Conceptual and phenomenological macroecological models of current global fire activity have demonstrated the overwhelming control exerted by primary productivity. Fire activity is very high in savanna regions with intermediate primary productivity, and very low in both densely forested regions with high productivity and arid/cold regions with low productivity. However, predicting future global fire activity using such macroecological models of fire's global ‘niche’ may not be possible because of the feedbacks between fire, climate and vegetation that underpin the fire−productivity relationship. Improving forecasts of global fire activity demands the use of dynamic models to determine how climate, CO2, vegetation (i.e. canopy closure and plant functional types) and primary productivity constrain fire and evaluation of the strength of feedbacks amongst these variables.