Multihazard Risk Assessment from Qualitative Methods to Bayesian Networks: Reviewing Recent Contributions and Exploring New Perspectives

Abstract

Natural processes are interacting components of natural systems. Under certain circumstances, they can be transformed into threats for humanity, environment, and development. Examples such as the 2006 Pangandaran earthquake–tsunami and the 2011 Tohoku earthquake–tsunami–flood–nuclear catastrophe point out the necessity for an integrated multihazard risk assessment tool. This paper presents the critical steps and improvements in approaches to multihazard risk management. From the first qualitative, semiquantitative techniques with which risk is calculated through individual processes to more powerful techniques which try to capture and evaluate the interactions (trigger, cascade effect) among the natural hazards, such as Event Tree (ET) and Bayesian Networks (BNs). Especially Bayesian Networks and recently, their extensions as Dynamic Bayesian Networks (DBNs) and Hybrid Bayesian Networks (HBNs) offer a great opportunity for a more realistic and flexible multihazard risk assessment.