Wang, HS; Lin, H; Chen, JS; Chen, FL (2012). Study on the relationship between sub-pixel percentage cover and multi-temporal NDVI. INTERNATIONAL JOURNAL OF REMOTE SENSING, 33(17), 5615-5628.
Normalized difference vegetation index (NDVI) time series have the potential to delineate vegetation phenology change. For a mixed pixel, time-series fluctuations representing phenological change in a year indicate percentage cover changes. In this study, noise in the Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI time series was removed by a Savitzky-Golay filter and wavelet filter. Principal component analysis was conducted to generate new variables characterizing the variations of percentage cover. The results show that the first principal component (PC1) is not sensitive to filters and that PC1 and percentage forestland cover are significantly correlated, with a coefficient of determination (R-2) greater than 0.55, while PC1 and percentage cropland cover are uncorrelated, with R-2 < 0.01. This study also indicates that MODIS percentage forestland cover can be roughly estimated by multi-temporal NDVI data in cloud-prone areas and that the spatial features of mixed pixels can be characterized by temporal phenological features.