Soil maps are crucial for habitat and species distribution modeling under present and future conditions, thereby providing relevant background information for conservation planning. In Amazonia, soil conditions are highly heterogeneous, which has important implications for the distribution and dynamics of the area's exceptional biodiversity. Unfortunately, available soil maps for this region suffer from inaccuracies and lack of ecologically relevant variables. Here, we develop a map of the sum of exchangeable base cation concentration (SB) in the surface soil by applying machine learning to a comprehensive set of over 10,000 field data points of SB values directly measured from soil samples or inferred using indicator plant species occurrences. As predictors, we used rasters of soil type probabilities, elevation, biomass and reflectance values from Landsat satellite images. Random Forest (RF) models were trained and tested using two different cross-validation strategies. We also assessed in which areas the map was more reliable using the area of applicability approach and compared the results with two other soil layers. The best predictors of SB variation were Landsat bands 7, 4 and 3, elevation, and proba-bility of Histosols. The regional patterns observed across Amazonia were consistent with current geological understanding; lower SB values tended to occur in central Amazonian soils and higher values in western Amazonian soils, with considerable variation within each region. The model was found applicable over most of the Amazonian biome, especially in non-inundated (terra-firme) forest, but not over coastal areas, floodplains of major rivers and wetlands, which were poorly represented in the training data. Our new SB map over performed previous SB map and represent an accurate and ecologically meaningful variable. It is available as a digital GIS layer and can be used in habitat mapping and in modeling the current or future distributions of biological communities and species. This will advance general understanding of Amazonian biogeography and help in conservation planning.