Here I present a new approach to forecasting the effects of climate change on catastrophic events, based on the 'self-organised criticality' concept from statistical physics. In particular, I develop the 'self-organised critical fuel succession model' (SOCFUS), which deals with wildland fires. I show that there is good agreement between model and data for the response pattern of the whole fire size statistical distribution to weather fluctuations in a boreal forest region. I tentatively predict the fire regime in this region for an instance of possible climate change scenario. I show that the immediate response is sharper than usually thought, but part of the added burning rate might not persist indefinitely. A large fraction of the extra burning in the transition period is likely to be concentrated in a few 'climate change fires', much larger than the largest fires that could currently occur. I also suggest that the major fire events recently observed in some tropical rainforest regions belong to a qualitatively different, even more abrupt type of response, which is also predicted by the model.