ITHACA, N.Y. -- As small earthquakes can be omens of larger ones and landslides can be precursors to avalanches, Cornell University geologists have shown in a computer simulation that forest fires display the same natural behavior. Their findings, they believe, could be used to predict where large forest fires can occur -- and how to prevent them.
The researchers' findings appear in the latest issue of the journal "Science "(Sept. 18, 1998).
"What is surprising to me is that an event like a forest fire is so similar to other natural events," says Donald L. Turcotte, the M.M. Upson Professor of Engineering in the Cornell Department of Geology. "And humanity really plays a small role in these events."
Turcotte and his fellow researchers, Bruce D. Malamud, a Fulbright Scholar and visiting lecturer in geology at Cornell, and Gleb Morein, a Cornell graduate student, built their computer model of forest fires and analyzed data sets from a number of forests and wildfires from around the world, including Yellowstone National Park.
Until 1972 Yellowstone had a policy of suppressing forest fires. This resulted in a large accumulation of dead trees, undergrowth and very old trees that became perfect tinder for fires. The researchers contend that the large Yellowstone fire of 1988, which burned 800,000 acres, could have been prevented if the policy of letting smaller fires burn to completion had been in place before 1972. The smaller fires would have eliminated the underbrush and dead wood earlier, thus reducing the likelihood of a large fire, they say.
In analyzing how forest fires start and propagate, the researchers found that the frequency distribution of small and medium fires can be used to assess the risk of larger fires, as small tremors are routinely used to assess the risk of larger earthquakes.
Turcotte explains that for natural occurrences there is a return period for events of different magnitudes. For example, meteorologists classify storms as 5-year, 10-year, 50-year or 100-year. This is an example of behavior on the part of weather. "It's an interesting class of phenomena," he says. "Small landslides build up to large landslides, and I suppose this model can be applied to something like the stock market, which also shows a degree of self-organized critical behavior."
To model forest fires, the scientists developed a computer grid of a forest, then simulated dropping a match. If the match landed on a tree, the tree and its neighbors would burn. If the match landed on an area without trees, no fire ensued. Where the grid was packed with trees, a large fire ensued. The scientists say that the only previous application of this type of model was to study measles epidemics in isolated populations.
The scientists ran the model under different scenarios of forest density. Interestingly, for each scenario, the researchers found a range of small to large fires and many more small fires than larger ones, which correlates with the law of a fractal distribution. (Fractal distribution is the magnification of certain things, like the size of a fire, in proportion to their original size.)
Large forest fires are dominant when the forest is densely populated. This was demonstrated by the researchers when the computer grid was filled with trees: The fires spanned the grid. This is what happened in Yellowstone National Park in 1988. The scientists say that forest-fire professionals now recognize that the best way to prevent the largest fires is to allow the small and medium fires to burn.
The scientists found that actual forest fires have fractal distributions over many orders of magnitude. However, the environmental- and human-related variables that affect the size of wildfires are many, including the proximity and type of combustible materials, weather conditions and firefighting efforts to extinguish certain fires, the researchers say.
"Despite these complexities, the predicition capability of the forest fire model appears to be robust," says Turcotte.
The Science article is titled "Forest Fires: An Example of Self-Organized Critical Behavior." Funding for the research was provided by a grant from the National Aeronautics and Space Administration (NASA).
Journal
Science