We review the state—of—the—art of models of forests and grasslands that could be used to predict the impact of a future climate change arising from increased atmospheric carbon dioxide concentration. Four levels of resolution are recognized: physiologically based models, population models, ecosystem models, and regional or global models. At the physiological level a number of important processes can be described in great detail, but these models often treat inadequately interactions with nutrient cycles, which operate on longer time scales. Population and ecosystem models can, on the other hand, encapsulate relationships between the plants and the soil system, but at the expense of requiring more ad ho formulations of processes. At the regional and global scale we have so far only steady—state models, which cannot be used to predict transients caused by climate change. However, our conclusion is that, in spite of the gaps in knowledge, there are several models based on dominant processes that are well enough understood for the predictions of those models to be taken seriously.