Fire detection

The Silvanet system continuously monitors air quality. The detection of a smoldering fire triggers fire alerts.

If a Wildfire Sensor determines with a high probability that a smoldering fire has occurred, then fire alerts are immediately sent to notify users of the fire.

Fire Alerts

For details about Fire alerts displayed in the Site Management app, sent via email and available with the MQTT interface for fire alerts, see Alert Center.

False alerts may occur before end of 14 day calibration period

Silvanet Wildfire Sensors are ready to detect fires ONLY after the sensors have been calibrated for 14 days after deployment.

Consequently, before the calibration period ends, the Wildfire Sensors do not provide useful information and may generate false alerts.

Smoldering fires A smoldering fire is defined to be a slow, flameless combustion of a biomass material such as forest floor material, branches or leaves.

Continuous monitoring

After the Wildfire Sensor has been calibrated, the gas sensor continuously monitors the microclimate of the forest air to measure:

• Air pressure

• Temperature

• Humidity

At the same time the Wildfire Sensor "smells" the air around the sensor for the presence of Volatile Organic Compounds (VOCs) and carbon monoxide.

VOCs

Volatile Organic Compounds (VOCs) are compounds that have a high vapor pressure and low water solubility. Wildfire smoke contains many toxic VOCs depending on the type of fuel source of the fire.

Index of Air Quality (IAQ)

The sensor “smells” the air on a regular basis to determine if the air quality has deviated from the saved Normal Air Quality (see below). Air quality is quantified as an Index of Air Quality (IAQ) which has a range of values that indicate the quality of air within range of the sensor.

This continuous monitoring allows the Wildfire Sensor to distinguish between normal and declining Air Quality:

• Normal Air Quality:

  • Normal Air Quality is the baseline reading of the Wildfire Sensor. It is the result of the 14 day calibration period of the sensor.

  • This process continues after calibration.

• Declining Air Quality:

  • When the Air Quality deviates from the baseline reading (Normal Air Quality), this is considered a decline of Air Quality.

  • However, this decline could have many causes. It could be the result of similar gasses which the sensor has detected. These gases could be diesel fumes from a passing truck, cigarette smoke or other factors that can cause a decline in air quality.

  • The role of the Wildfire Sensor is to determine if this decline is the result of a smoldering fire or some other source.

Triggered Fire alerts

Once per minute, the Sensor wakes up and performs an environmental monitoring gas scan. During each gas scan, hydrogen, carbon monoxide and other Volatile Organic Compounds (VOCs) are measured. It also measures air pressure, temperature and humidity.

Fire detection process

The Wildfire Sensor uses a default sliding time window of 5 minutes to "smell" the air. If the Wildfire Sensor detects a significant deterioration of air quality within the past four gas scans, a prescribed set of gas scans are triggered to determine if the observed deterioration is from a smoldering fire or from some other source.

The Wildfire Sensor only needs only 2-3 minutes to determine if a fire is present.

For each gas scan, the results of the gas scan are compared with with a pre-trained Machine Learning (ML) Model. This is done to determine if the observed decline in air quality can be classified as wildfire smoke or from some other source. As these gas scans consume a lot of energy, they are kept to a minimum.

• Low probability of fire (Sensor Normalization): If a source other than a fire caused the decline in air quality (diesel fumes or some other similar gasses), then the sensor stops the gas scans and does not trigger a fire alert. In this case the sensor normalization is required which takes 30-60 minutes. See Sensor Normalization.

• High probability of fire (Fire Alerts): If a smoldering fire caused the decline in air quality, then the gas scans are terminated and the sensor immediately triggers a stream of fire alert packets to the Silvanet Mesh Network. Users are then notified via email and through the Site Management app.

Returning to normal values after tests

After the sensor performs gas scans, it needs approximately 30-60 minutes for the sensor to return to baseline. During this time, the sensor does not perform fire detection.

Machine Learning models

These models have been trained in Dryad's lab to detect typical gas compositions of smoke from burning trees.

ML (Machine Learning) Models are pre-trained in the Dryad laboratory and can be fine-tuned / programmed specifically for the species of trees present in a particular deployment. Updated models can be remotely installed in the sensors by Dryad’s support team using FUOTA.