Predicting Phenology Using Time Series Remote Sensing Data: Initial Results for the Indian Forests

Time series (2003 to 2007) MERIS Terrestrial Chlorophyll Index (MTCI) products were used to predict the phenology of different forest types in India. The MTCI data were corrected for noise using a temporal moving window filter and then a Fourierbased smoothing was applied without compromising annual phenological cycle. Finally, the phenological variables i.e. onset of greenness and end of senescence, were predicted through iterative search for each pixel using 1.5 years of Fourier smoothed data. Different forest types were extracted from a global land cover map (GLC 2000) and corresponding phenological variables were clipped. Finally, for each forest type, median of phenological variables was derived from four year results and then a spatial majority filter was applied to the 1◦ x 1◦ tiles covering complete India. This study presents the initial results derived for the evergreen, semi-evergreen, moist deciduous and dry deciduous forest in India.