Forest Fire Detection

Early warning systems are a very new area in IoT technology, and few are using much in the way of Internet aware design, but many more plan to. One example of an early warning system would be a forest fire monitor. Very little has actually been accomplished here, most is still in the planning or proof of concept stage. A few reasons for this could be the difficulty of the problem, the scope, and the expense required to do it well. Currently, a few systems that are in the works are using a variety of methods to detect a fire. The vast majority of them use satellite telemetry data. While the vast majority of  them are using satellite telemetry data, and while this does not qualify as IoT, these systems are augmenting this data using information available from the Internet. Other systems are experimenting with visual fire detecting algorithms and overall monitoring of forest environmental conditions in an effort to predict fires.

Currently the biggest forest fire monitoring and reporting application is Active Fire Mapping which is run by the USDA Forest service. This application has been in use for a number of years and watches fire activity all over the United States. The system gathers data by once-a-day satellite flyovers that contact ground stations in fixed areas. The ground stations report conditions to the satellites and they in turn determine if there is a fire and where it’s projected path is in it’s projected path. Currently, using this method, the system is accurate within about 1km. Recently, it was announced that the USDA is looking to augment the system using smaller network attached sensors on prop planes. These could fly over known fire incidents and collect data down to 30m of accuracy [10].

Another fire monitoring solution comes out of Russia. The Nizhny Novgorod Forest Fire Center is an application that monitors several large sections of state-owned forest. Their existing technology had been in place for a while, but response times to fires was still too slow. Recently, they installed several network-attached cameras to critical points within the forest. When a fire is detected, operators can use this streaming data to gather information and better fight the fire [11]. There are no details if the cameras have a fire detection algorithm (really, there were few details on how the system works at all), but this is an example of how fires could be programmatically identified. By monitoring a fixed section of forest and using an image analysis algorithm, it is possible to identify when the section is on fire. Again, it is unclear if that is what is happening here, but it is a logical next step.

A slightly different approach to monitoring forest fires was developed in China. Several researchers there set up the FEFCP (Forest Environment Factors Communication Platform). This is another Zygbee-based sensor network that monitors environmental conditions in several forested areas of China. The sensor network regularly reports light levels, soil conditions, humidity and temperature in an effort to capture overall forest health. While not specifically targeted at identifying fires, it is suggested in the study that data could be used to predict “fire risk days”. The researchers said this data could be used to choose which nodes to power in an effort to conserve energy, but the data could also be used to prevent fires from starting [12].

When trying to locate an actual functioning fire monitoring system, I came across a mock-up of a system presented by a 14 year old owner of the startup company Zenbotica. The interface was presented at the IoT Innovation World Cup 2015 in Munich. It shows how a single dashboard could be used to track sensor readings over a wide area, plot incidents on a map, as well as report information about firefighters in the field. The system has not been implemented, but has been noted several times as a good concept for such an application [13].