FLOOD SUSCEPTIBILITY MAPPING USING RANDOM FORESTS AND BOOSTED TREES IN SEOUL, KOREA

摘要

Recently, the frequency of floods has increased due to climate change such as global warming; the damage highlighted in the flood hazard map is based on real surveyed data. Therefore, the purpose of this study is to create flood susceptibility maps of the Seoul metropolitan area using Random forest and Boosted tree models in a geographic information system (GIS) environment. Topographic, geology, soil and land use data were used to build the spatial database. Inundation areas in 2010 were used to train the models and inundation areas in 2011 was used for validation. Flood susceptibility maps were created using Random forest and Boosted tree models for the regression algorithm. Finally, the flood susceptibility maps were validated using the flooded area in 2011, for prediction rate. The random-forest and boosted-tree models showed validation accuracies of 78.78 % and 77.55 for the regression algorithm, respectively. This information and the maps generated from it could be applied to flood prevention and management. In addition, the susceptibility map provides meaningful information for decision-makers regarding priority areas for implementing flood mitigation policies.