Global methane emission estimates depend highly on the models, techniques, and databases used. Since emissions cannot be measured directly at large scales, it is impossible to judge which estimate is more realistic. In this paper, different aspects of uncertainty in upscaling methane emissions from rice paddies are discussed. These aspects are visualized by a case study on the spatial upscaling of methane emissions from the island of Java, Indonesia. The first aspect concerns process information. An approach to incorporate this information in a simplified but process-based way in predictive models is discussed. Sources of uncertainty include the methane emissions measurements, processes quantification, process simplification, and the use of data transfer functions. Data availability of input parameters, the second aspect, is uncertain because of differences between different data sources, the use of data sources for purposes not originally planned for, and the scale at which data are available. Data interpolation in combination with nonlinear model responses introduces scaling errors, the third aspect. Data accuracy introduced the highest uncertainties in emission estimates but is rarely accounted for in the estimation of global emissions.