International Journal of Advances in Computer Science and Its Applications
Author(s) : ADARSH S, JANGA REDDY M
This paper presents the multiscale spectral analysis of four water quality time series data from an Indian river. First, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is employed for multiscale decomposition and the resulted orthogonal modes namely Intrinsic Mode Functions (IMFs) are subsequently subjected to the Normalized Hilbert Transform (NHT). The spectral representation clearly depicts the nonlinearity and nonstationarity of the datasets and the time varying behavior of dominant frequency. The marginal Hilbert spectrum of different parameters shows that the dominant frequency of most of the pollutants is at high frequency range which indicates the significant anthropogenic impacts in the study area. Also the trend analysis performed upon the instantaneous amplitudes show that the high frequency components are responsible for overall trend of the four time series during the study period under consideration. The multiscale decomposition process and the results of spectral analysis may improve the modeling efforts on the river system.