| dc.description.abstract | The vulnerability of Amazonian rainforest, and the ecological services it provides, depends on
an adequate supply of dry-season water, either as precipitation or stored soil moisture. How the
rain-bearing South American monsoon will evolve across the twenty-first century is thus a
question of major interest. Extensive savanization, with its loss of forest carbon stock and
uptake capacity, is an extreme although very uncertain scenario(1-6). We show that the
contrasting rainfall projections simulated for Amazonia by 36 global climate models (GCMs) can
be reproduced with empirical precipitation models, calibrated with historical GCM data as
functions of the large-scale circulation. A set of these simple models was therefore calibrated
with observations and used to constrain the GCM simulations. In agreement with the current
hydrologic trends(7,8), the resulting projection towards the end of the twenty-first century is for a
strengthening of the monsoon seasonal cycle, and a dry-season lengthening in southern
Amazonia. With this approach, the increase in the area subjected to lengthy-savannah-pronedry
seasons is substantially larger than the GCM-simulated one. Our results confirm the
dominant picture shown by the state-of-the-art GCMs, but suggest that the model democracy
view of these impacts can be significantly underestimated. | |
