Wildfire prediction and detection: what each one brings and why both are needed
12 November 2026
In wildfire management, detection and prediction answer different questions and support decisions taken at different moments. Detection reacts to reduce the damage of each active fire. Wildfire prediction aims for fewer ignitions and fewer fires escaping initial attack. They are complementary layers of the same system.
At SilvIA Earth we work on predicting the probability of ignition, using artificial intelligence on satellite, weather, terrain and human-activity data to produce wildfire risk maps and action plans. This article explains our approach: why we chose that focus and why detection remains an essential part of any wildfire defence system.
What is the difference between detecting and predicting a wildfire?
Detecting a fire means identifying an active fire as soon as possible after ignition, locating it precisely and passing the alert to whoever must respond. Predicting means estimating, before there is any fire, the probability that an ignition will occur in a given area and period.
Detection works with observable facts: a column of smoke, a thermal anomaly picked up by a satellite or a call to the 112 emergency number. Prediction works with conditions: the state of the vegetation, the weather forecast, the terrain, human activity and the patterns that have historically preceded fires in each area.
| Aspect | Prediction | Detection |
|---|---|---|
| Question it answers | Where is a fire most likely to start in the coming days? | Where is there a fire right now? |
| When it acts | Before ignition | After ignition |
| Time horizon | Days | Minutes to hours |
| Main data | Weather, vegetation condition, terrain, fire history, human activity | Thermal and visible imagery, cameras, lookouts, public reports |
| Output | Map of ignition probability by area | Geolocated alert of an active fire |
| Decision it enables |
|
|
What early wildfire detection technologies are there?
Early wildfire detection is one of the factors that most shape the outcome of a fire. A fire spotted in its first minutes can be controlled with initial-attack resources, but the same fire, detected an hour later and with strong wind, can require a deployment lasting several days.
That is why response systems combine several layers: lookout towers, camera networks increasingly backed by computer vision, drones, aircraft, satellites and reports from the public. Among satellite systems, NASA's FIRMS distributes active-fire data from the MODIS and VIIRS sensors within about three hours of observation, at a resolution of 375 metres in the case of VIIRS (NASA, n.d.).
None of these layers is superfluous. The shorter response times achieved over recent decades are one of the great achievements of firefighting services. However, detection has a limit that does not depend on technology: it can only act once the fire already exists. On days of extreme conditions, a fire can outgrow initial-attack capacity before resources arrive, however early the alert was.
Can wildfires be predicted?
Yes, but it is worth being precise about what is predicted. It is not possible to anticipate a specific fire, that is, that a fire will start at an exact time on a particular hillside. What can be estimated with the available data is the probability of ignition in each part of the landscape over a given period and, therefore, which areas concentrate more risk than others.
Europe has had public fire danger forecasting systems for years. In 2007, the network of the European Forest Fire Information System (EFFIS) adopted the Canadian Fire Weather Index (FWI) as the harmonised method for assessing fire danger across Europe, with a scale of several classes from very low to extreme, based on weather forecasts (European Forest Fire Information System [EFFIS], n.d.).
How is wildfire risk predicted with artificial intelligence and satellites?
At SilvIA Earth we have developed a model that produces an ignition risk map at 1 km resolution, updated every day and covering the next 10 days. The model shows where a fire larger than 5 hectares is most likely to start, since those are the fires that call for planning. That resolution reveals risk differences within a single municipality or district, the scale at which many management decisions are planned, and the multi-day horizon leaves room to organise prevention work. The SilvIA Earth assistant answers questions about the risk and the data for each area, and proposes action plans tailored to the forecast risk and to the conditions of each landscape.
In internal evaluations over Andalusia, Castilla-La Mancha and Castilla y León (seasons 2017 to 2024, each with a model trained only on the previous years), the 10% of the land that the model flagged as most likely accounted on average for 73% of the burned area.
Why does wildfire prevention have to come before ignition?
Because once started, a fire in extreme conditions can overwhelm any response capacity. What reduces damage most is acting beforehand: managing fuel, watching where the probability is highest and restricting the activities that can cause an ignition.
Several experts consulted by the Science Media Centre Spain agree that simply adding firefighting resources does not solve the underlying problem and that the effort should shift towards preventive land management (Science Media Centre España, 2026).
What the latest seasons showed
According to the annual report of the European Commission's Joint Research Centre (JRC), based on EFFIS data, more than one million hectares burned in the European Union in 2025, the worst season since EFFIS records began (Sedano et al., 2026).
Much of the damage was concentrated in a very short time. During the heatwave of the first three weeks of August, 22 large, almost simultaneous fires in Spain and Portugal burned more than 460,000 hectares, 43% of all the area burned in the EU that year (Joint Research Centre [JRC], 2026).
In Spain, EFFIS mapped 1,359 fires that burned 401,266 hectares, four times the average of the last fifteen years. 85% of that area burned in August, after a quiet start to the year. Spain was the hardest-hit country: together with Portugal and Romania it accounted for around 75% of the area burned in the EU (Sedano et al., 2026).
The 2026 season has been less extensive than the previous one, but it is again well above normal. According to provisional EFFIS data reported by Europa Press, between 1 January and 4 November 301,681 hectares burned in Spain, almost 43% of all the area burned in the European Union (Busto, 2026). Spain's Ministry for the Ecological Transition, with data up to 9 November, puts the affected forest area at 280,276 hectares, 120.6% above the average of the last decade, and counts 47 large fires, against an average of 29 over the last decade (Ministerio para la Transición Ecológica y el Reto Demográfico [MITECO], 2026).
The trend shows no sign of easing. In its 2022 report, the United Nations Environment Programme projects a global increase in extreme fires of up to 14% by 2030 and 50% by the end of the century (United Nations Environment Programme [UNEP], 2022). The same body notes that direct response to fires usually takes more than half of related spending, while planning receives less than 1%, and recommends devoting two thirds of investment to planning, prevention, preparedness and recovery (UNEP, 2022).
Prediction does not replace forest management, but it makes it more efficient: it shows where and when to concentrate preventive resources, which are always limited.
What decisions does a wildfire risk map support?
A wildfire risk map is valuable to the extent that it changes concrete decisions:
- Targeted surveillance. Reinforcing lookouts, patrol routes or reconnaissance flights in the areas and on the days of highest probability. It is the most direct link between prediction and detection.
- Pre-positioning of resources. Placing crews and initial-attack resources close to where the risk is highest, to cut arrival time if an ignition occurs.
- Fuel management. Prioritising brush clearing, thinning, grazing or prescribed burns in the areas that contribute most to the risk.
- Restricting high-risk activities. Adjusting the schedule of work with machinery, agricultural burning or other activities that could start a fire. This matters all the more because, according to WWF, up to 95% of fires in Spain are caused by people (WWF España, 2026).
- Planning in special-use areas. Organising work, access or activities in natural parks, large estates or military land according to the forecast risk.
The usefulness of prediction is not limited to public authorities. Any organisation responsible for forest land, public or private, has to decide every week where to watch, where to work and when it is better to postpone an activity. A risk analysis of your land lets you make those decisions with data.
What are the limitations of wildfire prediction?
Prediction works with probabilities, not certainties. A high-probability area may record no fire at all, and a fire can start in an area of moderate probability. Model performance varies between regions, depending on the fire regime and the quality of the available data. And human intent introduces a component that no model can fully anticipate.
That is why a risk map does not replace the judgement of technical staff: it is a decision-support tool. At SilvIA Earth we evaluate our models on seasons that were not used to train them, and we consider it essential to complete that evaluation with the teams who know the ground.
Detection and prediction: two layers of the same system
Prediction shows where to look and where to prepare over the coming days. Detection confirms when and where an ignition has occurred and triggers the response. A system that only detects always arrives after the fire; one that only predicts lacks confirmation on the ground. Combining both is what makes it possible to move from a reactive model to one that also anticipates.
That is our focus at SilvIA Earth: adding anticipation to existing systems, so that preventive resources are concentrated where the probability of ignition is highest, reducing risk and damage.
If your organisation manages forest land and wants to know how ignition risk is distributed across your area, the SilvIA Earth team can show you the tool applied to your own land. Request a meeting.
Frequently asked questions
Does wildfire prediction replace detection?
No. Prediction acts before ignition and detection afterwards, and each strengthens the other. SilvIA Earth is designed to complement existing detection systems, directing surveillance towards the areas with the highest probability of ignition.
Where can I check today's wildfire risk?
For weather-driven fire danger, Spain's state meteorological agency (AEMET) publishes daily forecast maps for Spain and EFFIS publishes them for all of Europe; many Spanish regions also have their own indices. These maps show how favourable conditions are for fire. SilvIA Earth adds ignition risk at 1 km, combining weather with vegetation, terrain, human activity and fire history.
How far ahead does SilvIA Earth predict wildfire risk?
Every day, our model shows where a fire larger than 5 hectares is most likely to start over the next 10 days.
How is SilvIA Earth different from the FWI index?
The FWI measures the danger associated with weather conditions and fuel moisture. SilvIA Earth estimates ignition risk in each part of the landscape by combining weather, vegetation condition, terrain, the human factor and fire history. They are complementary tools.
Who can use SilvIA Earth?
Any organisation that has to decide where and when to act on forest land. On one side, companies and landowners with land or assets exposed to fire, such as owners of forest estates or operators of critical infrastructure, who need to plan surveillance, work and activities according to the forecast risk. On the other, public authorities: fire prevention and firefighting services, civil protection and managers of protected natural areas.
Why does prediction matter if most fires are caused by people?
Because a human-caused fire can become a large fire if conditions on the ground allow it. The SilvIA Earth model takes human-activity variables into account, such as settlements and the wildland-urban interface, so it can show where that activity is most likely to cause an ignition. Knowing this helps concentrate surveillance and prevention where the consequences could be most serious.
References
- Busto, M. (2026, November 5). Un total de 301.681 hectáreas ardieron en España hasta el 4 de noviembre, el 42,9% de UE, según Copernicus. Ifomo Noticias. https://www.ifomo.es/cantabria/cantabria-total-301681-hectareas-ardieron-espana-4-noviembre-429-ue-copernicus/20261105131941363729.html
- European Forest Fire Information System. (n.d.). Fire danger forecast. Copernicus Emergency Management Service. Retrieved November 12, 2026, from https://forest-fire.emergency.copernicus.eu/about-effis/technical-background/fire-danger-forecast
- Joint Research Centre. (2026, March 31). 2025 was EU’s most destructive wildfire season on record. European Commission. https://joint-research-centre.ec.europa.eu/jrc-news-and-updates/2025-was-eus-most-destructive-wildfire-season-record-2026-03-31_en
- Ministerio para la Transición Ecológica y el Reto Demográfico. (2026). Informe MITECO-CCAA-EFFIS 2026: Situación de los incendios forestales en España [Provisional estimate, data from January 1 to November 9, 2026]. https://www.miteco.gob.es/content/dam/miteco/es/biodiversidad/temas/incendios-forestales/estimacion_provisional.html
- NASA. (n.d.). Fire Information for Resource Management System (FIRMS). Retrieved November 12, 2026, from https://firms.modaps.eosdis.nasa.gov
- Science Media Centre España. (2026, March 31). Los incendios forestales en 2025 quemaron más de un millón de hectáreas de la Unión Europea, casi la mitad, en la península ibérica. https://sciencemediacentre.es/los-incendios-forestales-en-2025-quemaron-mas-de-un-millon-de-hectareas-de-la-union-europea-casi-la
- Sedano, F., Maianti, P., Boca, R., Suarez-Moreno, M., Broglia, M., de Rigo, D., Roglia, E., Branco, A., San-Miguel-Ayanz, J., Durrant, T., Ferrari, D., Scionti, N., Nikonovas, T., Oom, D., & Artés Vivancos, T. (2026). Advance report on forest fires in Europe, Middle East and North Africa 2025. Publications Office of the European Union. https://doi.org/10.2760/3859043
- United Nations Environment Programme. (2022, February 23). Number of wildfires to rise by 50% by 2100 and governments are not prepared, experts warn [Press release]. https://www.unep.org/news-and-stories/press-release/number-wildfires-rise-50-cent-2100-and-governments-are-not-prepared
- WWF España. (2026). Incendios extremos: El reto de adaptar el territorio. https://wwfes.awsassets.panda.org/downloads/informe-2026-wwf-incendios-extremos--el-reto-de-adaptar-el-territorio.pdf