This study evaluates both the impact of two deforestation scenarios on the distribution of clouds and precipitation in the Amazon, and how these shifts in the regional climate could affect the productivity of the remaining forests. We use a fully coupled ecosystem dynamics-atmosphere model (ED2-BRAMS), which provides a thorough and realistic representation of the vegetation heterogeneity as well as the atmospheric dynamics. The present day ecosystem state of tropical South America is generated using offline simulations (ED2) forced with an improved reanalysis product developed by Sheffield et al (2006), increasing CO2, and historic land use changes. Two future ecosystem states were then generated based on two different deforestation scenarios: SimAmazonia1 in which large areas of the Amazon were converted to pastures (Soares-Filho et al. 2006); and SimAmazonia2 in which most of the future changes in land use were due to logging, with lower contribution of land clearing (Merry et al. 2009). The impact of these three different ecosystem states (present day, SimAmazonia1, and SimAmazonia2) on the coupled forest-climate system is being investigated using the fully coupled ED2-BRAMS model runs. Preliminary results suggest that widespread land clearing leads to a significant reduction in rainfall across most of the Eastern Amazon. Logging, on the other hand, had a reduced impact on the regional rainfall pattern. Our results indicate that higher rates of deforestation could cause loss of productivity in the remaining forest due to increased water stress on the remaining forests.