SIT GmbH

Our subsidiary Harbour Light Software Development Ltd. in Nova Scotia, Canada had a terrible experience in 2023 - the company had to be evacuated after a wildfire got out of control and threatened the associated buildings. Here some information about this fire: A major wildfire started around May 27, 2023, near Barrington Lake in Shelburne County. The fire has burned around 23,015 hectares (56,871 acres) and between 30 and 40 structures have been destroyed. The response to the fire included widespread evacuation orders of the surrounding area, displacing around 5000 people. This wildfire is the largest recorded in the history of Nova Scotia. Efforts to stop the fire included the use of American water bombers and additional firefighters from the U.S. and Costa Rica. The triggers of these wildfires are almost always (95%) human influences. The problem is the late detection of the fires. The longer a wildfire remains undetected, the greater the danger that it gets out of control. Therefore, we believe that systems are needed to detect and report wildfires at their inception. This requires analyzing, typifying, and alerting based on the gas composition of the air. Do we need to install sensors everywhere in the forest? No. Since fires usually occur where people are, it is sufficient to monitor the frequented areas closely. This way, most hazardous wildfires can be detected early. Basic components for early detection To ensure safe early detection on the ground, we need sensors made from various components that perform different tasks: - Flexible detection of gas compositions - each region "burns" differently. AI and learning models help detect fires effectively and regionally specifically. The basis for this is the BOSCH BME688 sensor from Bosch Sensortec, which allows training for gas composition detection using AI and can take over models for detection. - Flexible radio technology - energy-efficient and reliable - LoRa/LoRaWAN with GPS technology. - Batteries and solar cells for long-term operation in inaccessible terrain. We are still at the very beginning of our own development of a wildfire sensor. But - prototypes already exist! Currently, we are testing the prototype for functionality - afterwards, adjustments to environmental conditions will follow. The sensor acts very flexibly - via LoRaWAN, a detection pattern can be transmitted to the BME688 sensor for its area of use. Upon "release" of the sensor, its GPS coordinates are stored. The sensor automatically registers itself with the LoRaWAN gateway infrastructure upon startup, thus connecting to its data network. At predefined intervals, the sensor analyzes the ambient air - if it detects a pattern from its recognition model, it alerts a notification chain via LoRaWAN. If no detection occurs, the system shuts down until the next cycle. The sensor is powered by a battery that is charged by solar cells and has a usage capacity of about 2-3 years. A sensor can detect wildfires within a radius of about 1/2 hectare. Our goal is to develop a sensor as an open-source model that can be produced easily and cost-effectively, thereby achieving widespread distribution. Other components like LoRaWAN gateways are already available as open-source hardware.