Exploring the interannual variability of vegetation activity using multitemporal AVHRR NDVI imagery

Interannual qualitative changes were found to be determined by the timing of phenological events. However, kappa coefficients indicated that the timing of phenological events exhibited relatively low year-to-year consistency. Results suggest that the methods currently used to compute AVHRR/NDVI seasonal phenological metrics may need to be modified and/or regionally calibrated. Considerable variation is probably associated with noise in the original AVHRR-NDVI dataset. Stratification by similar phenological groups could result in a better characterization of seasonal and phenological events. The changes were, however, less pronounced when NDVI values were compared among years, especially when the noise in the data was removed by principal component analysis. PCA loadings for each annual vegetation component had almost the same magnitude for every year. The magnitude of interannual variability in INDVI was not greater than 10%. The maximum variation in s2NDVI was in the order of 20%