Some portions of land in Brazilian Amazon are forested, and other portions used in agriculture. Deforestation (reforestation) emits (captures) carbon, which has consequence for the global climate. The social and private productivities for these alternative land uses vary across locations within the Amazon region. In this research, we build and analyze a spatial/dynamic model of socially efficient land allocation to establish a benchmark for ‘ad-hoc’ policies. We show how to incorporate the stochastic evolution of agricultural prices, and we explore the consequences of ambiguity in the location-specific productivities on the socially efficient policy. Finally, we assess the consequences of imposing alternative social costs of carbon emissions on the spatial/dynamic allocation of land use. Our results indicate that with modest transfers per ton of net CO_2, Brazil would find it optimal to choose policies that produce substantial capture of greenhouse gasses in the next 30 years, suggesting that the management of tropical forests could play an important role on climate change mitigation in the near future.