A new AI wildfire spread prediction tool from New York University aims to tell fire commanders where a blaze is heading within minutes rather than days. The NYU Fire Research Group calls it a “digital twin” of an active wildfire: a system that keeps updating its forecast as fresh data arrives from satellites, weather feeds, drones, ground sensors and crews in the field. In early tests on three US wildfires, one forecast matched the observed burned area by 94.6 percent.
The lead team is Research Associate Professor Prabodh Panindre, Research Scientist Anandu Nair and Global Network Professor Sunil Kumar. Panindre and Kumar hold appointments at both NYU Tandon School of Engineering and NYU Abu Dhabi. They say the system is built to solve one of the hardest problems in wildfire response: giving incident commanders timely, explainable forecasts while conditions keep changing.
How the AI wildfire spread prediction system works
The platform shows a map of the United States. A user picks a fire location, sets the size of the area to forecast and chooses a time horizon of up to 12 hours, with updates at set intervals. The map then draws the predicted perimeter of the fire.
Under the hood, a physics-informed fire-spread model is paired with an AI layer. The AI continually solves the equations that govern fire propagation, combustion, heat transfer, fuel type and properties, wind, slope, weather and known fire-transition thresholds. It uses live conditions instead of relying only on historical fire data, as many machine-learning tools do.
“Our system is not built from training data. It is grounded in physics,” Kumar said. “We solve the fundamental equations of how fire burns and spreads using AI, instead of solving them only in the conventional way.”
Why combine physics and AI?
According to the researchers, physics-based simulators can be accurate but are computationally heavy and can take hours to update. Machine-learning tools are faster but only as good as their training data, and they may struggle in conditions they have not seen before. The digital twin is designed to take the strengths of both, syncing with new observations to produce updated forecasts in near real time.
A chatbot-style assistant for commanders
The tool also has an AI-enabled plain-language interface, similar to a chatbot. Users can ask direct questions, such as what happens to the eastern flank if the wind shifts north, where to cut a firebreak or which zones to evacuate first.
It recommends containment strategies, evacuation plans and training scenarios. It cites fire department procedures and published research so commanders can review its reasoning.
Test results on real wildfires
The team tested the AI wildfire spread prediction system on three active wildfires: the Grasshopper Fire in Oregon, the Widemouth 2 Fire in Utah and the Sacaton Fire in New Mexico.
In one Grasshopper Fire test, predicted and observed fire areas reached a 94.6 percent match on a standard measure of spatial overlap. Results varied in other tests. In one Grasshopper Fire forecast, agreement fell to 72.4 percent after a longer gap between satellite observations.
The researchers said the shape of the forecasts impressed them most. Rather than treating a fire as a rough circle, the model reproduced the irregular, jagged edge of the real wildfire. Panindre explained why that matters: the exact boundary decides where crews go and which neighbourhoods are evacuated. “A small difference on a map may not look very significant,” he said, “but it may be very significant in real life, because there may be a critical structure there.”
Still a research prototype
The system has not been deployed as an operational tool on active wildfires. Panindre said the team wants more testing with fire departments and federal agencies before making it public. The researchers have filed a provisional US patent through NYU.
As a writer, I feel the explainability piece is the most important part here. A forecast that shows its reasoning is far easier for a commander to trust under pressure than a black box, but the drop to 72.4 percent shows how much the system still depends on fresh satellite data.
FAQ
What is NYU’s Digital Fire Twin?
It is a research system from the NYU Fire Research Group that pairs a physics-informed fire-spread model with AI to forecast wildfire spread in near real time, updating as new data arrives.
How accurate is the AI wildfire spread prediction tool?
In one test on Oregon’s Grasshopper Fire, it achieved a 94.6 percent spatial overlap with the observed fire area. Another forecast on the same fire reached 72.4 percent after a longer gap between satellite observations.
Can fire departments use it now?
Not yet. It remains a research prototype, and the team wants further testing with fire departments and federal agencies before releasing it publicly.