Fighting Wildfires Before They Start

Satellite maps and sophisticated computer simulations are helping scientists spot places in danger of wildfires well before the blazes begin.

Listen to this story via streaming audio, a downloadable file, or get help.

August 28, 2001: It begins with a quick flash of blue lightning. The weather in this remote wooded valley is stormy, but rain hasn't fallen on the parched forest in weeks.

Fanned by the high winds, the small flame sparked by the lightning spreads quickly in the thick mat of dried twigs, pine needles, and fallen branches. Standing dead trees catch, and within hours the blaze has lit the canopy. A modest spark is quickly escalating into a blistering firestorm.

"Hmmm, that's no good," says the forest manager as she regards the computer screen alongside her colleague. "Let's try it again. But this time," she muses, "let's thin out the dead trees and trim all the lowest branches up to 8 meters from the ground."

Sign up for EXPRESS SCIENCE NEWS delivery

The scientist makes the changes with just a few clicks of the mouse, then restarts the fire with another bolt of virtual lightning. This time the fire just creeps along the ground, clearing out the excess clutter, and sparing the adult trees.

Above: Burning virtual trees is helping forest managers choose the best methods for reducing the risk from forest fires before they start. Image courtesy USDA Forest Service.

This scenario might sound futuristic, but computer simulations of forest fires are already transforming how land managers protect their forests and the people who live near them. By combining satellite-derived vegetation data with topographic maps, weather data, and ecological knowledge, forest scientists can construct digital landscapes on which these virtual fires burn.

The computer-assisted approach to fire risk assessment is still relatively new and only partially adopted by the fire management community, but the advantages of using computers have become widely recognized, and the technology is spreading like -- well, like wildfire.

"It's well recognized now that this is what needs to be done," says Mark Finney, a forest researcher with the USDA Forest Service's Missoula Fire Sciences Laboratory in Missoula, Montana.

"Having a digital map of forest characteristics and simulating the fire behavior of the whole map in a computer is really the future of planning efforts," Finney says.

Left: Forest managers and firefighters are able to run fire simulation programs on commonly available computers. This program -- called FARSITE -- was originated by Finney. Image courtesy USDA Forest Service. [more information]

As a ballpark estimate, Finney suspects that only 15 percent of the forestry community utilized these high-tech tools just 2 years ago. Now, he estimates, that figure would be closer to 40 percent, and within 5 years he expects the technology to be nearly pervasive.

Simulating the spread and intensity of a wildfire clearly has important uses for officials who decide how best to battle an on-going blaze. But perhaps an equally important use of these computer models is to aid decisions about how to reduce the risk from forest fires before they happen.

Forest managers have a few tricks up their sleeves -- called "fuel treatments" -- that allow them to lessen the chance that a potential forest fire will be dangerous. Low-intensity fires are a normal part of many forest ecosystems. They clear underbrush, open seed pods, and return nutrients to the soil. Mature trees normally survive these "ground fires." But if an area contains too much dead wood and dry leaves -- that is, fuel -- the tops of the trees can be ignited, starting a "crown fire" that kills everything in its path.

Right: Clearing some of the organic "fuel" from an area is one way that firefighters try to keep wildfires from getting out of control. Image courtesy National Interagency Fire Center in Boise, Idaho.

The solution is often to remove some fuel from the area. There are several ways to do this, such as controlled burns, selective logging, or trimming low branches and underbrush, to name a few. By using the computer models, managers don't have to merely guess at the best choice. They can run simulations of each option and compare the results.

Often the effects of an alteration can be complex and counterintuitive.

"We can evaluate the effectiveness of a treatment over time, because we can simulate the regrowth of the forest. Treatment A might be more effective for reducing the spread of a fire for the first 10 years, but treatment B might be more effective than A for years 10 to 50," says Kevin Ryan, project leader for the Fire Effects Project, also at the Fire Sciences Laboratory.

In the case of prescribed burns, these simulations can be particularly important.

"Fire managers are quite reluctant to set prescribed fires if they can't determine how those fires will spread," says Steve Running, a professor of forest ecology at the University of Montana who specializes in remote sensing applications. "In the Los Alamos fire [May 2000], the big problem was that fire managers guessed how the fire would spread, and they guessed wrong, so the fire burst out of control."

In order for the models to produce predictions similar to the behavior of a real fire, scientists must provide the computers with "virtual landscapes" sufficiently similar to the real terrain. With millions of acres to map, scientists must rely on remote-sensing satellites. The workhorse for land mapping in recent years has been Landsat 7, which was prepared for operation and launched by NASA's Goddard Space Flight Center, then turned over to the U.S Geological Survey for daily operations. The 30m-resolution maps produced by Landsat lay the foundation for the vegetation data fed to the models.

Above: The imaging instrument aboard Landsat 7 isn't just a camera. Called the Enhanced Thematic Mapper Plus, this instrument is a "spectrometer," which means that it takes images of the land in several different frequencies of light, including some in the infrared part of the spectrum that the human eye cannot see. This allows scientists to analyze the spectral properties of the light reflecting off the land, which provides much more information about the vegetation than a normal photograph. These two false-color images show different combinations of the frequencies detected by Landsat. Image courtesy NASA's Goddard Space Flight Center.

"Data from the Landsat imager can be used to produce vegetation index maps," says Darrel Williams, a forest ecologist and the Landsat project scientist at GSFC. "The vegetation index for a given area can vary from season to season and from year to year, especially for grasslands and scrublands. One can assess if there is a greater fuel load than normal." That can be a danger sign if conditions become drier.

Landsat maps reveal the boundaries between forests, grasslands, deserts and cities. But that's not enough. To a wildfire, not all forests are created equal. The virtual terrain needs to include more detail about the forest ecosystems than 30m pixels alone can provide.

Right: Superimposing the Landsat data onto a digital elevation map produces a 3-dimensional terrain. Using ecological principles, scientists can infer more about the vegetation from these hybrid maps than from Landsat data alone. Image credit: NASA's Goddard Space Flight Center.

Land managers need to know what is the dominant tree species? How dense is the canopy? Do the trees range in age, or is the stand more uniform? When was the last fire in the area? Some of these questions can be answered by inspecting the history of the area's vegetation. For others, scientists must rely on their extensive ecological knowledge and on fieldwork to find clever ways to infer such details from the satellite data.

"If two polygons on the map have similar spectral images, by knowing at what slope aspect and elevation and latitude that polygon is, our ecological knowledge tells us, well, it can't really be spruce, it has to be Douglas fir, for example," Ryan says.

This contour information is deduced by laying the Landsat data over the top of a digital topographic map produced by the USGS, creating a 3-dimensional landscape. Artificial structures like roads and buildings are also added, which allow the forest managers to see how a potential fire would affect the local communities.

Finally, the crucial ingredient is added: weather. Moisture and wind can make or break a forest fire. Since moisture lingers on the terrain, historical records of the area's weather must be woven into the model along with current conditions.

Like a forester's SimCity 3000^TM, the computer software integrates the vegetation, the terrain, and the weather into a "virtual forest" that simulates the interactions of all these variables.

Left: A firefighter setting fires? Sometimes the best way to avoid a big fire later is to set a small one now. Learning to use computers to predict the extent and intensity of these controlled burns will help keep them under control. Image courtesy National Interagency Fire Center in Boise, Idaho.

Trying out fuel treatments in silico holds great promise for helping forest managers mitigate fire risk, but the technology has only been in routine use for about 5 years and still has much room for improvement.

"It's an inexact and emerging science, obviously," Ryan says. "There still is a lot of research and development going on in fire behavior modeling and the effects of the different treatments."

New satellite data should also bring some improvements to the field. For example, researchers are learning how to better estimate forest canopy structure using the Multi-angle Imaging Spectro-Radiometer (MISR) on NASA's Terra satellite, launched in 1999. MISR provides 9 different views of the terrain below, each from a different angle.

"If you just look straight down at the trees, it's often hard to detect how thick the canopy is. But if you can look from various directions, the 3-dimensional nature of that canopy is much more clear," Running says. Canopy structure is an important factor, since it partly determines whether a ground fire will escalate into a raging crown fire.

Right: By looking forward and backward at several different angles as it flies, the MISR instrument aboard Terra has the potential to help scientists decipher the structure of the forest canopy -- important information for estimating if a ground fire will become a more dangerous crown fire. Image courtesy of the Jet Propulsion Laboratory.

In the future, this marriage of computer and satellite technologies should become a robust tool for helping reduce the risk posed by wildfires.

"As computers get faster and models get better, then prediction will get better, too," Finney says.

Wildland Fire Assessment System -- wide variety of information about current fire conditions in the US

Farsite.org -- home page for the FARSITE fire simulation software package, with information about the software and download links

Fire Management Tools -- links to other fire management software

Landsat 7 -- home page, from NASA's Goddard Space Flight Center

Landsat data and news -- from the U.S. Geological Survey's EROS Data Center

NASA's Terra Project -- home page for the satellite carrying the MISR instruent

Smoke Sentry in Space -- Science@NASA article: A new program provides firefighters with daily satellite images to aid in their efforts to control wildfires

Watching Wildfires from Space -- Science@NASA article: NASA's Earth Probe satellite is keeping an eye on smoke from wildfires raging across the Western US

Fuel Model and Fire Potential from Satellite and Surface Observations -- technical paper about the use of satellite technology to assess fire risk nationwide