European investigators, including Portuguese, are using artificial intelligence to revolutionize forest fires.
The European Meteorological Prediction Center has developed a new model called “Fire Probability”, which uses Machine Learning to analyze large data volumes and predict more accurately where fire can start.
In Portugal, artificial intelligence is placed at the service of fire prevention with the Developed by researchers from Nova FCT, which monitors the cleaning of fuel management ranges.
Predict the exact location where you can trigger a fire
A It is revolutionizing the forecast of forest fires, allowing a more accurate assessment of risk, not only based on weather conditions, but also on factors such as the amount of vegetation available to burn and the potential sources of ignition.
The model developed by scientists from It promises to more rigorously identify the places with the most potential to burn.
It is more than predicting danger: it is pointing the exact place where you can derive a fire.
“In a way, it is a radical change in the forecast of fires, because we are suggesting where a danger of fire may occur, to say: This is where we think the fires will actually occur. So it is a great advance,” the climate scientist of ECMWFJOE McNorton told Reuters, told Reuters to Reuters.
An intelligent model to predict fires
O developed a new designated model Probability of firewhich uses automatic learning (Machine Learning) to analyze large data volumes and predict more accurately where fire can start.
Current forecasts are based on a system that considers four variables: temperature, wind, precipitation and moisture. However, this method has several limitations, explained Joe McNorton.
“Historically, to predict forest fires, we use what we call the weather index. It is a simple model, based on physics, which uses four meteorological variables – temperature, wind, precipitation and humidity – to predict, if a fire happens, what will be its intensity. sources of ignition, “said the scientist.
By integrating data on the present vegetation – fuel – and possible sources of ignition, such as lightning or human activity, the ECMWF model can go beyond simple danger identification – It indicates the actual probability of a fire starts in a specific location.
“With all this, we can predict not only the danger of fire, but the actual probability of a fire starting in a specific place,” said McNorton.
New model predicted the start of fire in Los Angeles
One of the most striking recent examples occurred in They killed 28 people and destroyed about 16,000 structures.
ECMWF’s main scientist, Francesca di Giuseppe, explained how the new model managed to greatly predict the most affected areas.
“For example, in a recent case in Los Angeles, the fire began even in the transition zone between the urban area and the forest, which made the situation particularly severe. This is because the previous season had been marked by very wet conditions, which led to abundant growth of vegetation – that is, fuel – which ended up during the fire.”.
This new method, based on the probability of fire, “Incorporates the ‘memory’ of this abundance of fuel in its formulation, which allowed us to identify the regions that could be more severely affected – unlike models that consider only the weather conditions. Therefore, in this case, our forecast was much more accurate and was able to indicate the exact place, very close to Los Angeles, where the fire eventually triggered,” he said.
The weather conditions contributed to the worsening situation: unusual rains in the previous period made excess vegetation grow, which became highly flammable when winter arrived with abnormally dry temperatures.
Monitor the cleaning of fuel management ranges
In Portugal, artificial intelligence is also placed at the service of fire prevention.
The project coordinated by bet on continuous surveillance of Fuel Management ranges (FGCI), Using artificial intelligence and automatic learning algorithms to detect failures and monitor the evolution of fuel material. The platform automatically assesses compliance with the legislation that requires the cleaning of these zones to lock fire spread.
One of the main innovations is the mobile application for municipalities and population report, consult and follow the state of cleaning FGCI in municipalities as Almada, Mação and Santarém, with the ambition to widen the coverage to the entire national territory.
The system integrates satellite data, namely vegetation index standards, and allows you to collect updated and reliable information on the state of the lanes, helping to plan cleaning actions and reinforcing supervision.
The hottest year ever aggravated the fires
In 2024, the record levels of greenhouse gases contributed to , with extreme weather events affecting various continents. Forest fires forced the move of 800,000 people, the highest number since the beginning of systematic data collection in 2008.
