Erika Schiappapietra

Traditional seismic hazard and risk analysis tools usually employ ground motion prediction equations (GMPEs) to estimate the earthquake ground motion at a given site. The earthquake ground motion of interest to engineering is often the transient ground shaking that occurs during an earthquake. This ground motion is invariably evaluated in terms of one or more scalar intensity measures (IMs), such as the maximum ground acceleration occurring during the earthquake (known as the peak ground acceleration), the peak ground velocity and the peak ground displacement. The ground motion is also often expressed in terms of response spectral acceleration, which represents the maximum response of a single degree-of-freedom system of a given oscillator period and damping subject to the ground motion time-history. GMPEs provide an estimate of the ground shaking by relating an IM to a set of parameters describing the earthquake source, such as the magnitude, the propagation path and local site conditions. Standard Probabilistic Seismic Hazard Assessment, which provides IM estimates at a single site, is now a well-established tool.

The quantification of the seismic performance of spatially-distributed infrastructures, such as bridges and lifelines, and prediction of earthquake-induced damage over a region requires not only the estimation of independent IMs values at different sites, but also the knowledge of the joint probability of occurrence of such IMs at multiple locations during the same earthquake. Indeed, stakeholders, such as government, search-and-rescue organizations and private companies, require a reliable evaluation of the ground-motion field to support decision making for civil protection emergency planning as well as long-term and rapid loss and risk assessment. Therefore, understanding the spatial characteristics of the ground motion is a fundamental prerequisite. Although the number of earthquake recordings has increased dramatically over the last decades, earthquake ground motions in epicentral areas are often recorded at only a handful of seismometers separated by many kilometres. Therefore, knowledge of how the ground motion varies spatially is required to predict ground-motion IMs at unobserved locations to generate shaking scenarios. The importance of defining spatially-correlated ground-motion fields has also been demonstrated in other contexts, such as loss estimation. Different assumptions can lead to significantly different outcomes on the predicted ground motions, which in turn affect loss estimates. Numerous studies have shown that neglecting the spatial correlation may cause a bias in loss estimates, overestimating the most likely losses and underestimating rare losses. On the contrary, overestimating the correlation may lead to the opposite result.

Several spatial correlation models have been published since the 1990s. Although it is widely accepted that the correlation decreases with increasing separation distance, significant discrepancies in terms of the decay rate (characterised by the range – the higher the range the slower the decay rate) exist, leading to large uncertainties in the assessed risk. Further analyses are thus required to draw firm conclusions on the spatial correlation. To address these issues, we use large databases of recorded strong ground motion from previous earthquakes and ground-motion simulations. We find that: (1) magnitude and range do not show any relationship (we believe that other source effects should be accounted for to predict the range for a specific earthquake); (2) there is a positive correlation between the range and the response spectral period; (3) the range is also regionally-dependent, so that region-specific spatial correlation models should be derived. Our findings can be used to advance understanding of the spatial correlation of ground motion. The reader is referred to Schiappapietra and Douglas (Earth-Science Reviews, 2020) for further details.