Spatio-Temporal Reconstruction of MODIS NDVI Data Sets Based on Data Assimilation Methods

Consistent Normalized Difference of Vegetation Index (NDVI) time series, as paramount and powerful tool, can be used to monitor ecological resources that are being altered by climate and human impacts, since its temporal evolution is strongly linked to changes in the state of land surface. However, the noise caused mainly by cloud contamination, heavy aerosol, atmospheric variability and signal of background soil and bi-directional effects impedes NDVI data from being further applied. In this work, data assimilation method for NDVI was proposed to reconstruct high-quality spatially and temporally continuous MODIS NDVI data. The historical MODIS NDVI data are used to generate the background field of NDVI based on a simple three-point smoothing technique, which can generally capture the annual feature of vegetation change. At every time step, the quality assurance (QA) flags in MODIS VI products were adopted to determine empirically the weight between background field and observation of NDVI. Additionally, the gradient inverse weighted (GIW) filter algorithm is adopted further to remove spatial discontinuity. Finally, the more reliable NDVI data can be generated. This method is implemented by the 16-Day L3 Global 1km SIN Grid NDVI data sets covered west China during 2003-2006. Results indicate that the newly developed method is easy and effective in reconstructing high-quality MODIS NDVI time series.