Comparison of two fitting methods of NDVI time series datasets

The fitting of NDVI time series datasets is to minimize the effects of anomalous values caused by atmospheric haze and cloud contamination, and to reconstruct high-quality NDVI data for parameters inversion and information extraction. A comparative study on two well-known fitting approaches of NDVI time series data, namely, Savitzky-Golay filtering and asymmetric Gaussian function fitting, was described in China using both direct and indirect comparison methods. The results showed that these two methods generally presented a similar performance and a high agreement in data filtering in China, but regional differences existed between them. The main reason for their discrepancies was related to the methodological differences between these two methods, as well as the landscape heterogeneity in different regions. This comparative study can help users to understand the advantages and limitations of each fitting method, and choose the appropriate method to reduce the errors and improve the accuracy in their applications.