An Innovative Approach to Predicting Forest Fires
Winners of the Junior Academy Innovation Challenge Fall 2022: “Forestry for a Sustainable Future”
Sponsored by Royal Swedish Academy of Engineering Sciences (IVA)
Prolonged droughts, caused by climate change, have amplified the risks of forest fires around the globe– making blazes bigger, more frequent, and more intense.
These fires devastate vast swathes of forests and often spread into residential areas, threatening lives and housing. Research by the University of Maryland suggests that fires cause forests to lose 3 million more hectares annually than in 2001. Furthermore, the UN Environment Program estimates that by 2100, the number of forest fires will increase by 50%.
The team Intelligent Forest — Chinmay R. (India, Team Lead), Rohan S. (India), and Soumik P. (India) — worked under the guidance of their mentor Malarvizhi Arulraj (United States) to tackle this critical issue as part of the “Forestry for a Sustainable Future” Fall 2022 Junior Academy Challenge, sponsored by the Royal Swedish Academy of Engineering Sciences (IVA). Intelligent Forest bested the field among 175 competitors. Their innovative method to predict the risk of fire helped them to win.
“It was great taking on real world problems and using our intellect to solve them. I learned various things throughout the course of the challenge such as AI, weather patterns, machine learning applications and much more,” says Rohan. “We worked hard as a team and came up with a solution in the end together.”
Understanding Forest Fires
Forest fires can be triggered by natural factors, such as lightning, or by human factors, such as the careless dropping of a cigarette or the lighting of an unnecessary fire in severe drought conditions. Crown fires burn the entire length of the trees while surface fires only scorch dried leaves and grass.
In some cases, fire can rage under the ground. As the team discovered over the course of their research, climatic conditions play a critical role– the hotter and drier the weather, the more destructive the fire is likely to be.
Finding ways to mitigate the impact of these now-frequent infernos required hard work, but the team members worked collaboratively to achieve results.
“There were times when I was uncertain as to whether we would even reach the end, but here we are,” says Soumik. “It was a fun experience working with my team members, and I had the opportunity to add and develop my skills. My main contribution was helping with the research side of things and suggesting ideas and edits.”
Utilizing Artificial Intelligence
With support from their mentor, the students decided to focus on harnessing the power of Artificial Intelligence (AI) to analyze forest and temperature data, in the hope that it would be possible to predict the risk of fires.
“I was impressed by the plans and ideas the team put together and was absolutely delighted to mentor the team,” says their mentor, Malarvizhi. “They chose a problem and approach that was hard and challenging. Especially, finding the best dataset and creating working machine-learning algorithms needs a lot of effort.”
Using data on fire alerts and meteorological information (minimal and maximal temperatures, rainfall, solar radiation and daily evaporation) collected in the Brisbane area in Australia between 2012 and 2022, the team tested two different AI approaches: Decision Tree and Random Forest.
The Results
The goal was to create four categories: no risk, low risk, medium risk or high risk of fire. The results provided the proof-of-concept the team expected. With the Decision Tree approach, they were able to predict fire risk with 70% accuracy, while the accuracy was 79% using the Random Forest approach.
These findings demonstrated that with the help of AI, it is possible to predict the risk of forest fires with 70–80% accuracy, which, in turn, allows for increased preparedness and limited impact.
“The project was a great learning experience for me,” says Team Lead Chinmay. “I had taken Artificial Intelligence as a subject in high school and this project taught me how I could apply what I had learned in a real-life situation.”
