Spatial modelling of natural disaster risk reduction policies with Markov decision processes

Abstract

The 2010/2011 floods in Queensland, Australia inflicted significant damages to government's critical infrastructures, private properties and businesses reaching an estimated amount of AU$16 billion. Mitigating the devastating effects of floods to the community and critical infrastructures entails competing financial requirements at the different levels of government. Hence, the main objective of this study was to examine the financial optimality of disaster risk reduction measures by integrating Markov decision processes (MDP for short) with geographic information system (GIS). Conducted in the core suburbs of Brisbane City, we organised the MDP variables using the following: 1) flood risk levels as the states of the urban system; 2) Queensland's disaster risk reduction measures as the action variables; 3) percentage of government expenditures by disaster risk reduction category as the state transition probabilities; 4) total lost earnings to businesses affected by the flood events as the reward variables; and 5) the weighted average riskless rate of return, the weighted average rate of return, and rate of return for a riskier asset as discounting factors. We analysed 36 MDP scenarios at four-level iteration and then calculated the expectimax values to find the optimal policy. The results from the analyses revealed that the Commonwealth government optimised the use of its natural disaster risk reduction expenditures to recovery while the State government focused on mitigation. When both government expenditures combined, the mitigation measure was identified as the optimum natural disaster risk reduction policy. The methodology presented in this study allowed a spatial representation and computationally feasible integration of complex flood disaster risk model with government expenditures and business earnings. The insights from this integrated approach emphasise the viability of finding optimum expenditures, and re-examine if necessary, in implementing natural disaster risk reduction policies and climate adaptation strategies.