Application of Artificial Neural Networks for Regional Flood Estimation in Australia: Formation of Regions Based on Catchment Attributes

摘要

Most of the traditional regional flood frequency analysis (RFFA) methods are basedrnon linear models. Artificial neural networks (ANNs) can be used to develop nonlinearrnmodels in RFFA. This paper uses data from 452 gauging stations from easternrnAustralia to identify the optimum regions in the ANN-based RFFA modeling inrnAustralia. From an independent testing, it has been found that that K-Means clusterrnanalysis generate the best performing regions in the catchment characteristics datarnspace with two regions. However, the best ANN-based RFFA model is achievedrnwhen all the data set of 452 catchments are combined together.