Practical Test of a Forest Fire Early Warning System
The article that follows is a translation of an article which first appeared in AFZ-DerWald (Issue 11/2023). The German version of the article can be accessed here. Image credits belong to Prof. Dr. Axel Göttlein.
This article explores the practical test of our Silvanet forest fire ultra-early warning system that utilizes advanced sensor technology to detect wildfires at their earliest stages. By combining gas sensors capable of detecting changing gas compositions and artificial intelligence, Silvanet offers preventive and post-extinguishment monitoring capabilities.
The article highlights a controlled burning experiment conducted in a representative pine forest area in Nuremberg, Bavaria, and describes the installation of a network of sensors that communicate through a wireless infrastructure. The system connects to a cloud platform, enabling real-time data consolidation, monitoring, and analysis. Additionally, in this particular test, an automated drone equipped with thermal imaging enhances the system's capabilities.
The article discusses the activation of the sensors during the experiment, the influence of wind direction and speed on detection, and the successful response of fire brigades. It concludes by emphasizing the system's practicality, potential for monitoring high-risk areas, and the importance of early detection in mitigating wildfires.
Practical Test of a Forest Fire Early Warning System
The sooner a forest fire is detected, the better the chances of containing it within a limited area and preventing it from becoming uncontrollable. Modern sensor technology, along with a comprehensive wireless connectivity infrastructure for the sensor units, offers the possibility of detecting smoke development both preventively and in areas that have already been extinguished, and reporting it directly to the command center.
With the increasing summer droughts resulting from climate change, the risk of forest fires is also on the rise. In Bavaria, the Alpine region and the pine forests of Central Franconia are particularly vulnerable in this regard . To create a test area representative of the pine forests in the Nuremberg region, a controlled burning experiment was conducted.
The Bavarian State Forestry, specifically the Nuremberg and Allersberg forest districts, initially identified 14 potentially suitable areas for the experiment. After overlaying these areas with publicly available data on water protection zones, habitat mapping, etc., five areas were shortlisted.
After evaluation by the forestry administration and the local nature conservation authority, a 70-year-old pine stand with limited natural regeneration and no significant lichen presence was selected as the test area. The stand is located on dry sands of the middle Keuper formation in the Brunnauer Forest near the municipality of Allersberg.
The designated burning area, measuring 30 m x 30 m, was surrounded by an approximately 1 m wide milled strip to create a barrier against the underground spread of the fire. On the eve of the experiment, the milled strip and an adjacent safety strip, up to 10 m wide, were thoroughly watered (Fig. 1).
For the irrigation of the safety strip and extinguishing the fire, two reservoirs with a capacity of 4,000 liters each were set up along the forest road by the participating fire departments. These reservoirs were filled with water using tanker trucks and manure spreaders. The firefighting operation was supplied with water from these reservoirs. We would like to express our gratitude to Fire Chief Inspector Egbert Petz for his excellent organization of the firefighting operation and to the 50 volunteers from the participating fire departments and the involved farmers for their assistance.
"The fire alarm early warning system has proven its practicality." - Axel Göttlein.
The Early Warning System
A network of gas sensors was installed, capable of detecting multiple smoke gases (hydrogen, carbon monoxide, and volatile carbon compounds) and reliably identifying forest fires through built-in artificial intelligence (AI) . With their design and energy-autonomous, battery-free operation using integrated solar panels, they can be easily, reliably, and inconspicuously deployed in the forest.
They communicate through a solar-powered mesh network infrastructure (LoRa) and ultimately connect to the public safety agency (BOS) portal via 4G/LTE or satellite communication (BOS: authorities and agencies with security responsibilities, in this case, the fire department). The BOS portal for the fire department is used to consolidate, monitor, and analyze the data provided by the sensor network. Additional information such as location maps, path information, fire hydrants, local weather stations, or cameras is also available on the BOS portal. They enable emergency responders to prepare for optimal operations and initiate firefighting measures quickly and in a coordinated manner.
To enable an early comprehensive assessment of the operational situation, the early warning system has been enhanced with an automated drone, which can be stationed at a fire station or control center during standard operation. The drone is provided with the GPS position of the triggering sensor, and it autonomously flies to the designated location. The flight can be initiated and monitored by a fire department control center. Real-time images from the drone's optical and thermal cameras are transmitted to the control center. In the future, additional sensors such as soil moisture or tree growth measurement can be easily integrated into the early warning network. Currently, these valuable data need to be manually collected.
With increasing summer drought, the risk of forest fires also rises.
A controlled burning experiment was conducted to create a test area representative of the pine forests in the Nuremberg area.
Four minutes after igniting the fire, the sensors sent out a preliminary alarm.
The fire alarm early warning system proved its practicality.
Just four minutes after the fire was ignited, sensors No. 9 (located directly at the fire site) and No. 3 (east of the fire site) sent out a preliminary alarm, followed by a fire alert three minutes later. Shortly after, more distant sensors also reported the occurrence of the fire (see timeline in Fig. 2).
The drone equipped with a thermal imaging camera was activated and automatically flew from its stationed position to the fire location upon receiving the fire alert. Even during its approach, the drone detected that it was not a false alarm but a real fire based on the heat radiation (Fig. 3).
During the controlled fire, the wind was blowing from a range between 270 and 360 degrees, with an average of 319 degrees, which corresponds closely to the northwest direction. The outdoor wind speed ranged from 3.3 to 5.9 m/s, with an average of 4.5 m/s. These data were obtained from the nearest weather station of the German Weather Service (DWD) in Roth (station number 4280, 10-minute values from https://opendata.dwd.de), located approximately 7.5 km southwest of the experimental fire site. The sequence of activation of the fire sensors can be explained by the wind direction and their position relative to the fire site. After sensors No. 9 and No. 3, which were located near the fire site, sensors No. 5, 7, and 6 (within a distance of less than 33 m from the fire site) and sensor No. 4, located in the eastern direction within the smoke plume (56 m from the fire site), reported the fire event with a delay of 15 to 25 minutes.
The last sensor to activate was sensor No. 1, placed 29 m north of the fire site, with a delay of approximately one hour. However, sensor No. 8, located 40 m downwind, and sensor No. 2, placed 74 m to the north, did not detect the fire event. Although the attending fire brigades extinguished the fire extensively, with no smoke rising from the ground when they left, sensor No. 5 reported the presence of smoke again the following day (Fig. 2). At that time, the wind was coming from varying directions with a mean outdoor wind speed of 2.7 m/s, ranging from west to northeast. This explains why sensor No. 5, located south of the fire site, detected the smoke development instead of sensor No. 9, directly installed at the fire site. As a result, the fire brigades from Allersberg and Lampersdorf mobilized eight personnel to extinguish the remaining reactivated smoldering nest.
The fire alarm early warning system has proven its practicality. From a financial and logistical perspective, it is hardly feasible or sensible to comprehensively monitor vast fire-prone areas such as the entire Nuremberg Reichswald forest. However, considering that the majority of fires are caused by human activities , it would be worth considering prioritizing heavily frequented areas by visitors for monitoring using this system. Furthermore, the controlled fire experiment demonstrated that despite intensive and diligent firefighting efforts, smoldering nests may remain and pose a renewed threat even days later. Therefore, on extinguished fire sites, the early warning system can fully utilize its capabilities to monitor the area in detail over an extended period. This significantly reduces the personnel and time required for monitoring a fire site, thus relieving the burden on volunteer emergency responders. In the event of rekindling in individual smoldering nests, precise deployment planning can be achieved by leveraging the knowledge of the triggering sensors, particularly on larger fire sites.
 ZIMMERMANN, L., HOLZAPFEL, K. (2022): Ent-wicklung der Waldbrandgefahr in Bayern. LWF- aktuell 2/2022, 4–8.
Prof. Dr. Axel Göttlein heads the department of Forest Nutrition and Water Balance at the Technical University of Munich in Freising/Weihenstephan.
Roman Laniewski is working on the "Emergency Mixture" project funded by the Bavarian State Ministry of Food, Agriculture, and Forestry at the university, where the experiment was conducted.
Carsten Brinkschulte is the CEO of Dryad Networks GmbH, the company that developed the fire warning system.
Heiko Schwichtenberg is an employee at Bosch Sicherheitssysteme GmbH, which organized the integration of the fire sensors into the fire alarm infrastructure and the drone flight.