Reconstruction of Temporal NDVI Dataset:Evaluation and Case Study

Abstract

NDVI time series datasets have been successfully applied to global and regional environmental change research,vegetation dynamic,land cover change and extraction of surface biophysical parameters.Due to the limitation of atmospheric conditions and sensors,although temporal NDVI dataset has been preprocessed rigorously,there is still so much noise within the NDVI time series datasets which would affect further analysis and applications.This study introduced six common methods for reconstructing high-quality NDVI time series dataset in recent years:Mean-value iteration filter,the modified best index slope extraction,Fourier Transform,Savitzky-Golay filter,Asymmetric Gaussian function fitting and Harmonic analysis of time series.Then,the six methods mentioned above were applied to reconstruct the SPOT VEGETATION 10-days-maximum-NDVI composite data in 2007 and 2008 in Zhangye.The reconstructing results are compared and evaluated separately.Finally,a temporal profile with artificial noise was also reconstructed with the six methods respectivly in order to evaluating the merits and demerits of every method.