317 Urban Water-log Simulation and Prediction based on Multi-Agent Systems Shitai Bao 1 , Changjoo Kim 2 *, Wenping Ai 1 , Zehui Lai 1 , Jianfang Wang 1 1 Department of Geo-information, South China Agricultural University, Guangzhou 510642 2 Department of Geography, University of Cincinnati, OH 45221-0131 Telephone: (513-556-3424) Email: changjoo.kim@uc.edu *corresponding author Abstract Urban water-logging often occurs along with rapid urbanization and becomes a huge challenge on drainage emergency decision in many Chinese cities. Due to lack of the functionality for process modeling in terms of simulation in GIS, this paper proposed a method of agent-based water-logging simulation and prediction. Because of the fragmentation and space complexity on urban surface, urban water-logging is predicted based on simulating by multi-agent systems instead of traditional hydrological equations. The method can simulate uncertain flow and calculate water depth of any position at any moment, and seek optimal reservoirs for storing the excessive floods. Many factors of rainfall, flow, sink, interception, infiltration, gully outflow are considered, and multi- agents of water-log, drainage system and environment are utilized. Interaction rules among agents are defined. This approach is verified by historical data in DongHaoChong basin, central urban of Guangzhou city, and applied to forecast water-logging with a 100- year storm. The geocomputation process is supported by NetLogo tool system, an agent - based programming language and integrated modeling environment. The results are compared with those of a CA model developed in Matlab. The improvement of the methodology is discussed with respect to rules, parameters and data accuracy. The experiments demonstrated that the method is efficient for modeling the process of urban water-logging and drainage optimization to decrease huge loss of life and properties. Keywords: Urban water-logging; Surface water simulation; Agent-based model; Multi- agent; Netlogo system. 1. Introduction Agent-based modeling and simulation becomes increasingly popular in social sciences, due to its elegant and explicitly represent entities, environment, and relations between them (Gilbert, 2008). Although many agent-based models were built to simulate the complicated urban phenomena, few are done for urban water-log. In recent years, urban water-logging is very frequent in many cities that have long been a major concern in various areas throughout the world. Alongside rapid urbanization in People's Republic of