Volume XLII-4/W18
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 923–929, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-923-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-4/W18, 923–929, 2019
https://doi.org/10.5194/isprs-archives-XLII-4-W18-923-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  19 Oct 2019

19 Oct 2019

ANALYSIS OF THE PRECIPITATION CLIMATE SIGNAL USING EMPIRICAL MODE DECOMPOSITION (EMD) OVER THE CASPIAN CATCHMENT AREA

F. Sabzehee1, V. Nafisi1, S. Iran Pour2, and B. D. Vishwakarma3 F. Sabzehee et al.
  • 1Geomatics Engineering, Faculty of Civil Engineering and Transportation, University of Isfahan, Isfahan, Iran
  • 2Institute of Geodesy, University of Stuttgart, Stuttgart, Germany
  • 3School of Geographical Sciences, University of Bristol, University Road, Bristol BS8 1SS, UK

Keywords: Empirical Mode Decomposition, Precipitation, Intrinsic mode functions, Trend, Hilbert Transform, Frequency

Abstract. In this paper, we employ Empirical Mode Decomposition (EMD) together with Hilbert Transform to analyze precipitation time series over the Caspian Sea catchment. Several studies have shown that EMD can extract nonlinear and non-stationary signals better than Fast Fourier Transform (FFT) and Wavelet Transform. EMD decomposes the time series into a finite number of Intrinsic Mode Functions (IMFs) in the time-frequency domain, while FFT helps us operate either in the time or the frequency domain, which fuels limitations such as the inability of nonstationary signal processing and the lack of time transparency. Although Wavelet Transform is shown to be better than FFT, it fails to detect the instantaneous frequencies and needs to have prior information about characteristics of the data. On the other hand, EMD has shown that it is almost able to determine the signal characteristics with no previous assumptions to estimate the instantaneous frequencies of the signal. In this work, EMD is applied to identify the main frequencies of precipitation time series. Thereafter, a statistical procedure is used to identify the prominent IMF of the original signal.

We use the correlation coefficient, Minkowski distance and variance test to extract the relevant and prominent IMFs. The results show that IMF 1–3 are the relevant components and are related to annual and biennial variations of precipitation time series over the Caspian catchment during 2003–2016, respectively.