This article is authored by A. Enis Cetin, Dennis Akers, Ilhami Aydin, Nurettin Dogan, Osman Gà¼nay & B. Ugur Toreyin. To enlarge the images so the charts are readeable, click on the Photos tab above.
We are losing the fight against forest and wildland fires. Here’s why:
- Wildfires are increasing in frequency, duration and intensity worldwide. Drought and other factors have increased not only the susceptibility to wildfires, but on an actual basis, the duration of the wildfire season. In many regions, the season has increased by 20% to 40% and the area burned by up to four times.
- Nowhere is this more prevalent than in the United States. Over the last three full years according to the National Interagency Fire Center (NIFC), the average number of U.S. wildfires was 87,004 and the average area destroyed by fire was in excess of 8 million acres. This is an increase over the average of previous “worst years” by 13% and 8% respectively.
- In addition to the increase in frequency, duration and intensity of wildfires, the risk of financial loss has increased significantly. Increasing intensity of wildfires has placed urban areas in danger as the fire penetrates the wildland/urban interface (WUI). In 2007, one of the worst wildfire years on record, more than 3,000 structures were lost in Southern California and insured losses (not total) exceeded $2.5 billion. And the exposure is growing because development in WUI areas is increasing. One recent study, which mapped wildfire occurrences against building development, found that the WUI had increased by 52% in the U.S. between 1970 and 2000. It projected at least an additional 10% in the coming 30 years.
Early Detection Can Be More Important Than Prevention
Unfortunately, most wildfires aren’t easily preventable. Although most wildfires are initiated by either human carelessness or arson, some of the most severe are the result of lightning and power line interference. A 2009 study by the California Department of Forestry and Fire Prevention (CAL FIRE) stated that of the 20 largest wildfires in California, half were caused by lightning or damage to power lines, both unpreventable.
To significantly mitigate the risk from and cost of wildfires, they must be detected and suppressed prior to reaching an uncontrolled state. In other words, the faster a wildfire is suppressed, the lower the cost. Again referring to Southern California in 2007, the rapid growth of wildfires and the multiple concurrent fires quickly overwhelmed the available resources. The fires were then only suppressed when the wind direction changed and the fires were blown back into themselves. (A review of recent major wildfires in California, Montana and Georgia indicates that these fires, too, were only brought under control with the help of a weather change–wind direction and velocity, humidity increase or rain.) According to some academics, a typical wildfire will double in size every five minutes. In high-wind conditions or extremely dry fuel conditions, the rate of growth will be much greater.
Determining Fire Location, Size, Direction & Burn Rates
Current wildfire suppression approaches often hinge on trying to predict where the fires are most likely to occur and “pre-staging” interdiction resources accordingly. But even when these activities are paired with important fire prevention aspects of public education, fuel load control and eliminating ignition sources during high-risk periods, the need for rapid detection and suppression is still one of the most important tools for containing wildfire damage. This typically means:
- Already having the right equipment in the right place
- Access to the most capable tools to assist suppression
- Having an accurate, up-to-date management view of the fire (e.g., wind direction, speed and other weather conditions, ongoing view of fire direction and growth rate)
- Simulation and forecasting of the fire size, speed and direction
- Making the correct resource allocation decisions (e.g., Do we send two firefighters in an SUV or a C-5 aerial flame suppression tanker?)
Although it’s certainly possible to gain this knowledge from a variety of “manual” sources, the first few minutes of fire inception are the most important, because they will determine fire deployment strategy. The fastest, most accurate data is increasingly coming from 24/7 automated detection equipment.
Automated Wildfire Detection Technologies
There are four major types of automated wildfire and forest fire detection:
- Ground-based visual systems
- Ground-based non-visual sensors
- Manned and unmanned aircraft
- Satellites
Each has its own advantages and disadvantages and strengths for different phases, types and locations of wildfire fighting.
Terrestrial (ground) visual systems
These systems monitor smoke and fire presence using the visible light spectrum. Typically this means using video cameras of some type. (Some systems require more specialized video cameras; others can use more common video cameras used for general security and surveillance, which can lower implementation costs.) These are often “tower-mounted” systems and have the advantage of continuous operation, relatively near the potential fire source. Because of their proximity placement, they can support ongoing operations even after a fire is initially detected.
With wireless data networks now available (and those that can even use cellular telephone networks for data transmission), ground-based visual systems can be deployed in remote areas. However, these platforms need to be fairly stable to prevent vibration-induced sensing errors, and they’re highly dependent on the software used to detect and report fires. Some systems are highly automated (essentially executing most monitoring and signaling tasks for the operator), while others require more direct human operation.
Although the range for these systems is growing (up to 300 square kilometers for some visual camera-based versions), they do require being deployed in networks if “massive” areas require monitoring. Conversely, for areas with definite boundaries–such as home developments or areas around power and communication lines–the detection distance is less of an issue. The higher quality versions of these systems resemble a simple “alarm” that provides coordinates, size, speed and direction of fire without any special interpretation or even human intervention.
There are also specialized detection systems such as “under the canopy” heat/flame sensors that can be mounted at ground level or on trees. These can be relatively inexpensive and can fill in “dead spots” that may occur with other systems (e.g., if wireless communication cannot penetrate a mountain into a valley). Some have the advantage of being powered biologically (e.g., gaining power from soil biochemical reactions), and can provide immediate detection if heat or flame is present. Unfortunately, they are relatively short-ranged (typically up to 500 meters), which require hundreds to cover a reasonable area. They require some type of repeater or “daisy chain” communication network, which can be less dependable if a node is lost. And of course, being ground- or tree-mounted makes them more susceptible to loss from fire or even theft.
Terrestrial (ground) non-visual systems
Sharing many of the features of fire monitors employing visible light as noted above, non-visual sensing systems typically use infrared light to sense the presence of fire. Their chief advantage is that they can sense heat and fire even in occluded conditions, such as through smoke, fog or heavy rain, which can cause issues for some visual-light spectrum monitors. Conversely, these systems will not detect smoke, which can mean a delay in fire detection until high heat or open flame is produced.
Because the technology employs a specialized infrared sensor, there are typically fewer choices for “off the shelf” style camera sensors. Sensing range performance has typically been a bit less powerful than visual camera systems, but each system varies in performance depending upon set-up and environment. Because they use the infrared spectrum, there is sometimes no actual visual “read out” of the fire scene for an operator to confirm. Other configurations essentially combine a visible light camera with infrared, or use a display that interprets the heat in a visual representation like a handheld thermal imaging camera (TIC). Some systems subdivide the infrared spectrum into distinct frequencies. These can then be used to “filter” non-infrared flame sources to lessen false alarms.
Aircraft
Both manned and unmanned aircraft can be extremely flexible fire detection tools and can move to affected areas quickly. They also have the advantage of potential fire spotting within larger areas if they’re capable of higher altitudes. Although special equipment is typically used to detect and document fire locations, even “lower tech” versions, using human spotters and radio transmission, can be quite helpful once a fire has been detected.
Limitations of aircraft can include a relatively high cost to deploy as pre-fire detection, since ongoing operation is high (fuel costs, pilot cost). If an area has a high elevation and is prone to low clouds, visibility can be an issue. Depending on pilot navigational resources and skill, detection of fire in specific areas can be a problem if they use an inconsistent or inaccurate search pattern, which means they could easily miss a fire inception event.
Satellites
Earth-orbiting satellite systems have large area coverage capabilities and can monitor on a 24-hour basis with little operational limitation. They are particularly well suited for “post fire” assessments of damage. However, depending on the type of technology, they may have difficulty in spotting fires if weather conditions are cloudy. Obviously, the cost to deploy these is very high and their use has to be shared with other critical functions, such as weather detection. Although resolution is increasing, some satellites cannot provide a very high resolution or closely focused view of a fire event. Due to orbital limitations, such as only being able to pass by a fire every four or more hours, many cannot provide continuous views of a fire. Satellite operation and data interpretation also require highly trained personnel.
The Cost of Early, Automated Detection
The bottom line for most jurisdictions is cost of monitoring per square mile vs. accuracy. Local, state and federal firefighting entities, insurance companies, utilities and land owners/users (e.g., logging companies) need to be able to assess whether this technology will lower their fire damage control costs, and if so, by how much. In these areas, automated wildfire detection systems have made many gains and have begun to prove their cost effectiveness over manual reporting approaches (spotters, citizen reporting, etc.) including:
- Reaching a daily cost to monitor a square mile as low as $.05, including deployment, hardware, software, power and personnel.
- Improved pattern recognition algorithms that can recognize smoke or flame more quickly, even at night or under adverse conditions, such as fog, resulting in the detection of a 1-meter-square plume of smoke at a distance of nearly 15 kilometers within 25 seconds of its presence.
- Lowering or eliminating problems with “nuisance alarms” which can be caused by camera/sensor vibration or weather/cloud conditions.
- Lowering technical levels for operators, including the leveraging of operators to those who merely confirm the results of a detection event before deploying response teams.
- Increasing (leveraging) network operations personnel through the ability to “gang” many networks of sensors together, meaning potentially thousands of sensors (representing millions of square miles) can be monitored in one center.
- Lowering the cost of system acquisition and deployment by using standard “off the shelf” sensors like standard security cameras or repurposing cellular networks for data transmission.
One Locality’s Determination to Decrease Wildfire Damage
An example of such increased automated monitoring effectiveness comes from Turkey, a country with high levels of urban-style development infused into relatively rural, wooded areas that has been plagued in the past with high property and casualty losses due to wildfires. By deploying a network of tower-mounted, optical camera sensing systems using an advanced algorithm for automated, early sensing of smoke or fire, the country has reduced the average area burned per fire by 80% through combining auto detection with other program improvements.
The chart (photo 4 in the photo viewer above) shows that achieving these results required a deployment of 154 sensing cameras to monitor nearly 8 million acres of forested land. Notably, because of the increase in the monitored area in the five years since the program’s inception, the fire interdiction response times decreased by nearly 50%.
So Why Aren’t We Using This Technology to Fight Wildfires?
One of the main barriers to deployment of automated wildfire detection is lack of familiarity with the systems. The other is the sense that the cost/benefit ratio isn’t high enough based on past performance. However, given the advances in technology, plus the rising occurrence of wildfires, it seems reasonable to revisit this technology.
The cost for implementing trials in relatively large areas is now within the cost and scope of even limited local, state and federal budgets. Plus the value to non-governmental units such as insurance companies, utilities, pipeline companies, logging operations and warehouses invite the potential for offsetting deployment and operational costs through public-private partnerships.
With the fury of wildfires increasing, it would seem a good time to take a second look at automated wildfire detection.
This article was primarily researched, written and edited by the following individuals:
Professor A. Enis Cetin is chief technology officer of Wildland Detection Systems, where he is responsible for guiding the overall development efforts for the company in the remote/automated sensing/monitoring applications. He obtained his PhD from the University of Pennsylvania and as served as assistant professor of electrical engineering at the University of Toronto. Dr. Cetin was an associate editor of the Institute of Electrical and Electronics Engineers (IEEE) Transactions on Image Processing between 1999-2003. Currently, he is on the editorial boards of journals Signal Processing, Journal of Advances in Signal Processing (EURASIP), and Journal of Machine Vision and Applications (IAPR), Springer. He is a fellow of IEEE. Dr. Cetin’s research interests include signal and image processing, human-computer interaction using vision and speech, and audio-visual multimedia databases. (His full biography can be accessed here.)
Dennis Akers, PE, is chief executive officer of Wildland Detection Systems (WDS). Akers has worked in the automated fire detection arena since 2003 in installations in the United States and abroad. Leveraging his training as a professional engineer, he launched WDS by helping to expand the use of other pattern recognition and signaling systems used for leak and intrusion detection used by its parent company, Delacom Detection Systems (DDS). Since then, WDS has grown from the venture level to the production stage and is expanding globally, with more than 150 installations. Prior to creating both DDS and WDS, Akers worked for more than 20 years in various technical development roles, including technologies such as broadband, wireless communication systems, subsystems and service networks. (His complete biography can be accessed here.)
Ilhami Aydin is manager of communications for the General Directorate of Forestry, Turkey. He received his bachelor’s degree in Electronics Engineering from Erciyes University, Kayseri, Turkey in 1985.
Nurettin Dogan is vice president of the General Directorate of Forestry, Turkey. He received his bachelor’s degree in Forest Engineering from Istanbul University, Turkey in 1977.
Osman Gà¼nay is a Ph.D. student in the Department of Electrical and Electronics Engineering at Bilkent University, Ankara, Turkey. He received his bachelor’s and master’s degrees in Electrical and Electronics Engineering from Bilkent University, Ankara, Turkey. His research interests include computer vision, video segmentation and dynamic texture recognition.
B. Ugur Toreyin, PhD, is assistant professor at Cankaya University, Turkey. He received his bachelor’s degree from the Middle East Technical University, Ankara, in 2001, and his master’s and doctorate degrees from Bilkent University, Ankara, in 2003 and 2009, respectively, all in electrical and electronics engineering.
