Forest classification based on MODIS time series and vegetation phenology

The distribution and phenologies of vegetation are largely associated with climate, terrain characteristics and human activities. Satellite data provide an opportunity to map vegetation and monitor its dynamics continuously. This paper describes the application of Moderate Resolution Imaging Spectroradiometer (MODIS) time series data for forest classification in Northeast China. Fourier analysis was used to identify forest types based on temporal changes in MODIS NDVI values. Firstly, Fourier transform was applied to 36 MODIS 10-day maximum NDVI composite images during 2002 in the study area. Then the amplitude and phase data from the first and second harmonics of the Fourier transform were analyzed in amplitude-phase space. It shows that the introduction of these characteristic phenology parameters extracted from this series of MODIS NDVI into feature space improves the separabilities of forest types. Finally, the mean NDVI, first- and second-order amplitude and phase were used to produce an unsupervised classification map of basic forest formations in the study area. The Fourier analysis approach shows promising for identifying vegetation types based on multi-temporal remotely sensed data.