A Computational Agent Model of Flood Management Strategies

A Computational Agent Model of Flood Management Strategies

Lisa Brouwers (The Royal Institute of Technology, School of ICT, Sweden) and Magnus Boman (The Royal Institute of Technology, School of ICT, Sweden & The Swedish Institute of Computer Science (SICS), Sweden)
DOI: 10.4018/978-1-4666-0882-5.ch307
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

A geographically explicit flood simulation model was designed and implemented as a tool for policy making support, illustrated here with two simple flood management strategies pertaining to the Upper Tisza area in Hungary. The model integrates aspects of the geographical, hydrological, economical, land use, and social context. The perspectives of different stakeholders are represented as agents that make decisions on whether or not to buy flood insurance. The authors demonstrate that agent-based models can be important for policy issues in general, and for sustainable development policy issues in particular, by aiding stakeholder communication and learning, thereby increasing the chances of reaching robust decisions. The agent-based approach enables the highlighting and communication of distributional effects of policy changes at the micro-level, as illustrated by several graphical representations of outputs from the model.
Chapter Preview
Top

Background

While the agent area started off with artificial intelligence assistants, service robots, and many other forms of one-agent systems (Agre & Rosenschein, 1996), the potential for societal applications naturally reside with multi-agent systems (O’Sullivan & Haklay, 2000; Gimblett, 2002). In geography, physical components of a complex system—vegetation, fauna, soils, and climate/hydrology—are usually separated from socio-economic ones, like demography, culture, economy, and policy (Reenberg, 2001). An agricultural management strategy then advices and implements a policy by carefully considering those two sets of components separately. An agent-based model like our own, by contrast, lets the stakeholders be represented by agents taking actions in various states of information and in various geographical spaces, in situations where agents are informed of all components to very different extents. Heterogeneity in the agent population is thus an important feature (Wooldridge, 2002).

In microsimulation, the population is usually homogeneous. Early models (Orcutt 1957; Orcutt et al., 1961; Hägerstrand 1975) purported to address the shortcomings of the macro models then reigning in economics, demography, and geography (Merz, 1991), a promise delivered on as computing power became available with which to execute models in scenario analysis fashion (Caldwell & Keister, 1996). The development in computer hardware and software has naturally also benefited agent-based simulations, making them a widely used complement to mathematical theorizing (Axtell, 2000). When to go for agent heterogeneity and when the simpler microsimulation models suffice is a question part of a methodological scientific development still in rapid progress (Boman & Holm, 2004), but agent-based social simulations have proved increasingly important to policy making (Polhill, Gotts & Law, 2001).

Complete Chapter List

Search this Book:
Reset