From Near-Real-Time Information Processing to Near-Real-Time Decision Making in Risk Management of Natural Hazards

Authors

Kazuyoshi Nishijima, Mathias Graf and Michael H. Faber

Abstract

A framework for near-real-time decision making concerning risk management of natural hazards is presented taking basis in a conditional probability representation for probabilistic models, e.g. Bayesian probabilistic networks or regression models. Two decision situations are considered; one-time decision making and sequential decision making. In the former case decision makers are required to make decisions only once for terminal actions, whereas in the latter case decision makers can postpone the decisions for terminal actions and optionally collect more information before making the terminal decisions. For sequential decision making the sequential theory is employed. The presented framework can be facilitated in decision situations where the information becomes available in near-real-time and decision makers need to make decisions also in near-real-time. Two examples are shown to illustrate how the present framework is implemented into practice considering a decision situation in the emergence of a typhoon.

Published in/by

Proceedings EM08, Inaugural International Mechanics Institute Conference, Minneapolis, USA, May 18-21, 2008.