Economic incentives and improving infrastructure in tropical countries are creating land-use mosaics wherein small-scale farming and industrial plantations are embedded within native ecosystems. Practices such as agroforestry, slash-and-burn cultivation, and oil palm monocultures bring widely different impacts on carbon stocks. Characterizing these production systems is both crucial to attribute deforestation to particular drivers and essential to understand the impact of macroeconomic scenarios, national policies, and land tenure schemes on carbon fluxes and biodiversity. The aim of this project is to develop and test a systematic methodology that integrates L- and C-band Synthetic Aperture Radar as well as optical instruments to map oil palm plantations in tropical biomes. We have selected study sites in Costa Rica, Peru, Indonesia and Malaysia in order to assess model generality and characterize error sources. We propose an approach to continuously update maps through a learning Bayesian algorithm and crowd-sourced validation.