Author(s)

K. A. Wojciechowska

Abstract

In 1995, because of danger of flooding, a massive evacuation in the province of Gelderland in the Netherlands took place. The process of the evacuation went well, however, the life-threatening flood did not occur. The decision was based on deterministic information, experience and expertise. A robust technique, where uncertainties (e.g. in water level forecast) are explicitly incorporated, was not used (and is still not used). There are several methods in the decision theory, which allow the inclusion of uncertainty in decision-making in an explicit way. This paper gives a description of such methods in the context of operational flood management. We focus on decision trees, decision influence diagrams and Markov Decision Processes. In the context of operational flood management, the methods usually require specification of conditional flooding probability, i.e. a flooding probability given some (uncertain) information. We present application of such methods to operational decisions like evacuation or activation of an emergency storage area. We compare the methods with respect to ease of understanding and we briefly discuss whether there is a chance to apply the methods in real life situations. Keywords: decision-making, decision trees, decision influence diagrams, Markov Decision Processes, uncertainty, conditional flooding probability. 1 Introduction The Netherlands is a country located in the delta of three European great rivers: the Rhine, the Meuse and the Scheldt. Over 60% of the country is situated below mean sea level or is vulnerable to flooding by sea or by rivers. Natural barriers like dunes and high grounds, and man-made constructions like dikes and storm surge barriers protect the area from flooding.

Keywords

decision-making, decision trees, decision influence diagrams, Markov Decision Processes, uncertainty, conditional flooding probability