How AI helps put down forest fires

Abhishek Srivastava was aware of just how ruinous forest fires could be. But it was a meeting with the District Forest Officer at the Indian Institute of Technology (IIT) Indore in 2017 that sensitised him to the importance of early intervention in forest fire management. A slight delay in detection can turn a tiny spark into a raging inferno.

Satellite-based fire detection systems, a low-cost solution for detecting fires over a large geography, are now the backbone of every forest fire management plan, but they have their limitations. "By the time the fire is big enough for the images to be captured by the satellite, it's usually too late. Once the flames are that big, it spreads very quickly. In a few seconds, a large area gets engulfed," explains Srivastava, Professor, IIT Indore.

Satellite sensors cannot detect fires through thick clouds or smoke, and are ineffective in spotting small fires and surface fires in dense forests. As they capture the ground from a great height, they are often unable to spatially resolve the several forest fire clusters and may show them as one, creating confusion for the firefighting team. Depending on the sensor and processing time, the time delay in fire detection and relaying is 1 to 1.5 hours – time enough for a fire to blaze out of control.