In Portugal, Axians has developed EcoSentinel, a new, highly ambitious solution designed to warn of impending natural disasters such as forest fires and floods.
In Portugal, forest fires are becoming increasingly frequent and intense, often burning thousands of hectares in just a few hours. Flooding in Portugal is also becoming more severe and less predictable. Because of these changes, says Arlindo Ribeiro, Chief Architect Manager at Axians Portugal, “Our systems for combating climate change urgently need solutions that not only address but can also anticipate these increasingly dramatic events.”
One such solution is EcoSentinel, a new tool the VINCI Energies ICT brand has been developing since 2023. Currently at the POC (proof of concept) stage, this web application is designed to predict the locations and dates of extreme climate events – in this case, forest fires and floods.
AI-assisted prediction
EcoSentinel appears as a dashboard displaying various data linked to these events gathered by weather stations and sensors.
“We are using artificial intelligence and automatic learning for the prediction tasks,” explains Arlindo Ribeiro. “Our technology learns from every fire and refines our modelling. This makes it possible to react more quickly to extreme events, but also to reduce the amount of carbon dioxide released by fires and protect our forests, which are vital for absorbing carbon.”
“One of our objectives is to enable EcoSentinel to adapt to different environments and needs on a global scale”
EcoSentinel uses different prediction methodologies depending on whether it is looking at a forest fire or a flood. But the automatic learning models work in the same way. “We use clustering algorithms to divide the country into geographic zones. Once these zones are defined, we calculate the daily averages of climate indicators in each one, and this data is fed into the pre-trained model.
Once trained with historical data, the forest fire model is for example capable of predicting a fire based on data from a previously recorded day when the temperature was X, the drought index was Y and the fire index was Z, a combination of parameters that coincided with a forest fire at latitude X and longitude Y in zone A. In the future, if we see the same climate indicators, our model will attempt to predict the fire with the greatest possible precision.”
On a global scale
With the project still in development, Axians does not yet have all the data required to maximise the model’s precision. It takes time to train the system.
But EcoSentinel is unique, the only solution on the market that attempts to predict the location of a forest fire. Other tools are currently limited to displaying risk attribution maps or predicting the behaviour of forest fires to calculate their spread.
As Arlindo Ribeiro explains, “One of the objectives of our system is to adapt to different environments and needs on a global scale, whether in regions subject to brush fires, such as Australia, or in zones regularly affected by flooding, such as Southeast Asia.” As Chief Architect Manager at Axians Portugal, he works closely with Portuguese government agencies (national security agency, forest protection institute, fire-fighting associations, etc.).
01/16/2025
