EFFECT OF WAVELET-BASED DE-NOISING METHOD ON CHANGE POINT DETECTION PROCESS IN HYDROLOGICAL TIME SERIES

The amount of de-noising method effect upon change point detection process of hydrological time series was studied. Two different stations were studied; first case study was Vanyar station in Iran that the annual average of precipitation, stream flow, temperature and moisture time series of this station were used in this paper, and another case study was Tifton station in the United States of America that the annual average of precipitation, stream flow and NO3 concentration time series of it were studied. Data was considered as two types, at first the natural – noisy – form of them was there forestudied data was modified and reformed – de-noised – form of them was studied again. Wavelet-based de-noising method was used to make de-noised time-series data. Since both stations have had a known change during 30 recent years, no significant effect of this change has been identified in phenomenon time series and the differences have been cancelled out during the comparison through classic change point detection methods. The change point detection process shows a significant difference before and after of de-noising. The change point detection is strictly improved through applied wavelet-based de-noising on hydrological time series. The detected points of noisy data have variancewhile the results present that the detected points of de-noised data are more reliable. A comparison of applied methods shows that de-noising is most effective on the Wilcoxon sum rank test, Pettit test, two-phase regression test, and standard normal homogeneity test, respectively.