Sample Entropy‐based Adaptive Wavelet De‐noising Approach for Meteorologic and Hydrologic Time Series
Dong Wang,Vijay P. Singh,Xiaosan Shang,Hao Ding,Jichun Wu,Lachun Wang,Xinqing Zou,Yuanfang Chen,Xi Chen,Shicheng Wang,Zhenlong Wang
DOI: https://doi.org/10.1002/2014jd021869
2014-01-01
Abstract:De‐noising meteorologic and hydrologic time series is important to improve the accuracy and reliability of extraction, analysis, simulation, and forecasting. A hybrid approach, combining sample entropy and wavelet de‐noising method, is developed to separate noise from original series and is named as AWDA‐SE (adaptive wavelet de‐noising approach using sample entropy). The AWDA‐SE approach adaptively determines the threshold for wavelet analysis. Two kinds of meteorologic and hydrologic data sets, synthetic data set and 3 representative field measured data sets (one is the annual rainfall data of Jinan station and the other two are annual streamflow series from two typical stations in China, Yingluoxia station on the Heihe River, which is little affected by human activities, and Lijin station on the Yellow River, which is greatly affected by human activities), are used to illustrate the approach. The AWDA‐SE approach is compared with three conventional de‐noising methods, including fixed‐form threshold algorithm, Stein unbiased risk estimation algorithm, and minimax algorithm. Results show that the AWDA‐SE approach separates effectively the signal and noise of the data sets and is found to be better than the conventional methods. Measures of assessment standards show that the developed approach can be employed to investigate noisy and short time series and can also be applied to other areas.