Phenologies from harmonics analysis of AVHRR NDVI time series

The parameters from the Fourier harmonics analysis based on discrete Fourier transform(DFT) algorithm concisely summarize the development signals of annual or biannual vegetations embedded in time-series of AVHRR NDVI data. In this paper, as the characteristics of agricultural vegetation dynamics is different from that of natural vegetation, the adjustment algorithm of NDVI time series for natural vegetation developed by Sellers et al.(1996) was modified. Then, the authors applied and evaluated the DFT using the adjusted 10 days composited Pathfinder NDVI data(1992) in the southern Hebei Province, China. The results show the modified algorithm can be used to reconstruct the NDVI time series either for agricultural vegetation or natural vegetation. The pollution of cloud and the abnormal values in NDVI time series can be successfully removed with the modified algorithm, the harmonics wave can reconstruct a new NDVI time series which providing a basis for linking the analysis results to basic vegetation types according to their characteristic phenologies. The mean NDVI indicated overall productivity, allowing the differentiation of unproductive, moderately productive, and highly productive areas. The amplitude of harmonics indicated the variability of productivity over the year. The phase of the first harmonics summarized the timing of green-up for annual natural vegetation. The first two phases of the third harmonics show the time of highly growth period of biannual agricultural vegetation. The point analysis provides the foundation for the regional analysis. The results from harmonics analysis of NDVI time series could be used to land cover classification and crop type identification.