Sreeram Reddy Kotha (ISTerre) : "SEISMIC GROUND-MOTION PREDICTION IN EUROPE: PHYSICS, STATISTICS, AND ENGINEERING PRACTICE"

Seismic hazard analysts perform seismic hazard assessment of a built environment to ascertain ground-motion levels that can probably be exceeded at least once in the buildings’ design lifetime. Earthquake engineers rely on these predictions to optimize building design for seismic resilience. Seismic risk analysts integrate this knowledge and convey, to policy-makers, insurance-reinsurance firms, disaster response networks, and society at large, a motive to mitigate the risk of a seismic catastrophe. A critical prior to this chain of decisions is the one of probabilistic ground-motion prediction; which is an engineering seismologist’s domain of assimilating expertise from civil engineers, seismologists, geophysicists, and statisticians. Here, I present the needs and deeds of an engineering seismologist in developing the Ground-Motion Models (GMM).

GMM is an empirical relation that seeks to explain the observed ground-motions at seismic stations from past earthquakes, and to predict ground-motions at new sites from future earthquakes. It is necessary that these predictive models are statistically well-constrained and physically well-behaved for a variety of prospective seismic sources, seismic wave propagation paths, and seismically shakable sites and structures. For a good predictive model, we prefer an enormous sample of high quality, uniformly processed, ground-motion data collected at several thousands of sites over a long history of seismic activity. To circumvent such impractical expectations, a traditional practice has been to liken the long-term sampling of ground-motion at every single site to the short-term sampling at several sites scattered across the globe. A decade ago, this so-called ‘ergodic assumption in ground-motion prediction’ addressed the GMMs’ ‘well-constrained’ requirement quite reasonably. In this decade however, the ‘well-behaved’ problem has aggravated with the exponential increase of data and the apparent increase in its natural randomness. The GMMs’ ergodic assumption governs its prediction accuracy and precision, which in turn has a profound impact on seismic hazard and risk assessments, and subsequent decisions. It has thus become essential to review the ergodic assumption and to resolve the apparent inflation of aleatory variability in ground-motion data.

I will present to you the progress we have aided so far: the engineering needs, ground-motion datasets, statistical analyses, patent and latent physical phenomenon, development and deployment of GMMs in hazard and risk assessments, and the open challenges in the entire process – with an intent to enhance cross-disciplinary interactions.