MULTIAGENT DECISION SUPPORT FOR FLOOD EMERGENCY MANAGEMENT Norita Md. Norwawi 1) , Ku Ruhana Ku Mahamud 2) , Safaai Md. Deris 3) 1) Universiti Utara Malaysia, School of Information Technology (rita@uum.edu.my ) 2) Universiti Utara Malaysia, School of InformationTechnology(ruhana@uum.edu.my ) 3) Universiti Teknologi Malaysia, Head of Software Engineering Department, Faculty of Computer Science and Information Science (safaai@fsksm.utm.my) Abstract An emergency situation is a complex and dynamic environment whereby the situation is always changing, uncertain and unpredictable. Consequences of an emergency or crisis whether natural or man-made disaster are loss of human lives and damages to property. In many instances, human responses to emergency are reactive in nature and in an ‘ad-hoc’ manner. The response operation also involved a lot of people from various level of jurisdiction and organization. Decision makers are under stress, have to response urgently without considering all possible alternatives due to the time limitation and incomplete information, at the same time preserving reliability, consistency and coherency in their decisions and coordinated activities. An assistant technology rooted from an intelligent agent based system is an appropriate approach in guiding and assisting decision making processes of emergency managers especially during an emergency response operation. The nature of agents behavior which are autonomous, cooperative and proactive makes them a suitable method for designing and deployment of a dynamic and complex system. This paper presents a conceptual multi-agent architecture in the area of active decision support systems for flood emergency management. This model is aimed at providing an autonomous monitoring system that will assess river and reservoir water level based on an online hydrological data. If the system detects a potential flood occurring, a self-triggering mechanism will form an agent task force that will assist the response operations. The lifetime of this team of agents ended when floods are over. An ontology on flood warning has also been developed using Protege2000 ontology development tools. This ontology was reused in the multi-agent framework developed. 1. Introduction Loss of lives and damages to properties are the two main concern in a crisis. In this paper, crisis refers to natural disaster such as floods, earthquakes, typhoon or man-made disaster such as industrial accidents or terrorist attack. The terms crisis, disaster and emergency will be used interchangeably in this paper which refers to the state of danger or calamity. Number of death, evacuation operations, emergency centers, roads closed, food supplies are among the usual subject of communications in an emergency. Normally, the population will react on the onset of a disaster and response in an ad-hoc manner. Emergency situation is a dynamic environment where events are always changing with a lot of uncertainties on what is/will be happening. The situation is unpredictable, full of stress and at the same time requires an urgent and rapid response in order to save lives and reduce further damages to properties. Emergency response operation also involves many people from various jurisdictional level and organizations from local, state and federal government including voluntary organizations. It is a complex system where the complexity arises from four factors: dynamism, interaction among many parts, uncertainty and risk. [1] 2. Emergency Environment and Complex Adaptive Systems Comfort [2],[3],[4],[5],[6],[7],[8] has done extensive studies in crisis management in particular natural disaster like earthquakes from a complex adaptive systems (CAS) perspectives. CAS contains unique properties such as non- linearity, self-organizing and emergent behavior. Table 1 below shows the characteristics of complex adaptive systems.