Model estimates of methane emission from regional and global rice paddies have been gaining international interesting. Structurally, methane emission models are characterized in terms of their component processes and their inter-relationships. Once a model is developed, it is necessary to take into account the many sources of uncertainty in the processes modeled. Uncertainties of model estimates come primarily from spatial database errors when up scaling is performed, while the magnitude of the uncertainties induced by the database errors depends straightly on the sensitivity of model output to the input parameters. The objective of this paper is to develop a methodology of sensitivity analysis applicable to a methane emission model CH4MOD, and to further link the sensitivity analysis with the spatial database errors for assessing the uncertainties of model estimates of methane emission from Chinese rice paddies. Sensitivity of CH4MOD output to the model input is characterized by a sensitivity index defined as a ratio of relative change of model output (dy/y) to that of model input parameter (dx/x). By applying the statistical analysis on a vas
t amount of model outputs against the random combinations of the model parameters, the sensitivity index was first calculated for each parameter in CH4MOD. The uncertainties of CH4MOD output were then assessed by linking the sensitivity index with the spatial database errors. The parameters in CH₄MOD include soil sand percentage (Sand), rice variety index (VI) and grain yield (GY), rice aboveground biomass at transplanting (W0) and the intrinsic growth rate of aboveground biomass (r), organic matter amendment (OM) and irrigation regime (WPtn). Sensitivity analysis showed that the irrigation regime is the most sensitive factor with a sensitivity index of 0.64. The sensitivity of remaining parameters is ranked by their sensitivity indices as Sand (0.50) ＞ VI (0.48) ＞ OM (0.43) ＞ r (0.42)＞ GY (0.32) ＞W₀ (0.08). By running CH4MOD, methane emission from Chinese rice paddies was estimated to be 6.02 Tg in the year of 2000. The uncertainties ranged from 3.09 to 10.61 Tg based on the calculations from the sensitivity indices and the spatial database errors. These values are similar to previous model estimates reported by other scientists. It was concluded that the sensitivity index proposed in this paper is capable of quantifying the sensitive difference among the model parameters. Moreover, the uncertainties of model estimates on a regional scale can be evaluated by linking the sensitivity index with the model input parameters in terms of spatial database errors.