Can Big Data Analytics and Machine Learning Help Contain Wildfire Next Time?

Researchers are developing tools to predict wildfire leveraging technology, real-time data to prevent devastation

In the last three years alone, wildfires have devastated forests in Australia, Brazil and the U.S. With climate change affecting rainfall and creating a likely scenario for forest fires around the world, researchers are turning to technology to contain such devastating situations.

Three major wildfires in three countries alone have killed over a billion animals and close to 50 million acres of forest. Australian bushfire, Amazon rainforest fire and California fire are the three largest recorded in history. And such incidents are likely to occur again in the coming years as intense heatwaves and deforestation continue. Thus, scientists are turning to computer prediction to contain before it spreads.

Fire Prediction

In the U.S., after a wildfire burned forest just outside Glacier National Park in Montana, a fire safety team was deployed in a town nearby as a precautionary measure. The town and forest were divided by Hungry Horse Reservoir but soon thunderstorm stoked the fire and it was able to spread to the northern tip of the lake.

California fire
A helicopter drops water on the wildfire Wikimedia commons

The fire-safety team saw this coming and could manage the fire even before it spread. They could do it thanks to a fire prediction model that simulates the situation to stop such incidents from happening again. Although fire is essential to keep a forest ecology thriving, sometimes it gets out of hand and the destructive force wreaks havoc.

Mark Finney, a U.S. Forest Service researcher realized that and created a fire prediction program FarSite in 1992. He used that software to predict the 2003 fires. "That wasn't a forecast that it would happen. It was a scenario that showed what could happen," Finney told Cnet.

While Finney's software wasn't able to predict future fires, current ones are using NASA's wind prediction, weather data to simulate wildfire situation. Apart from that, drones, infrared cameras and satellites are being used to look for fire. There are already software and apps that can leverage the data.

Team Awareness Kit or TAK is one mobile app, developed by the U.S. Airforce, that works as a data gathering and sharing platform for first responders. It can map the growth of fire while also provide real-time insight on wildfire behavior.

Big Data Analytics

But all the above mentioned technologies are used when the inferno is raging to stop the spread. But what about the technology to predict fire ahead of its time and stop the fire even before it happened? While predicting wildfire is difficult considering human-caused infernos, there is software that can leverage big data to predict a wildfire ahead of time so that it can be stopped.

Machine learning and big data analytics are already being used to carry out such predictions and prepare firefighters. Analysts gather data from fire finders, satellite imaging, historical data and weather predictions to create algorithms that would forecast.

However, the existing prediction models aren't exactly accurate. The algorithm fire models are derived from the movement of waves. But as the past incidents suggest, a forest fire can be extreme too. In California's 2018 Camp Fire, it spread over a football field just in a second. That essentially creates a problem in predicting fires in extreme cases.

Hence, the need is of prediction software that can also crunch real-time data to control and stop a fire. California's Wifire lab is developing a program that can crunch data in real-time about wildfire. The program is run on San Diego's Super Computer Center in partnership with the University of California to feed into exiting fire-modeling programs like FarSite.

Park Williams, an associate research professor at Lamont-Doherty Earth Observatory in New York, is developing another tool that will be able to understand the nature of wildfire in a region based on past incidents. His tool will compile data from thousands of wildfires that happened in the last 30 to 40 years and link that to the change in vegetation, weather patterns and habitations. The resulting program will be able to project large fires and the impact on the vegetation and lives.