Résumé
During the Late Pleistocene and early Holocene 59 species of South American megafauna went extinct. Their extinction potentially triggered population declines of large-seeded tree species dispersed by the large-bodied frugivores with which they co-evolved, a theory first proposed by Janzen and Martin (1982). We tested this hypothesis using species range maps for 257 South American tree species, comparing 63 species thought to be primarily distributed by megafauna with 194 distributed by other animals. We found a highly significant (p < 0.001) decreased mean range size of 26% for the megafauna dispersed fruit (n = 63 species) versus fruit dispersed by other animals (n = 194), results which support the hypothesis. We then developed a mathematical model of seed dispersal to estimate the theoretical impact of megafauna extinction on tree species range and found the estimated dispersal capacity (phi(seed)) of a 2 g seed decreases by > 95% following disperser extinction. A numerical gap dynamic simulations suggests that over a 10 000 yr period following the disperser extinctions, the average convex hull range size of large-seeded tree species decreased by approximate to 31%, while the estimated decrease in population size was approximate to 54%, indicating a likely greater decrease in species population size than indicated by the empirical range patterns. Finally, we found a positive correlation between seed size and wood density of animal-dispersed tree species implying that the Late Pleistocene and early Holocene megafaunal extinctions reduced carbon content in the Amazon by approximate to 1.5 +/- 0.7%. In conclusion, we 1) provide some empirical evidence that megafauna distributed fruit species have a smaller mean range size than wind, water or other animal-dispersed species, 2) demonstrate mathematically that such range reductions are expected from megafauna extinctions ca 12 000 yr ago, and 3) illustrate that these extinctions may have reduced the Amazon's carbon storage capacity.