Detection of Mesoscale Seasonal and Interannual Variation in the Vegetation of the Amazon Basin
In the Amazon basin, seasonal and interannual spectralchanges measured by satellites result from anthropogenic disturbance and from the interaction between climate variation and the surface cover. Measurements of spectral change, and the characterization of that change, provide information concerning the physical processes evident at this mesoscale. A 17-yr sequence of daily Advanced Very High Resolution Radiometer (AVHRR) global area coverage (GAC) images were analyzed to produce a monthly record of surface spectral change encompassing El Niņo-Southern Oscillation (ENSO) cycles. Monthly cloud-free composite images from daily AVHRR data were producedby linear filters that minimized the finescale spatial variance and allowed for awide range analysis within a consistent mathematical framework. Here the useof a minimized local variance (MLV) filter that produced spatially smoothimages in which major land-cover boundaries and spatial gradients are clearlyrepresented is discussed. Changes in the configuration of these boundaries and the composition of the landscape elements they defined are described in terms of quantitative changes in landscape pattern. The time series produced with the MLV filter revealed a marked seasonal difference in the pattern of the landscape and structural differences over the length of the time series. Strikingly, the response of the region to drier El Niņo years appears to be delayed in the MLV series, the maximum response being in the year following El Niņo with little or no change seen during El Niņo.