Seismic risk has social and physical aspects: although very different in nature, they are intertwined and share striking similarities. Seismic risk assessment involves models and data, and judgments of the potential impact of earthquakes, which must be weighed by the probabilities of occurrence. Models for physical and social impacts all involve considerations of resilience: material in one case, psychological and social in the other. For purposes of risk assessment, statistical models that capture the occurrence of major earthquakes over the long term suffice -- typically different models for different types of sources of seismicity --, and may be the only ones that are practicable. Data include historical and pre-historical seismicity, mechanical properties of structures, and the social conditions, including preparedness, and the responses of affected populations. We argue that social vulnerability plays a central role in this context, and review the risk factors that influence social vulnerability. We focus on the risk of post-traumatic stress syndrome induced by seismic events, and discuss its mitigation. After reviewing approaches and methods for the measurement of social impact, we suggest that automatic content analysis of electronic texts are a new, potentially revolutionary source of indications about mental states induced by traumatic events, complementing or even replacing traditional surveys, and offering the ability to conduct such analysis in near real-time. We illustrate the assessment of the probability of large earthquakes generated by recurrent faults using the Brownian Passage Time model, and illustrate how censored observations of waiting times may be used, alongside uncensored ones, to calibrate this model and make predictions for various future time horizons.
RISK Models and Applications, Selected Papers, Berlin 2011, CODATA-Germany
seismic risk, earthquakes, seismicity, Brownian Passage Time, resilience, vulnerability, text mining, censored observations, post-traumatic stress syndrome.