Exploring the impact of spatial correlations and uncertainties for portfolio analysis in probabilistic seismic loss estimation

Abstract

The significant potential for human and economic losses arising from earthquakes affecting urban infrastructure has been demonstrated by many recent events such as, for example, L' Aquila (2009), Christchurch (2011) and Tohoku (2012. Within the current practice of seismic loss estimation in both academic and industry models, the modelling of spatial variability of the earthquake ground motion input across a region, and its corresponding influence upon portfolios of heterogeneous building types, may be oversimplified. In particular, the correlation properties that are well-known in observations of ground motion intensity measures (IMs) may not always be fully represented within the probabilistic modelling of seismic loss. Using a case study based on the Tuscany region of Italy, the impacts of including spatially cross-correlated random fields of different ground motion IMs are appraised at varying spatial resolutions. This case study illustrates the impact on the resulting seismic loss when considering synthetic aggregated portfolios over different spatial scales. Inclusion of spatial cross-correlation of IMs into the seismic risk analysis may often result in the likelihood of observing larger (and in certain cases smaller) losses for a portfolio distributed over a typical city scale, when compared against simulations in which the cross-correlation is neglected. It can also be seen that the degree to which the spatial correlations and cross-correlations can impact upon the loss estimates is sensitive to the conditions of the portfolio, particularly with respect to the spatial scale, the engineering properties of the different building types within the portfolio and the heterogeneity of the portfolio with respect to the types.

Earthquake Risk Assessment: Present Shortcomings and Future Directions

“…16.8) but that when both spatial correlation and cross-correlation are accounted for, the impact on the losses at low probabilities of exceedance can be significant. However, it is noted that the portfolio selected by Weatherill et al (2013) was highly heterogeneous and included building types with a very wide range of periods of vibration; should the portfolio be more clustered around a smaller range of periods of vibration then the impact of the inclusion or not of spatial correlation (without cross correlation) will have a significant effect on the resulting losses, as has been shown in other studies (e.g. Crowley et al 2008).…”

“…The way in which the ground-motion aleatory variability is spatially modelled affects the standard deviation of the loss; neglecting site-to-site ground-motion correlation leads to systematically underestimation of large, rare losses and overestimation of smaller but frequent ones (see e.g. Crowley and Bommer 2006;Weatherill et al 2013). Monte Carlo simulation is generally employed to simulate the seismicity of the next one hundred thousand years or so (see e.g.…”

“…When simulating spatial distributions of ground motion for loss assessment, if cross-correlation, for example between the spectral acceleration at 0.3 s (used for the vulnerability function of a low rise building type) and that at 1.0 s (used for a mid rise building type), is not modelled, and each ground-motion field is simulated independently, the impact of the spatial correlation is eroded when the combined damage/loss to both building types is estimated. Weatherill et al (2013) show that the impact of spatial correlation on the total loss to a heterogeneous portfolio is minimal when cross-correlation is not modelled (Fig. 16.8) but that when both spatial correlation and cross-correlation are accounted for, the impact on the losses at low probabilities of exceedance can be significant.…”

“…Comparison of spatial correlation (blue curve) and spatial cross-correlated losses (green and red curves) on the total loss to a heterogeneous portfolio of losses (Weatherill et al 2013) The use of instrumental intensity measures in vulnerability modelling is required when analytical modelling of the response of structures is employed. In this case the explicit nonlinear behaviour of structures of a given class under accelerograms with differing characteristics is evaluated.…”

Editorial for the VEESD 2013 special issue

“…It comprises 13 contributions, which were selected on the basis of the response and feedback of both mini-symposium organizers and participants. The papers of this special issue provide new insight into the state-of-the-art of different fields in earthquake engineering based on experimental, analytical, and numerical studies as well as field observations after earthquakes, as it becomes obvious from the subsequent short summary.The first paper by Weatherill et al (2015) illustrates the effect of the spatial variability of earthquake ground motions upon the resulting seismic loss of aggregated portfolios of heterogeneous building typologies. Iervolino et al (2015) present a model on life-cycle assessment considering the effect of damaging aftershocks.…”