ANU Home | Search ANU
The Australian National University
Centre for Mathematics and its Applications (CMA)
CMA Research Reports
Printer Friendly Version of this Document

CMA Research Report


MRR02-003

Christopher Zoppou, Ole M. Nielsen and Lu Zhang

Regionalization of daily stream flow in Australia using wavelets and k-means analysis


Abstract:  This paper describes an attempt to identify catchments within Australia that have similar stream flow signatures. The stream flow from a catchment represents the integrated effect of the catchment's physical characteristics, such as topography, soil and vegetation to external impacts such as climate. Therefore, stream flow represents the response of a catchment to these variables and has been called the signature of the catchment in this paper.

Characteristic features of a catchment's signature are identified using the wavelet transform. Unlike other traditional frequency decomposition methods, the wavelet transformation represents a step by step decomposition of the data signal by time and scale. Therefore, a wavelet decomposition often reveals structures that are obscured in stream flow records.

Integration of the wavelet signals over time yields the wavelet energy contained at a particular time scale. Plotting the wavelet energy for all time scales produces the energy spectra distribution. The energy spectra distribution summarizes the variability present in the stream flow series at a range of time scales. These can be used to categorize the response of catchments by grouping catchments with similar wavelet energy spectra distributions.

Wavelet energy spectra distributions were obtained for 99 years of synthesized daily stream flow records at 286 rivers throughout Australia. These were grouped using k-means analysis where twelve distinct signatures have been identified. While more work is needed to establish the role of wavelets in the regionalization of catchments, our preliminary results suggest that classifying daily stream flow using the energy of the wavelet spectrum may be an effective method to classify the behaviour of catchments to climatic and physical catchment parameters.


AMS Classification:  62P12
Date:  15 February 2002

Download PDF   ( 3.55M )


Return to MRR02 contents