Understanding how fire regimes change over time is of major importance for understanding their future impact on the Earth system, including society. The fire model intercomparison project aimes at improving our abilities to model global fire occurrence and its impacts, to assess its uncertainties and development needs. Several modelling groups provided historical simulations of fire-enabled vegetation models and senstivity simulations. I will show a brief evaluation of simulated burned area and then identify the main driving factors over the 20th century for inter-annual and long term variability of burned area based on the sensitivity simulations. Those simulations isolate the influence of land use and land cover change, population density, CO2, lightning and climate. Where possible I will relate the inter-model variation in the sensitivities to model assumptions.While fire occurrence is well observed by satellites, the impacts of fire are less well constrained by observations especially on the regional to global scale. Comparing a reference simualation with fire to a simulation without fire I will provide estimates of the impact of fire on vegetation and the carbon cycle. Based on these results I will discuss model limitations and highlight consequences for model development and applications.