At 25 additional locations, cameras are intended to detect smoke and fire. The technology watches around the clock, but people confirm the alarm – and calling 112 after spotting a fire yourself remains important.
What Croatia is actually expanding
Croatia’s environmental ministry has expanded the existing early wildfire-detection network by 50 surveillance cameras at 25 new locations. The authority places particular emphasis on protected areas, expressly including national and nature parks. The announcement does not provide a complete public map of the new positions or an investment figure.
The expansion is intended to enable three things earlier: detecting a fire, assessing how it might develop and informing emergency services more quickly. This is infrastructure for observing a situation. It does not extinguish a fire, prevent deliberate ignition or replace either firefighters or people who see smoke on the ground.
Two cameras watch while software searches for anomalies
OIV describes Fire Detect AI as a system comprising elevated panoramic cameras, data transmission, a data centre and a control-room application. The software automatically analyses the images: by day it primarily searches for smoke, and at night for fire. OIV states a possible detection range of up to ten kilometres in automatic mode and up to 25 kilometres in manual mode – provider figures, not a guarantee in every combination of weather and terrain.
According to the ministry, meteorological data is also incorporated into the risk assessment. Wind, temperature and other conditions help evaluate how a fire might spread. The camera therefore does not detect an abstract wildfire on a map, but supplies visual and positional information that is assessed together with weather and terrain.
The alarm is not passed on blindly
An important step lies between algorithm and response: OIV explains that operators additionally check a flagged situation. Confirmation means that a person has satisfied themselves that the software has probably detected a real fire. The responsible emergency services are then notified.
This principle matters because clouds, fog, dust or other visual phenomena can resemble a smoke pattern. Automatic detection is intended to focus attention; the operational decision remains within a control-room chain involving people and fire-service command.
Wildfire early detection
From a smoke pattern to a confirmed alarm
From the camera mast to the fire-service control room
The new expansion is intended to operate around the clock, every day of the year. Images and data are to be available to all county fire-monitoring centres. Incident commanders can thus receive not only an initial indication, but also continue observing developments remotely.
OIV also describes the ability to control cameras manually, simulate a fire front and provide information to teams in the field. Such functions are particularly useful in hard-to-reach karst and forest areas, but remain tools within incident command – not an independent decision by AI.
Why protected areas are specifically mentioned
National and nature parks combine sensitive ecosystems with large seasonal visitor numbers and, in some places, difficult access. An early view can save valuable minutes before a small plume of smoke becomes a broad fire front. The official publication nevertheless does not reveal exactly where the 25 sites are located.
According to the ministry, funding comes from the Competitiveness and Cohesion Programme 2021–2027. The measure is therefore part of climate adaptation and disaster protection. It should not be equated with comprehensive video surveillance of every forest path.
What travellers will see from the system – and what they will not
Holidaymakers will generally see neither the camera images nor an automated alert on their own phones. Public warnings will continue to come from the fire service, civil protection, police, local authorities and, where applicable, Croatia’s SRUUK warning system. The responsible traffic authorities additionally report road closures and diversions.
The technology is therefore primarily background infrastructure. Its value becomes visible when a control room locates a remote plume of smoke more quickly or assesses an already reported situation from several angles.
Anyone who sees smoke should not wait for AI
If you see smoke, flames or a newly started fire, the rule remains: keep a safe distance, determine your position as accurately as possible and call 112. Useful details include the road name, bay, hiking trail and prominent landmarks, and whether people or buildings are threatened. Do not drive into a smoke zone to obtain a better video.
The possibility that cameras may be present is no reason not to report your own observation. Several independent reports can help confirm a location. The right conclusion is therefore not “AI is watching”, but that Croatia is supplementing human attention with a permanent technical eye.
Evidence
Sources & date
Sources support the facts. Planning notes and recommendations are Golden Beach editorial assessments.
- Unaprjeđen sustav videonadzora za rano otkrivanje požara na 25 novih lokacijaCroatian Ministry of Environmental Protection and Green Transition · 24 August 2026, checked 26 August 2026number of cameras and locations, protected areas, weather data, funding, year-round operation and access for fire-service centres
- OIV Fire Detect AIOdašiljači i veze (OIV) · product and feature page, checked 26 August 2026system architecture, automatic and manual detection, provider range figures, human alarm confirmation and control-room workflow
