How forest fires regulate their own propagation in Alaska
A study reveals that a natural feedback mechanism between fires and vegetation regeneration limits fire spread in Alaska. Scientists decipher the mechanism using satellite data and AI models, but uncertainty grows with climate warming.
Alaska: how forest fires regulate their own propagation
Every decade, Alaska's boreal forests see thousands of square kilometers burned, but a natural mechanism limits fire frequency. This dynamic, long underestimated, has been brought to light by an international team combining field observations and artificial intelligence.
A vegetation feedback that tempers fires
Analysis shows that forest fires first eliminate mature trees, which are highly combustible, making way for younger, clearer, and less dense forest growth. These young plants, due to their high moisture content and low height, burn less easily, creating a window of several years where the probability of a new fire decreases. This feedback acts as a natural regulator, slowing fire progression in boreal expanses.
The underlying scientific principle is based on the notion of fuel load. In a mature forest, the accumulation of dead wood, thick bark, and low branches creates a layer of highly flammable organic material. When a fire occurs, it consumes this carbon-rich layer and releases a large amount of heat, but it also destroys the canopy that maintained a moister microclimate by shading the ground. After the fire passes, the ground is exposed, but initial regeneration is dominated by pioneer species, often shrubs or fast-growing conifers with more water in their tissues. This internal moisture acts as a "thermal buffer", absorbing some of the fire's energy and reducing its spread rate.
Moreover, the physical structure of young stands is less vertical: trunks are short, branches remain close to the ground, and canopy density is lower. This configuration limits the ability of fire to "jump" from tree to tree, a phenomenon known as crown fire, which drives large fires. Thus, as long as the forest remains in this regeneration phase, the probability of another major fire is temporarily reduced.
To untangle this process, the team exploited decades of satellite data from the Copernicus program, combined with ECMWF model simulations. A neural network trained by machine learning extracted fire and vegetation growth signatures at high spatial resolution. The predictive model thus constructed could compare recently burned areas with those where vegetation was in regeneration.
