The behavioral dependence of vegetation simulation models for spatially heterogeneous grasslands on simulationresolutionwas investigated. The dependence can be largely attributed to the non-linearity of themodels.We showedthat increasing scale or decreasing spatial resolution tended to overestimate the changing rate of an ecosystem usingour landscape simulation model for alkaline grasslands in northeast China. A technique for scaling up simulationmodels with diffusive transportation was developed in this study by means of expanding the nonlinear drivingfunctions in the model. The analysis showed that a simulation model for spatially heterogeneous landscapes mightnecessitate modification of both its mathematical structure and parameterization when applied to different scales.The scaling coefficients derived in this study were shown to be proportional to the variances or covariance of thespatially referenced variables, and can be estimated by running the model at a fine resolution for selected samplesof the coarser grid cells. The technique was applied to a grassland landscape in northeast China and the resultswere compared with five-year observations on community succession. The comparison indicated that the proposedtechnique could effectively reduce overall scaling error of the model by as much as 80%, depending on the scalingdifference between the fine and the coarse resolutions as well as the sampling scheme used.