Author(s): M. C. Peel; T. A. Mcmahon; R. Srikanthan; K. S. Tan
Linked Author(s):
Keywords: Ensemble Empirical Mode Decomposition; Time series analysis; Spectral analysis; Stochastic data generation; Rational spline EMD; Southern Oscillation Index
Abstract: Ensemble Empirical Mode Decomposition (EEMD), a recently developed improvement over traditional Empirical Mode Decomposition (EMD), utilises the concept of noise assisted data analysis to decompose a time series into Intrinsic Mode Functions (IMFs) and a residual (or trend). Because the EEMD algorithm is locally adaptive it is robust when applied to non-stationary and non-linear data and is suitable for decomposing hydroclimatic time series that appear non-stationary in terms of mean and variance. Using synthetic time series constructed with fluctuations of known frequency and a range of trends (linear, ramp, step and parabolic), we report results from our investigation into the ability of EEMD to decompose the synthetic time series and reproduce the embedded known components. We discuss objective automation of the EEMD algorithm based on the synthetic time series analysis. EEMD is then applied to Southern Oscillation Index time series to demonstrate the potential for using EEMD for non-stationary stochastic data generation.
Year: 2011