Annual and interannual (ENSO) variability of spatial scaling properties of a vegetation index (NDVI) in Amazonia

The space–time variability of the Normalized Difference Vegetation Index (NDVI) over the Amazon River basin is quantified through the bi-dimensional Fourier spectrum, and moment-scaling analysis of monthly imagery at 8 km resolution, for the period July 1981–November 2002. Monthly NDVI fields exhibit power law Fourier spectra, E(k)=ck−β, with k denoting the wavenumber, c the prefactor, and β the scaling exponent. Fourier spectra exhibit two scaling regimes separated at approximately 29 km, above which NDVI exhibit long-range spatial correlations (0<β<2), and below which NDVI behaves like white noise in space (β≃0). Series of monthly values of c(t) and β(t) exhibit high negative correlation (−0.88, P>0.99), which suggest their linkages in power laws, but also that Et(k)=c(t)k−β(t), with t the time index. Results show a significant negative simultaneous correlation (−0.82, P>0.95) between monthly series of average precipitation over the Amazon, 〈P(t)〉, and scaling exponents, β(t); and high positive lagged correlation (0.63, P>0.95), between 〈P(t)〉 and 〈NDVI(t+3)〉. Parameters also reflect the hydrological seasonal cycle over Amazonia: during the wet season (November–March), β(t) ranges between 0.9 and 1.15, while during the dry season (May–September), β(t)≃1.30. These results reflect the more (less) coherent spatial effect of the dry (wet) season over Amazonia, which translates into longer (shorter)-range spatial correlations of the NDVI field, as witnessed by higher (lower) values of β(t). At interannual timescales, both phases of ENSO reflect on both parameters, as β(t) is higher during El Niño than during La Niña, due to the more coherent effects of El Niño-related dryness, whereas NDVI spatial variability is enhanced during La Niña, due to positive rainfall anomalies. Results from the moment-scale analysis indicate the existence of multi-scaling in the spatial variability of NDVI fields. Departures from single scaling exhibit also annual and interannual variability, which consistently reflect the effects from both phases of ENSO. Furthermore, departures from single scaling are independent of the order moment, q, as the PDF of departures scaled by the mean collapse to a unique distribution. These results point out that ideas of spatial scaling constitute a promising framework to synthesize important hydro-ecological processes of Amazonia.