Task 3 - Checking and validation of the shakemap results and associated analysis


The aim of this task is to verify the accuracy and the robustness of the peak ground motion maps obtained through USGS-ShakeMap and to make comparisons with other methodologies aimed toward evaluation of the amount of ground motion shaking and of the macroseismic effects.
This task also entails the determination (or the refinement) of GMPEs for M<5.5 earthquakes regionally, the adjournment of the general GMPEs for larger earthquakes and the testing of different approaches for the site corrections.
This task is important because it attempts to answer some basic questions concerning the robustness and and the accuracy of the shakemaps as function of the input data, the adopted GMPEs and the role of the site effects. Specifically, the studies will address the following main questions:

  • To what extent the addition of new data affects the resulting shakemaps ?
  • What is the robustness of the shakemaps when different GMPEs are used for the same set of observations ?
  • What is the level of variation of the shakemaps when different VS30 site classifications are used ?
  • Can new approaches for the local site corrections reproduce more faithfully the observed ground motion ?
  • Is it possible to make fair comparisons between the instrumentally derived shakemaps and the “did you feel” (macroseismic intensity) reports in Italy ?
  • How do the instrumentally derived shakemaps compare to simplified methods driven toward the calculation of intensities ?

Accuracy, robustness and comparisons

For what concerns the accuracy and robustness tests, it is envisaged the implementation of the new version of the USGS-ShakeMap package (3.3) that will include an analysis the uncertainties of the resulting shakemaps. For accuracy we address the problem of predicting the ground motion as faithful as possible to the true one and for robustness the ability to replicate the same shakemaps independently from the data available. In practice, we want to investigate the dependency of the shakemaps on the available data set and we are planning the adoption of the jackknife technique to assess the reliability of the results. In this test, different subsets of the original data set are randomly selected and the statistics will determined on the distribution on the differences between predicted and observed PGM values in the resulting maps. For the evaluation of the shakemap robustness (i.e., the PGM predicted values) resulting from adoption of different GMPEs, for each region the maps will be determined using the same selected data set and comparisons between the different outcomes will be made. A similar strategy will be employed to assess the reliability of the VS30 classification. The comparisons with the other damage assessment approaches (KF and macroseismic intensities) will be made more qualitatively since it may be difficult to transpose the results of these analysis to a common intensity scale

Site corrections

Within this task, we also plan to develop and test different approaches toward correction for the site effects. The first will follow a recent pilot study by Malagnini et al. (2007) in which they demonstrated the usefulness of the information on the absolute site terms of a seismic network in the everyday routine, from the fast determination of moment magnitudes, to more sophisticated applications in the USGS-ShakeMap package. The mentioned work will allow the production of site terms for every station that will be thrown into the regression procedure. Absolute site terms will represent averages computed on all available azimuth and incidence angles, and may be used, within the ShakeMap package, for the deconvolution of the site effect from the recorded ground motion. A suitable grouping of seismic stations into classes related to the geological/geophysical characteristics of the instrumented sites (NEHRP classification, for example) would allow the definition of a typical site response for each class. Moreover, since the peak ground motions are carried by a dominant frequency that is, in turn, a function of magnitude and distance to the source, we will perform a feasibility study, for a limited number of stations, and determine a new site classification that would no longer be a mere amplification (deamplification) factor. The new classification will be tested for the site correction of the ShakeMap predictions where seismic stations are not available. The second approach addresses directly the non-linearity of the soil response at increasing levels of ground motion. Theory and experience show that non-linear soil behaviour affects the natural frequency and the amplification factor of peak ground acceleration, both decreasing with the seismic motion amplitude (e.g. Lanzo & Silvestri, 1999). In a broad sense, the whole transfer function of the subsoil is significantly non-linear and dependent on the energy of the seismic radiation. A shakemap including site effects should reliably account for soil non-linearity for strong-motion events.

Ground motion prediction equations (GMPEs)

For earthquakes M>5.5, we will develop (or refine the existing) GMPEs for the different regions. In the current regional classification of the GMPEs adopted within the ShakeMap installation at INGV, the Italian territory is subdivided into six different regions. The aim of this study is to refine both spatially the regions (i.e., re-define the perimeter of the regions) and the attenuation relations to obtain maximum coherency between predictions and observations for the smaller earthquakes. The project NGA (Next Generation of Attenuation relations) funded by PEER (Pacific Earthquake Engineering Research Center; http://peer.berkeley.edu/) seeks the determination of more sophisticated GMPEs that include various parameters of the seismic source and of the propagation in addition to magnitude and distance (e.g. Abrahamson and Silva, 1997). The aim is to obtain more accurate simulations of the PGM and reduce the scatter between observations and predictions (Somerville et al., 1997). Possible additional parameters are the directivity, faulting style, the fault-top depth, accounting for the local site effects through Vs30 and some parameterization of the non-linear soil response (Abrahamson and Silva, 2007). The use of more accurate GMPEs is greatly needed within ShakeMap. Introduction of a few additional parameters such as those above, will allow to step from a purely cylindrical (or spherical) geometry of the spatial variation of the PGM to geometries that replicate more faithfully the actual faulting and propagation process occurring in earthquakes. In this project, we will develop and implement predictive equations of the response spectrum that include directivity and faulting style (e.g. Somerville, 2003; Bommer et al., 2003; Baker, 2007; Spudich and Chiou, 2006). During the first part of the project, we will assess the compatibility of the different relationships introduced so far within the Italian context. In the second phase, we will verify that the corrections to the standard cylindrical GMPEs are effective in reducing the misfit between predictions and observations. These tests will be carried out also on the data provided by K-net (http://www.k-net.bosai.go.jp/) which is now the richest data base of strong ground motion. The presumption being that the cylindrical attenuation relations are likely different in Japan when compared to those applicable in Italy but the source corrections we introduce are theoretically independent of the geographical region. Finally, since GMPE depend also on earthquake depth and in Italy earthquakes do occur deep in the Calabrian Arc, we will make an attempt to determine attenuation relations depending also on depth. However, we anticipate that, because a very small number of earthquakes have been recorded at depths larger than say 30-40 km, determination of meaningful relations will be hampered by the paucity of the data set. To partially circumvent this problem, we will include data from other regions worldwide assuming that propagation from Mantle depths is somewhat similar from region to region. In summary, this task is multifaceted and involves i.) essential ingredients toward generation of the shakemaps (GMPEs and site effects), ii.) results verification and iii.) comparison with other methodologies that seek fast determination of ground motion shaking. It is envisaged that through the project the relevant and sound information gained from the activities in i.) and ii.) will be progressively transferred to Task 2.

First Annual Meeting: 19-21 October 2009