Principal component analysis applied to a time series of MODIS images: the spatio-temporal variability of the Pantanal wetland, Brazil

TLDR: In this paper, a principal component analysis (PCA) was applied to a complete annual dataset of filtered EVI2 images (based on a 12-year average over the 2001-2012 period).

Abstract: The present study aimed at analyzing the spatio-temporal variability of the Pantanal vegetation cover, the largest tropical wetland in the world. A principal component analysis (PCA) was applied to a complete annual dataset of filtered EVI2 images (based on a 12-year average over the 2001–2012 period). There was about 99 % variance concentration in the first three components, with the respective loading responses and distributions (maximum, minimum and changes in the sign of the eigenvector loadings) matching the most significant seasonal interruptions. The first three principal components showed the essential aspects of the spatio-temporal variability of the local phenology, i.e. the cumulative greenness throughout the year, the later and more generalized senescence associated with the drought season climax, and the early senescence associated with sandy portions. Our results enabled the detection of homologous areas regarding vegetation density and the time and intensity of senescence. As the water availability throughout the year—the most important parameter for regional vegetation—is largely a function of geology (sediment grain size and vertical neotectonic), a geobotanic analysis of the Pantanal wetlands was also possible. Our PCA-based approach was able to capture the essentials of the phenological/environmental variability, with potential for application in other ecosystems with complex vegetation cover and functioning.