnet-Locator procedure. Shakemaps are determined automatically after manual revision of the hypcenter location by the H24 on duty personnel (max 30 minutes and on average 15 minutes after the event) using an “in- house”procedure. WORK IN PROGRESS (below): ON-LINE: In-house procedure, manually revised locations check_new_data.sh get_event_fromS4.sh chk_WW.sh check_shake_single.sh ShakeMap Backnet-Locator DATA WEB publication TESTING - THE IMPORTANCE OF REAL DATA ABSTRAC: Since 2005, the Italian Civil Protection (Dipartimento della Protezione Civile, DPC) has funded several projects driven toward fast assessment of ground motion shaking in Italy.- the final goal being that of organizing the emergency and direct the search and rescue (SAR) teams. Similarly, in mid 2006, the EU SAFER project aimed toward rapid assessment of ground motion shaking started. To this end, the Istituto Nazionale di Geofisica e Vulcanologia (INGV) has started to determine shakemaps using the USGS-ShakeMap® package within 30 minutes from event occurrence using a manually revised location. In this work, we present the INGV implementation of USGS-ShakeMap® for earthquakes occurring in Italy and immediately neighboring areas. Emphasis is put on adopted ground motion predictive relations, site correction for the amplification of the ground motion and on fast procedures for data processing and the in-progress developments to determine near real-time shakemaps. Finally, a test on the fundamental role of observed data and of seismic station coverage is shown.