An Effective Compound Algorithm for Reconstructing MODIS NDVI Time Series Data and Its Validation Based on Ground Measurements

Abstract

In this study, a compound technique was developed using eight denoising techniques for reconstructing high-quality normalized difference vegetation index (NDVI) time series data. The new algorithm consists of two major procedures: 1) detecting noisy data according to variation in the modification rates of eight selected denoising techniques and 2) using the medians of the denoised values of the eight techniques to replace the noisy data. The eight techniques include the modified best index slope extraction (M-BISE) technique, the Savitzky–Golay (S–G) technique, the mean value iteration (MVI) filter, the asymmetric Gaussian (A-G) technique, the double logistic (D-L) technique, the changing-weight (CW) filter, the interpolation for data reconstruction (IDR) technique, and the Whittaker smoother (WS) technique. The technique was tested with moderate resolution imaging spectroradiometer (MODIS) NDVI time series data derived from MOD09GQ of the Heihe River Basin in China. In situ NDVI data were obtained during one nearly complete growing season for six land-use types in the study area. Analysis of the temporal and spatial characteristics of the reconstructed data revealed that the compound technique performs better than the other techniques. In addition, the lower root-mean-square error (RMSE) of the compound technique, which was calculated using ground measurements, demonstrated the improved performance of the new technique. The main advantage of the new technique is its ability to effectively denoise data and maintain fidelity such that it can be widely used for other NDVI time series data and for other study areas.