Models make it possible to investigate the factors which influence forest fires and to measure their importance. Using various forest fire models, the works presented here examine the influence of weather, forest composition, human activity and changes in legislation on the likelihood of forest fire ignitions in Ticino and Valais. A distinction was made between forest fires started by flash of lightning, and those resulting from human activity. The results show that the weather has the greatest influence where lightning starts, whereas in fires caused by people, the weather takes a subordinate place to human activities. Depending on the ignition causes, different weather indices best represent the danger of forest fires: for those caused by lightning, the Duff Moisture Code (DMC) drought index, and for those started by human activity, the Angstrom Index. In order to test the general validity of forest fire ignition models these were applied to Ticino and to Valais over two different periods of time. Results show that transferability of the models is limited, and that their use for the assessment of the future risk of forest fire is difficult under changing climatic conditions. The landscape model LandClim was used in order to simulate the observed patterns of fire frequency and size in Ticino and in Valais. Thanks to further development of the forest fire module, LandClim achieved a marked improvement of modelquality. Such dynamic landscape models should have an important role to play in assessing future forest fire regimes.