Spatial prediction of flood susceptibility using random-forest and boosted-tree models in Seoul metropolitan city, Korea

TLDR

Flood frequency is rising due to climate change, making flood‑risk maps based on actual flooded area data increasingly important. This study develops flood‑susceptibility maps for the Seoul metropolitan area using random‑forest and boosted‑tree models within a GIS environment. The authors compiled flooded‑area, topography, geology, soil, and land‑use datasets, derived 12 spatial factors, trained the models on 2010 flooded area and validated with 2011 data, and assessed factor importance. Distance from river, geology, and DEM were the most influential factors; the random‑forest model achieved 78.78 % (regression) and 79.18 % (classification) accuracy, while the boosted‑tree model achieved 77.55 % and 77.26 %, providing useful maps for prioritizing flood‑mitigation decisions.

Abstract

Since flood frequency increases with the impact of climate change, the damage that is emphasized on flood-risk maps is based on actual flooded area data; therefore, flood-susceptibility maps for the Seoul metropolitan area, for which random-forest and boosted-tree models are used in a geographic information system (GIS) environment, are created for this study. For the flood-susceptibility mapping, flooded-area, topography, geology, soil and land-use datasets were collected and entered into spatial datasets. From the spatial datasets, 12 factors were calculated and extracted as the input data for the models. The flooded area of 2010 was used to train the model, and the flooded area of 2011 was used for the validation. The importance of the factors of the flood-susceptibility maps was calculated and lastly, the maps were validated. As a result, the distance from the river, geology and digital elevation model showed a high importance among the factors. The random-forest model showed validation accuracies of 78.78% and 79.18% for the regression and classification algorithms, respectively, and boosted-tree model showed validation accuracies of 77.55% and 77.26% for the regression and classification algorithms, respectively. The flood-susceptibility maps provide meaningful information for decision-makers regarding the identification of priority areas for flood-mitigation management.