Sustainable forest management seeks to ensure that the behaviour of managed forest ecosystems is environmentally and socio-economically acceptable. In this synthesis paper I assess the actual and potential contribution of modellers to the development of sustainable forest management, from scientific understanding of the underlying processes to practical decision-support tools. On the scientific side, detailed forest ecosystem models have been developed to understand and predict management and climate impacts on forest behaviour, based on a comprehensive description of plant-soil and carbon-nutrient-water interactions. However, two key processes for sustainable forest management - plant growth allocation and soil nutrient cycling - continue to challenge modellers. Hydraulic homeostasis has recently emerged as a guiding principle for modelling growth allocation. Combining this with root-shoot functional balance leads to a simple scheme incorporating both hydraulic and nutritional constraints on allocation for use in simplified, process-based growth models. While this scheme predicts realistic growth and yield trends with stand age, the individual roles played by allocation and stomatal conductance require further study. Hydraulic homeostasis alone cannot separate these. Soil nutrient cycling models differ in how they describe the regulation of microbial growth and diversity, two key processes for sustainable management. Existing models of the former probably suffice (at least for N cycling), but uncertainty in the latter presents a major limitation for predicting long-term ecosystem behaviour; more experimental work is required here. Models should also incorporate non-microbial immobilization processes. On the practical side, mensuration-based growth and yield models have been incorporated into decision-support tools designed to find management practices satisfying mainly economic objectives (e.g. maximum net present value), subject to various constraints (e.g. even wood production). Information from process-based models has yet to make a significant contribution. I discuss three options for enhancing the information flow from process-based models to decision-support tools: (i) use of process-based growth indices to improve conventional growth and yield models; (ii) simplification and direct incorporation of process-based growth models; and (iii) use of mass balance analysis to generate robust constraints between, for example, wood yield, ecosystem C storage and site nutrient loss. This last option provides a particularly promising way forward, whereby the environmental criteria of sustainability might be incorporated as extra constraints into established decision-support tools.