Spatial correlation of earthquake ground motion and spatial variability of geological characteristics in seismic hazard and risk assessment
Final Report Abstract
Knowing probability distribution of the portfolio loss is necessary for a seismic risk assessment and risk management, because probability of extremely negative outcomes of an earthquake must be estimated. A sufficiently smooth and accurate estimate of the loss probability function is of great importance. The ground-motion variability, i.e. value of within-earthquake standard deviation and within-earthquake correlation, plays a significant role in estimations of loss distribution for extended objects (portfolio, lifelines). The correlation of ground-motion residuals reflects a non-random component in the residuals, which is caused by factors not accounted for by the ground-motion model. In this project we analyzed the within-earthquake correlation of ground motion using two datasets. The first dataset contains strong-motion records (more than 7000 records obtained from 54 earthquakes with moment magnitudes MW > 5) accumulated by the TSMIP (Taiwan Strong Motion Instrumentation Program) network in Taiwan during 1993-2009. The database, which contains shear-wave velocity data averaged for the top 30 meters of the soil column (Vs30) for the TSMIP stations, has been also collected. Two ground-motion prediction equations, which were recently developed for peak ground acceleration (PGA) in the region and based on moment MW and local ML magnitude and hypocentral distance, were used for the calculation and analysis of ground-motion residuals. The second dataset includes strong-motion records (more than 23700 records obtained from 84 earthquakes, moment magnitudes MW from 4.2 to 7.4) accumulated by the K-NET and KiK-net networks in Japan during 1999-2011. For the K-NET and KiK-net stations, the shear-wave velocity data up to 20 m and 30 m are available, respectively. Peak amplitudes of ground acceleration (PGA) and ground velocity (PGV) were considered for the Japanese records and two regional groundmotion models were used. We have shown that there is a prominent correspondence between within-earthquake correlation of ground-motion residuals and Vs30 values, which was estimated for an area characterized by a specific geological structure. The larger correlation between the Vs30 values corresponds to the larger correlation of PGA residuals; and the relationship is apparently linear. On the other hand, the level of within-earthquake correlation may vary significantly depending on site classes, general geological conditions and magnitude of earthquakes, records of which dominate in the analyzed dataset. The level of correlation increases with the decrease of average shear-wave velocity and with the increase of earthquake magnitude. The correlation is the smallest for areas with large spatial variations of thickness of deposits and the largest on thick sediments, which cover large areas. It has been found that application of empirical correction factors, which consider magnitude of earthquake, source-to-site distance and average shear-wave velocity for given station, and which can be developed for particular arrays, allow effectively reducing the level of within-earthquake correlation and the within-earthquake standard deviation. The proper choice of the correlation model is critical in estimations of portfolio loss distribution. Skipping the within-earthquake ground-motion variability and correlation leads to underestimation of loss. However, the impact of ground-motion variability and correlation on parameters of seismic loss distribution depends on the levels of damage, which in turn is a function of earthquake magnitude and location, ground-motion model and vulnerability characteristics of a portfolio. In this project we have shown that in practical calculation, when it is necessary to analyze loss distribution for a portfolio, a sensitivity analysis, which considers influence of possible variations in the characteristics of within-earthquake variability (standard deviation and correlation) on the parameters of the loss distribution, is an essential procedure. Estimations of seismic losses in seismically active regions, which are characterized by the shortage of the observed data, require appropriate within-earthquake correlation models developed for the other regions with large ground-motion databases. Selection of the correlation models, which allow reducing uncertainty in the choice and which result in realistic loss estimations, may be performed using the recently developed relationships between the correlation and gross geological and local soil characteristics, and earthquake magnitude. We demonstrated the relevance of geology-based models of spatial correlation for practical application and proposed a scheme for the selection of appropriate within-earthquake correlation models.
Publications
- “Correlated Ground Motion – Influence on Loss Estimates”, European Geosciences Union, General Assembly 2011, Vienna, Austria, 03 – 06 April 2011
Wenzel F. and V. Sokolov
- “Correlation of Strong Ground-Motion and Estimations of Seismic Hazard and Loss for Large Areas and Critical Elements of Networks”, 8th International Conference on Urban Earthquake Engineering, Tokyo, Japan, 7-8 March 2011
Sokolov V. and F. Wenzel
- “Ground-motion correlation and its influence on assessments of seismic hazard for extended objects”, Problems of Engineering Seismology, 2011, V. 38, № 4, 5-15 (in Russian)
Sokolov V. and F. Wenzel
- “Influence of Strong Ground-Motion Correlation on Estimations of Seismic Losses for Extended Objects: a Case of Scenario Earthquake”, 9th National Conference on Earthquake Engineering and Seismic Zonation, Sotchi, Russian Federation, 6-9 September, 2011
Sokolov V. and F. Wenzel
- “Probabilistic assessment of seismic risk for extended objects considering ground-motion correlation”, Earthquake Engineering, Safety of Structures, 2011, 5, 41-45 (in Russian)
Sokolov V. and F. Wenzel
- “Probabilistic Estimations of Seismic Hazard and Loss Considering Correlation of Strong Ground-Motion”, 9th National Conference on Earthquake Engineering and Seismic Zonation, Sotchi, Russian Federation, 6-9 September, 2011
Sokolov V. and F. Wenzel
- “On the influence of site conditions and earthquake magnitude on ground-motion within-earthquake correlation: analysis of PGA data from TSMIP (Taiwan) network”, Bulletin of Earthquake Engineering, 2012, 10 (5), 1401- 1429
Sokolov V., F. Wenzel, K.-L. Wen and W.-Y. Jean
(See online at https://doi.org/10.1007/s10518-012-9368-5) - “On the Modelling of Strong Ground Motion Distribution in Probabilistic Seismic Hazard Assessment and Loss Estimation”, European Seismological Commission 33-rd General Assembly, Moscow, Russia, August 19-24, 2012
Sokolov V. and F. Wenzel
