Amazonian rainforest is undergoing increasing rates of deforestation, driven primarily by cattle pasture expansion. Forest-to-pasture conversion has been associated with increases in soil methane (CH₄) emission. To better understand the drivers of this change, we measured soil CH₄ flux, environmental conditions, and belowground microbial community structure across primary forests, cattle pastures, and secondary forests in two Amazonian regions. We show that pasture soils emit high levels of CH₄ (mean: 3454.6 ± 9482.3 μg CH₄ m^-2 d^-1), consistent with previous reports, while forest soils on average emit CH₄ at modest rates (mean: 9.8 ± 120.5 μg CH₄ m^-2 d^-1), but often act as CH₄ sinks. We report that secondary forest soils tend to consume CH₄ (mean: -10.2 ± 35.7 μg CH₄ m^-2 d^-1), demonstrating that pasture CH₄ emissions can be reversed. We apply a novel computational approach to identify microbial community attributes associated with flux independent of soil chemistry. While this revealed taxa known to produce or consume CH₄ directly (i.e. methanogens and methanotrophs, respectively), the vast majority of identified taxa are not known to cycle CH₄. Each land use type had a unique subset of taxa associated with CH₄ flux, suggesting that land use change alters CH₄ cycling through shifts in microbial community composition. Taken together, we show that microbial composition is crucial for understanding the observed CH₄ dynamics and that microorganisms provide explanatory power that cannot be captured by environmental variables.