Tropical forests are subjected to severe seasonal and ENSO-related interannual fluctuations in precipitation, yet transpiration and CO2 exchange in these forests are resilient, partly due to their ability to extract water from deep soil reservoirs. The physiological response of forests and secondary vegetation covers to drought stress determines the seasonal and interannual variations in carbon and water exchange, which feeds back to regional CO2 concentrations, hydrology, and circulation patterns. Land-use change in tropical regions is characterized by widespread replacement of primary rainforest by pasture and croplands which have very different drought-response characteristics. Proper characterization of the responses of each of these systems to climate variability is therefore crucial for analysis and prediction of carbon and water cycle changes in these areas.

We explored the interactions between ecosystem-level drought stress, carbon exchange, and precipitation using the simple biosphere model (SiB2) and the Colorado State University GCM. We found extreme sensitivity of the GCM-simulated precipitation to the parameterization of rooting depth, soil hydrology, and the parameterization of ecosystem stress in SiB2. Model parameterizations were evaluated at several forest and pasture sites across the Amazon Basin where long-term flux tower time series were available. The standard model was found to be overly sensitive to drought stresses, and alternative parameterizations were explored. Finally, the impact of the alternative parameterizations on global precipitation and circulation patterns was explored in the fully coupled GCM.