A process-based model for methane emission predictions from flooded rice paddies

Estimation and prediction of methane emission from flooded rice paddies is impaired by the large spatial and temporal variability in methane emissions and by the dynamic nonlinear relations between processes underlying methane emissions. This paper describes a process-based model on methane emission prediction from flooded rice paddies that can be used for extrapolation. The model is divided into two compartments; rhizosphere, which is a function of root length density, and bulk soil. The production of carbon substrates drives methane emission and originates from soil organic matter mineralization, organic fertilizer decomposition, in both compartments, and root exudation and root decay, in the rhizosphere compartment only. It is assumed that the methanogens are completely outcompeted for acetate by nitrate and iron reducers but that competition takes place with sulfate reducers. Produced methane is transported to the root surface in the rhizosphere or the soil-water interface in the bulk soil. Transport time coefficients are different for the two compartments. Part of the methane is oxidized, a constant fraction of produced methane in the bulk soil, whereas the oxidation fraction varies according to root activity dynamics in the rhizosphere. The remaining methane is emitted to the atmosphere. The model was validated with independent field measurements of methane emissions at sites in the Philippines, China, and Indonesia with only few generally available site-specific input parameters. The model properly predicts methane emission dynamics and total seasonal methane emission for the sites in different seasons and under different inorganic and organic fertilizer conditions. A sensitivity analysis on model assumptions showed that the assumptions made in this model are reasonable and that the division into two compartments was necessary to obtain good results with this model. The combination of proper prediction and the necessity of few input parameters allow model application at regional and global scales.