Stochastic Design of Early Warning Systems
Early Warning Systems (EWS) are designed to avoid, or at least to minimize the impactimposed by a threat. Identifying information patterns generated by the different information sources isfundamental for predicting a catastrophic event, and thus for defining realistic warning levels. SinceEWS are time sensitive or stochastic, it becomes necessary to develop a methodology that integratesthe different monitoring information sources and that takes into account all possible performingscenarios of the system. This paper discusses the development of a Stochastic Design of an EarlyWarning System (SDEWS) introducing a risk measure as the reference variable which enables theintegration of the different effects captured by the monitoring instruments. In this way, the riskmeasure serves as a rational index for the definition of warning thresholds, but also introduces EWSwithin a decision-making framework. For this purpose, a Bayesian approach is proposed as a suitabletool for integrating and updating the joint states of information, for updating the warning level of thesystem, and for facilitating the decision-making required for the issuing of the warnings. Some of themethods proposed for implementing the SDEWS are also discussed (Bayesian smoothing, Bayesianfiltering and Bayesian networks).
Z.Medina-Cetina, F.Nadim F.Nadim
International Centre for Geohazards / Norwegian Geotechnical Institute, Oslo Norway
国际会议
上海
英文
2007-10-18（万方平台首次上网日期，不代表论文的发表时间）