Models are often used to estimate regional greenhouse gas emissions by scaling up verified point data to a region, and errors may arise in this spatial up-scaling process. To compare the error sizes of different up-scaling methods, we collected input data of 59 sites required by CH4MOD in Jiangsu Province of China to simulate CH₄ emission from rice paddies, and applied three upscalling methods: 1) Take the average of all input parameters of 59 sites as the input parameters to model, then calculate the regional CH₄ emission, which is average-parameter method; 2) Apply the parameters of one representative sample site to the model, and extend the estimation directly to the regional scale, which is typical-point method; 3) Generate 100 virtual sites randomly according to the statistical characteristics of the parameters of 59 samples, then input them to run the model 100 times and get the average of simulated results as the regional CH₄ emission, which is the Monte Carlo method. The results showed that, when compared with the average of simulated results of the 59 individual sites, the average-parameter method and the typical-point method have a relative error of -19.60% and -19.74%, respectively, whereas the Monte Carlo method can effectively reduce the relative error to 3.29%. We also analyzed the error sources of average-parameter method and the typical-point method.