A new fire model is proposed which estimates areas burnt on a macro-scale (10–100 km). It consists of three parts: evaluation of fire danger due to climatic conditions, estimation of the number of fires and the extent of the area burnt. The model can operate on three time steps, daily, monthly and yearly, and interacts with a Dynamic Global Vegetation Model (DGVM), thereby providing an important forcing for natural competition. Fire danger is related to number of dry days and amplitude of daily temperature during these days. The number of fires during fire days varies with human population density. Areas burnt are calculated based on average wind speed, available fuel and fire duration. The model has been incorporated into the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM) and has been tested for peninsular Spain. LPJ-DGVM was modified to allow bi-directional feedback between fire disturbance and vegetation dynamics. The number of fires and areas burnt were simulated for the period 1974–94 and compared against observations. The model produced realistic results, which are well correlated, both spatially and temporally, with the fire statistics. Therefore, a relatively simple mechanistic fire model can be used to reproduce fire regime patterns in human- dominated ecosystems over a large region and a long time period.