Geosfire brings round-the-clock satellite fire detection to Cyprus

A satellite-based system using machine learning to detect wildfires across Cyprus 24 hours a day has been presented in Limassol, seeking to give first responders earlier and more reliable warnings of possible outbreaks. The Geosfire project, funded by TechIsland, was presented on Thursday by Anna Zenonos, a researcher at the Cyprus Institute’s Climate and Atmosphere Research Centre, as part of the TechIsland Wildfire Initiative. Geosfire stands for Geospatial Earth Observation System for Fire Intelligence and Real-time Evaluation and is being developed specifically for Cyprus. Explaining its mission, Zenonos said that “this project aims to develop a satellite-based, machine-learning fire detection and alert system tailored to the geographic and climatic conditions of Cyprus, while addressing the operational needs of first responders”. Unlike traditional monitoring, which largely depends on someone seeing smoke or flames, Geosfire uses satellite data to monitor locations across the island. Zenonos noted that the system “enables 24/7 fire detection across all locations” and “overcomes limitations of traditional visual monitoring, which is restricted to daytime and populated areas”. This could be particularly important in remote or sparsely populated parts of Cyprus, where a fire may develop without immediately being seen by residents, drivers or patrols. Satellite monitoring also allows detection work to continue during the night, when traditional visual observation becomes considerably more difficult. Turning to the project’s technical approach, Zenonos explained that Geosfire “utilizes data from recently launched, state-of-the-art satellite systems” and “applies advanced machine learning algorithms for detection and alerting”. The satellite data is processed by the algorithms to identify possible outbreaks and generate alerts. However, the project goes beyond detection alone, with the information also feeding into a dedicated visualisation platform. According to Zenonos, the system “provides a visualization platform for real-time monitoring, situational awareness, and decision support”, bringing complex satellite information into a format that can be more easily understood and used by first responders. The platform displays detected incidents on an interactive map of Cyprus, allowing users to see where a possible fire has been identified and examine the available information. Details shown through the dashboard include the location and precise coordinates of the incident, the date and time of detection, its classification and the algorithm responsible for identifying it. Where available, the platform can also display information on the fire’s intensity. A demonstration included in the presentation showed how an incident near Limassol appeared on the map with its coordinates and timestamp, allowing users to open the report and view the available data. In this way, Geosfire is intended not only to issue an initial alert but also to support the decisions that follow, giving responders a clearer picture of the location and nature of a possible outbreak. This information could be particularly valuable during the early stages of a wildfire, when delays in confirming its location may affect how quickly firefighting and emergency resources are mobilised. The project also has a wider research dimension. Referring to the technology’s use elsewhere in Europe, Zenonos noted that “few research teams in Europe currently leverage such satellite data for fire detection”. Moreover, she explained that “no existing solution offers an operational near–real-time detection system based on these data sources”, emphasising the gap that the Cyprus-based project is seeking to address. Geosfire therefore aims to move beyond research and develop into what Zenonos described as “a scalable early warning system to enhance wildfire response and mitigation in Cyprus”. By combining continuous satellite observation, machine-learning detection and a real-time visualisation platform, the project seeks to identify possible fires beyond the limitations of daylight, populated areas and traditional ground-based monitoring.