Reconstructing MODIS vegetation index time-series data of cropping area

As an excellent dataset which reflects the coverage of vegetation on earth surface continuously, MODIS vegetation index time series have already become an important data source in crop measurement by remote sensing. However, there are always some noises caused by atmosphere variability and sensor angle in MOD13 vegetation product. For this reason, the time series of MOD13 should be reconstructed before application. According to phenology and cropping system, time series were separated into different periods related to the growth process of crops. Then, the vegetation indices in each period were reconstructed based on asymmetric gaussian function. After all periods were reconstructed, the indices in overlapping range between two adjacent periods were optimized. The above two procedures were repeated a certain times to restore the indices affected by noises. The proposed method were applied to reconstruct the NDVI time series of cropping area lies between Tong zhou District, Beijing, and Baoding City, HeBei Province with MOD13 data acquired in 2005. The same data were reconstructed by two step Savitzky-Golay filter. The comparison between two results show that the noise in time series can be evaluated and restored accurately. Meantime, the low vegetation indices caused by double cropping system are reserved effectively. The whole reconstructed NDVI time series can indicate vegetation coverage accurately.