AbstractFloodplains lakes are abundant in the Amazon basin and are important methane sources to the atmosphere. Existing biogeochemical models require modifications and inclusion of hydrodynamic processes operative in shallow, warm waters to be applied to these aquatic ecosystems. We modified a 1‐dimensional process‐based, lake biogeochemical model and combined a 3‐dimensional hydrodynamic model to suit Amazon floodplains. We evaluated the combined model's performance simulating methane concentrations and fluxes and several related processes in the open lake and an embayment of a well‐studied Amazon lake. Parameters for calibration were selected through sensitivity tests using a machine learning‐based algorithm, classification, and regression trees. Comparison between simulated and measured fluxes indicate generally good agreement in seasonal patterns and magnitudes. Comparisons of near‐surface concentrations varied with no clear patterns. Simulations of methane concentrations at near‐surface and near‐bottom, and diffusive emissions are most sensitive to carbon mineralization rate, Q10 factors for methanogenesis and oxidation, and methane oxidation potential. Modeled rates of planktonic photosynthesis were generally lower than measurements, though simulated planktonic respiration was often similar to measurements. Simulated rates of methane oxidation were considerably lower, with a few exceptions, than measurements of methane oxidation in oxic water of the lake. Improvements of results of the linked hydrodynamic‐biogeochemical model will result from inclusion of advective transport, use of parameter values appropriate for tropical waters, especially for methane oxidation and photosynthesis, and addition of changes in hydrostatic pressure to model of ebullition.