USA — Every year, wildfires burn through woods and meadows, trees and grass, scorching pines, firs and shrubs. At times, the blazes reach into residential areas as firefighters struggle to contain them, and researchers try to find better ways to respond.
The financial costs of wildfires add up to $ 1 billion or more every year, while the damage done to people, communities and ecosystems is more difficult to quantify. A new method for predicting the potential effects and characteristics of fires that a Penn State professor and graduate student recently tested has the potential to help fire managers better plan for how fires will behave in the long term and help them decide how to respond.
Known as Random Forest, the statistical modeling approach Alan Taylor, professor of geography, studied and found to be effective is anything but random. Instead, the method crunches large amounts of data on the slope, soil, vegetation and fuels in field plots and uses an algorithm to develop a model to predict where different types of fuel that influence fire behavior occur across a landscape.
Andrew Pierce, a graduate student who is now doing post-doctoral work, and Taylor described their findings in the journal Forest Ecology and Management. The method can generate continuous fields for modeling fire behavior and assessing fire effects over large landscapes. So, rather than getting a snapshot of how a wildfire could behave, the approach allows for the development of more refined information for fire management planning.
It has great promise, Taylor said. I expect well start using it more widely.
Predicting how fires will behave in forest canopies is particularly difficult, Taylor said, but important in determining whether a fire has the potential to be severe and climb to the tree crowns and kill the stands.
Wildfires are burning through more land each year. Overall, the past decade has seen the biggest wildfire years since 1960. In 2012, 9.2 million acres burned. That is up from the 3.57 million acres affected in 2001 and 2.95 million in 1991, according to the National Interagency Fire Center.
One contributing factor to the growth in wildfires is the fire suppression that has been the norm. Stopping and preventing forest fires in places that have burned in the past allow fuel for future fires to build up. Taylor said warming global temperatures will be another factor and another reason finding ways to manage forest fires will become increasingly important.
Developing predictive tools will be very helpful for fire planning and management and could help reduce the social and economic costs of fighting wildfires, said Taylor, who is part of Penn States Earth and Environmental Systems Institute.
Using the Random Forests modeling method, Pierce and Taylor quantified the characteristics of surface and forest canopy fuels what type and how much there is on the ground and how tall and dense forest crown fuels were at Lassen Volcanic National Park in northern California. That meant collecting data from tree ring scars, maps of contemporary fires, historical data and field data, together with satellite data, then bringing that together to predict where different types of fuel occur on the landscape. The team then used the fuel maps in fire behavior models to predict where fire effects were likely to be low or high in severity and compared the model results to a 2004 wildfire. There was a good match between their predictions and the patterns of severity in the 2004 wildfire.
Taylor said a breakthrough for fire services was a new standardized fuels map for the country that responders use when fires break out.
It really is important, he said. But, he added, If youre making tactical decisions about what to do, theres lots of room for improvement.
That is where their Random Forests fuels mapping approach could come in. The paper was essentially the proof of concept. Now, Taylor said, it could be used by government agencies to help make decisions and plan for the long run.
Funding for the research comes from the interagency Joint Fire Science Program, National Park Service and National Science Foundation.