Journal
JOURNAL OF HYDROLOGY
Volume 296, Issue 1-4, Pages 98-117Publisher
ELSEVIER
DOI: 10.1016/j.jhydrol.2004.03.018
Keywords
regional flood frequency analysis; L-moments; Ward's cluster method; fuzzy c-means method; artificial neural networks
Ask authors/readers for more resources
A regionalised relationship to estimate flood magnitudes for ungauged and poorly gauged catchments can be established using regional flood frequency analysis. The geographical approach (Residuals method), Ward's cluster method, the Fuzzy c-means method and a Kohonen neural network were applied to 86 sites in the Gan River Basin of Jiangxi Province and the Ming River Basin of Fujian Province in the southeast of China to delineate homogeneous regions based on site characteristics. Similar groupings of sites into sub-regions were obtained from all but the Residuals method. Since the Kohonen neural network can be employed to identify the number of sub-regions as well as the allocation of sites to sub-regions, this method is to be preferred over Ward's method and the Fuzzy c-means approach. For each sub-region, growth curves must be constructed and the value of an index flood must be related to catchment characteristics. The regional L-moment algorithm may be used to advantage both to identify an appropriate underlying frequency distribution and to construct sub-regional growth curves. However, the membership levels produced by the Fuzzy c-means method may also be used as weights to derive a regional at-site growth curve from those of all the sub-regions. The latter method is likely to be most useful where the sub-regional growth curves are of strongly contrasting shape. An index flood may be related to catchment characteristics using Multiple Linear Regression Analysis, but application to the Gan-Ming data demonstrates that estimates with lower standard errors of estimate can be produced using an artificial neural network (ANN). However, in order to apply such ANNs, sufficient sites must be available so that enough processing elements can be employed without impairing the ability of the network to generalise outside the training data set. (C) 2004 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available