We characterized the seasonal pattern of Normalized Difference Vegetation Index (NDVI) of 49 grassland and shrubland sites corresponding to areas with low human impact. Each site was characterized as a vector of 21 elements, where each element corresponded to the average NDVI for one date. We used the integral of the NDVI as an estimate of the total production, and the difference between the maximum and minimum NDVI over the year as a measure of the seasonality. Based on a Principal Component Analysis, the integral of the NDVI accounted for 63% of the variability of the SITE x DATE matrix. The difference between the maximum and minimum NDVI throughout the year accounted for 19%. Mean annual precipitation explained 67% of the variability of the integral of the NDIV curves (P < 0.01). The slope of the relationship between precipitation and integrated NDVI was significantly steeper for grasslands than for shrublands. The difference between the extreme values of NDVI increased with mean annual precipitation and the thermal amplitude, and decreased with mean annual temperature (P < 0.01).
Read full abstract