My research aims at characterizing the responses of forests (boreal, tropical and Mediterranean) and savannas to climate change (mainly Holocene and future) in relation to natural disturbances. To study these ecosystem dynamics, I work both with field data representative of current or past conditions, paleoecological archives and modelling. I seek to quantify the natural variability that ecosystems have recorded in their dynamics to date and then estimate whether their future dynamics will go beyond their natural envelope. If this is the case, it is necessary to be able to recommend suitable practices in the framework of sustainable ecosystem management. I use quantitative models of fire danger and fire behaviour, spatially explicit or not, different versions of dynamic global vegetation models (DGVM), but also since more recently, qualitative models such as petri nets and graph transformations.