1. It is widely appreciated that management shapes the dynamics of many ecological systems, but ecologists rarely consider the reverse interaction, that is, the ecological influences on management decisions. Reciprocal feedback between management and ecology can cause complex system behaviour. Therefore, better predictions about how external policy-drivers or climate change will affect semi-natural ecosystems may be made when both the ecological and human dimensions are considered. 2. We develop a spatially-explicit model of moorland vegetation dynamics and management decisions about sheep grazing and heather burning in the Peak District National Park, UK. Competition between dwarf shrubs, bracken and graminoids is mediated by grazing, dwarf shrub age (determined by burning rotation) and environmental gradients. Management decisions depend on vegetation cover in a model parameterized through interviews with upland managers. 3. Current management regimes are designed to reverse historical dwarf shrub losses, and simulations suggest that this reversal should occur in the future. After equilibration, grazing densities fall and dwarf shrubs have expanded from their current distribution, mainly at the expense of graminoids. This causes more land to come under managed burning, but current intensities are maintained. 4. Enforcing winter or summer grazing densities influences model vegetation cover and causes other aspects of the management strategy to adapt. For example, when summer grazing is banned, dwarf shrub cover increases and there is a shift towards grouse moor management. 5. Simulations with warmer temperatures indicate that climate change may increase bracken invasion of the moorland and prevent re-vegetation of bare peat. This is associated with a reduction in managed burning that causes the dwarf shrub community to become dominated by the older, degenerate growth phase. 6. Synthesis and applications. Our model suggests that current management paradigms could achieve their aim of restoring historically degraded moorland over the coming century, but that climate change may prevent this from occurring. One application of the model would be to try to design management paradigms that are robust to this. As such, models of coupled human–natural systems can provide a valuable tool for assessing the impacts of policy decisions and climate change on semi-natural ecosystems at landscape scales.