Numerous forest dynamics models have been developed to make reliable long-term and large-scale predictions using available short-term and small-scale empirical data on forest dynamics. Associated uncertainties in model projections, if reported, are usually restricted to model stochasticity, variation between plots, or parameter estimation uncertainty; few, if any, modeler includes the uncertainty due to model assumptions. Our objective was to quantify these sources of uncertainty and evaluate if uncertainty due to two very common modeling assumptions (i.e., dynamic equilibrium and maximum tree size assumptions) would be greater than the uncertainties due to the other sources. Apart from forest dynamics modeling, these two assumptions are pervasive in wildlife management and in ecological theory in general. We used an individual tree spatially explicit model called SYMFOR, calibrated for tropical rain forests within the Brazilian Amazon. Our results suggest that these two assumptions in forest dynamics modeling can result in significant changes in mean projected timber yield and biomass, both in heavily logged and unlogged forests. Also, depending on the way these assumptions are accommodated in the model, contrasting model projections might emerge. These assumptions also increase overall model uncertainty but, in contrast to other sources of uncertainty (e.g., model stochasticity, parameter and plot uncertainty), assumptions are an uncertainty source that cannot generally be reduced without specific studies targeting them, revealing key gaps in our knowledge about the ecosystem. The joint analysis of all these sources of uncertainty indicate that uncertainties associated with model projections are likely to be greater than the uncertainty frequently reported.
