AI created to accurately predict forest fires 16:04

Scientists at the University of Southern California (USC) have combined satellite imagery and artificial intelligence to create an AI model to accurately predict the spread of wildfires. The technology could be a breakthrough in wildfire fighting and emergency response. reports TechXplore.

The new model uses satellite data to track the spread of a wildfire in real time, then feeds that information into a sophisticated computer algorithm that can accurately predict the fire’s likely direction, intensity and growth rate.

The discovery comes amid California’s worsening wildfire season. Multiple fires are currently burning across the state, driven by a dangerous combination of wind, drought and extreme heat. The state’s largest wildfire is the Lake Fire, which has already burned more than 15,000 acres in Santa Barbara County.

The scientists began their project by collecting historical data on wildfires using high-resolution satellite imagery. By conducting a detailed analysis of the behavior of previous fire masses, the researchers determined how each fire started, spread, and was eventually brought under control. Their in-depth analysis revealed patterns influenced by a variety of factors, including weather, fuel (trees, shrubs, and grass), and terrain.

They then created an AI-based generative computer model, known as a conditional generative adversarial Wasserstein network, that models the impact of these factors on the development of wildfires over time. The model was trained to recognize patterns in satellite imagery associated with the spread of wildfires in its model.

Initially trained on simple simulated data under ideal conditions such as flat terrain and one-way winds, the system’s test results amazed its developers. The system performed excellently in tests on real wildfires in California.

Previously on the drone attached Robot arm to scare birds off wires.