Browse > Article

A New Model for Forecasting Inundation Damage within Watersheds - An Artificial Neural Network Approach  

Chung, Kyung-Jin (하와이대학교 공과대학 토목환경공학과)
Chen, Huaiqun (하와이대학교 대학원 토목환경공학과)
Kim, Albert S. (하와이대학교 공과대학 토목환경공학과)
Publication Information
Abstract
This paper presents the use of an Artificial Neural Network (ANN) as a viable means of forecasting Inundation Damage Area (IDA) in many watersheds. In order to develop the forecasting model with various environmental factors, we selected 108 watershed areas in South Korea and collected 49 damage data sets from 1990 to 2000, of which each set is composed of 27 parameters including the IDA, rainfall amount, and land use. After successful training processes of the ANN, a good agreement (R=0.92) is obtained (under present conditions) between the measured values of the IDA and those predicted by the developed ANN using the remaining 26 data sets as input parameters. The results indicate that the inundation damage is affected by not only meteorological information such as the rainfall amount, but also various environmental characteristics of the watersheds. So, the ANN proves its present ability to predict the IDA caused by an event of complex factors in a specific watershed area using accumulated temporal-spatial information, and it also shows a potential capability to handle complex non-linear dynamic phenomena of environmental changes. In this light, the ANN can be further harnessed to estimate the importance of certain input parameters to an output (e.g., the IDA in this study), quantify the significance of parameters involved in pre-existing models, and contribute to the presumption, selection, and calibration of input parameters of conventional models.
Keywords
inundation damage; artificial neural network; forecasting model; temporal-spatial information; significance of parameters;
Citations & Related Records
연도 인용수 순위
  • Reference