Local wildfire detection system shines

The FireHawk cameras mounted on the Marten Mountain Lookout Tower in Canada were monitored by the FireHawk ops room in Chile.

The South African-developed FireHawk wildfire detection system has been found to be the most effective fixed fire detection and reporting system in an international trial that took place in Canada recently.

FireHawk combines human skill with artificial intelligence to keep watch over two million hectares of forest land around the world, alerting landowners and triggering rapid response whenever a fire is detected.

Willem Oosthuizen, CEO of FireHawk South Africa, explains: “The Alberta Wildfire Detection Challenge was a collaboration between Alberta Wildfire, Alberta Innovates, and FPInnovations. Six commercially-available fixed fire detection systems were installed and operated on the Marten Mountain Lookout tower near Slave Lake, Alberta, Canada, during the 2022 wildfire season. The aim was to test these fire detection systems in an operational environment in an area that experiences an average of 20 wildfires annually.”

The trial programme organisers invited 17 leading fire detection service providers from around the world to submit applications. This was narrowed down to the best six applicants who were selected to take part in the challenge. The selected applicants were provided a contract with funding to demonstrate their systems using their own proprietary equipment. A total of 54 events were recorded during the test period, and of these 14 were real wildfires and 33 were test smokes.

The Marten Mountain Fire Lookout Tower manned by local fire watch personnel also participated in the trial in order to provide comparative data.

The FireHawk cameras mounted on the Marten Mountain Lookout Tower were monitored remotely from FireHawk’s operations base in Chile for the trial.

Study location
All systems were installed on the Marten Mountain Lookout tower in Alberta, Canada. The tower is 20 km north of the Town of Slave Lake. The 40 km radius of the lookout coverage area has the following views: to the west is the waterbody of Slave Lake; to the east is a hill that limits visibility; to the south is the town of Slave Lake and the Mitsue industrial area; to the north is forest. Boreal forest species are the primary vegetation cover. Many industrial activities such as forestry, oil, and gas are also spread across the landscape.

This boreal forest area experiences a high number of wildfires due to a high number of lightning strikes and human activities. In 2011, a wildfire entered the town of Slave Lake and burned around 500 structures.

Evaluation metrics
Four metrics were used to evaluate the performance of the detection systems: detection distance, reporting efficiency, location accuracy, and system availability.

Detection distance
Results show the Marten Mountain Lookout observer was effective at the 40 km distance for both test smokes and wildfires. In addition, the lookout observer also detected a wildfire smoke at 46 km.

Of the six fixed detection systems, FireHawk had the best performance with an 88% detection success rate between 10 and 20 km on test smokes, and 78% between 20 and 30 km.

Reporting efficiency
FireHawk had the fastest average time reporting test smokes among the six systems: 3 minutes @ 5-10 kms; 6 minutes @ 10-20 kms; 10 mins @ 20-30 kms; 7 mins @ 30-40 kms.

The Marten Mountain Lookout observer was the best performer of all trialists with an average of 4 minutes over all distances. The rest of the fixed systems exceeded 10 minutes.

System availability
The total number of operational hours between July 1 and September 15 (8:00 am to 8:00 pm daily) was 924 hours.

The Marten Mountain Lookout observer did not have any downtime. Neither did FireHawk, which recorded 100% availability.

Firehawk systems use human operators. The data shows that these systems perform better than systems that do not have human operators.

Location accuracy
IQ FireWatch had the best accuracy on reporting locations and was better than the Marten Mountain lookout’s performance. SmokeD was second overall, but the sample numbers were too low beyond 10 kms to determine the reliability of the result.

Night detection
Night detection is the detection service provided during low or no sunlight. ForestWatch, SmokeD, and FireHawk chose to provide detection coverage outside of the operational period which included night detection. Only eight detection messages from public reporting and FireHawk were considered as night detection.

These results show that FireHawk has the ability to detect fires during night time. Two successful night detections were the results of an attended fire and a car fire. These two detection successes suggest that this type of detection system could be used as a monitoring tool at night.

Human operators
Firehawk, IQ FireWatch, ForestWatch, and SmokeD systems used human operators. The data showed that these systems performed better than systems that did not have human operators.

FireHawk also had the lowest number of false alarms at 11 over the two and a half months trial period.

“We are proud that our FireHawk system experienced no downtime during the detection challenge, and had the highest number of detection successes among the systems taking part in the programme,” said Willem. “It was a privilege to take part in this international testing of different wildfire detection systems.”

Where it all began
After 16 years of experience in aerial spraying and aerial fire fighting, Jake Oosthuizen formed Zululand Fire Protection Services cc. (ZFPS) in 1994 with the object of providing a service to the Zululand timber industry. Initially, the operation was started by taking over the control centre of the Zululand Fire Protection Association (ZFPA), which coordinated all fire fighting operations in the Zululand Coastal Area (approximately 80 000 hectares of timber).

At that time, the Control Centre had a very low profile and provided a basic communication service to the timber growers in the area. Over time, the Control Centre became the heart of all firefighting operations in Zululand. ZFPS expanded to include the management of ZIFPA.

In 1994, ZFPS was instrumental in the development of the FireHawk system that is today used in South Africa, Chile, Brazil, Malawi and Ghana.

FireHawk was the first computerized fire detection system in the world and has been operating commercially for the past 28 years. The first system was installed in Richmond in 1994 and is still operational today.

This is where it all started … the first FireHawk system was installed in Richmond, KwaZulu-Natal in 1994, and is still operational today.

The FireHawk system uses 360 degrees rotating high definition digital cameras to monitor areas and transmits information to a base station that is manned by operators. The software differentiates between fire, smoke and glow, and then raises an alarm. Positioning is done from a single camera, with the ability to cross reference for improved accuracy.

FireHawk now monitors approximately 591 000 hectares of timber plantations in KwaZulu-Natal, the Eastern Cape and Mpumalanga.

It also monitors 650 000 ha of timber in Chile, 155 000 ha in Brazil, 4 000 ha of sugar cane in Malawi and 600 000 ha of Burmese Teak in Ghana.

Find out more at firehawk.co.za

Discovering the cause and origin of a destructive wildfire

Fire expert Dave Dobson was hired to find the cause and origin of a wildfire that left a swathe of destruction across farms and forestry plantations in the KZN midlands in 2007. Working years after the event, Dave followed the trail of smoke to uncover the origin of the runaway fire that burnt thousands of hectares to a cinder …

The dispute surrounding the ‘Kentucky’ fire of 25th June 2007 that devastated farms and commercial forests in the Curry’s Post area of KwaZulu-Natal has eventually been settled. The fire caused extensive damage to privately owned farms and commercial forest plantations in the area.

As a result of the fire one of the plantation owners instituted legal proceedings against the owners of two properties where the fire was purported to have originated. This case study deals with the case against the second Defendant, the owner of the farm Kentucky.

The challenges
The objective of every fire investigation is to establish the cause and origin of the fire and to determine the ignition sequence.

The first challenge, and one that is regularly encountered in work of this nature, is the fact that I was appointed to investigate the origin and spread of this fire on behalf of the Defendant some seven years after the event (i.e. in 2014).

The Defendant was not a member of the local Fire Protection Association. This raised two issues. The first was that the Defendant was deemed negligent in the event of a fire originating or exiting his property and the onus rested on the Defendant then to prove his innocence. The second issue was that the Defendant was denied access to information held by members of the Fire Protection Association. This was particularly important in respect of the origin of the fire under investigation.

Other issues that contributed to the charge of negligence against the Defendant related to the presence of a compartment of plantation waste/slash located on the boundary between the Defendant’s farm and a neighbouring commercial forest plantation. This was deemed to be a fire risk which contributed to a massive flare-up that resulted in the fire spreading to the Plaintiff’s plantations.

The fires
A number of fires occurred during this period. The first two occurred on the property adjoining Kentucky Farm (St Clair Estate) while the third, as will be shown later, arose on Kentucky Farm. This led unfortunately to the fire becoming known colloquially as ‘the Kentucky Fire’ which was not entirely correct!

The first fire (fire A) occurred on Sunday 24 June 2007. It was extinguished on St Clair Estate (Kentucky’s neighbour) and played no further role in the events that unfolded the following day.

The second fire (fire B) started on the morning of Monday 25 June 2007 on St
Clair Estate and swept through Kentucky Farm fanned by extreme weather conditions.

A third fire (fire C), that was not reported, started on Kentucky Farm during the afternoon of Monday 25 June 2007. This is the fire that resulted in the extensive damage to commercial plantations and farms in the area, which was the focus of this lawsuit.

Data collection
I approached the investigation in a systematic manner beginning with a site visit and an interview with the Defendant. This proved useful in providing circumstantial evidence tracing the progress of the fire through the Defendant’s property. Of particular importance however, was acquainting myself with the overall fire area which was to prove beneficial later in the investigation.

Empirical data was collected in the form of weather data during the time of the fire, interviews and fire reports from various parties who were involved in fighting the fire as well as photographs taken from the spotter aircraft monitoring the fire. A vital piece of evidence was pictures taken of the origin and spread of the fire recorded by Fire Hawk’s locally based fire tower. This information was not readily available to the Defendant since he was not a member of the local Fire Protection Association. However, the picture sequence of the fire was in the public domain as it was being used to conduct Fire Boss Training courses. A simple request for the pictures provided access to this vital piece of information!

In addition, and rather late in the day, an Expert Report was received from the Plaintiff’s legal team.

Cause and origin of the fire
Once the data had been collected and analysed it was possible to develop a hypothesis of the events that occurred on this day. This hypothesis was then subjected to systems analysis to develop a broad understanding of what happened and to avoid linear thinking. This initial hypothesis is depicted in the following diagramme.

In developing this initial hypothesis consideration was given to the summons which suggested that the fire originated from the Sunday night fire (fire A) that had – according to the summons - not been properly extinguished. Various witness statements however refuted this contention. Furthermore, picture evidence obtained from the Fire Hawk camera appeared to indicate that the fire that damaged the Plaintiff’s plantations originated near some homes on the neighbouring St Claire Estate and was the result of a member of the household throwing out ash from the previous night’s hearth fire during the morning of Monday 25 June (fire B).

When examining a problem systematically there are tools in systems thinking that can assist with an enquiry. Systems archetypes are one such tool. As one works with the problem, developing the story, identifying the key variables and sketching them in causal loop diagrams, patterns begin to emerge which provide keys to unlocking parts of the problem. These generic patterns are described in systems thinking as archetypes (Kim, D., H. and Anderson, V. 1998).

A typical “Escalation” archetype now emerges where discarded ash from a previous night’s fire catches alight and starts a fire (fire B). This fire later enters the plantation slash on Kentucky and under deteriorating weather conditions escapes causing havoc to farms and commercial forest plantations down wind. This served as an initial hypothesis for testing against further empirical evidence.

The crucial evidence related to the slash in an old pine compartment located on Kentucky Farm, adjacent to the boundary between Kentucky farm and a neighbouring commercial forestry estate. The trees in this compartment had been clearfelled six years prior to the fire and most of the plantation residue would have decomposed by the time of the fire. Nevertheless the Plaintiff’s Expert presented pictures showing clear lines of fire burning in the old pine compartment which were ascribed to the brush piles having caught alight. The picture also clearly showed a road in the area.

I was aware that no such road existed and began searching for an answer to this conundrum. It turned out that the fire depicted in the photographs was actually burning on a property on the other side of the Curry’s Post road some 1.5 km away! The owner of the property was contacted and confirmation was received that he had in fact cleared and stacked jungle wattle and gum in this area of his farm a few months prior to the fire! The picture of the plantation slash burning that was presented as evidence of negligence on the part of the Defendant was incorrect – the photos were of another property on the other side of the Curry’s Post road.

It was time to test the validity of the evidence of the fire entering and escaping from the old pine compartment on Kentucky Farm.

A closer inspection of the photographic evidence of the fire in the section of Kentucky farm bordering the commercial forest revealed further interesting empirical data. The first was clear evidence – i.e. straight line burns - of firebreak burning in kikuyu camps adjacent to the old pine compartment on Kentucky. Wild fires do not burn in straight lines. From the available photographs it was evident that the party responsible for initiating this firebreak burn lost control of the operation resulting in it entering the old pine slash compartment on Kentucky farm, and subsequently the neighbouring forest where it was brought under control by the land owner. However, the fire in the old pine slash compartment on Kentucky continued to burn. This was fire C.

The events leading to the origin of this fire were never reported, although mention was made in a fire report of drip torches being issued to a helitack ground crew who were dropped in the area in an effort to mop up and contain the fire.

Soon after this, evidence of the fire break between Kentucky and the commercial forest plantation being half burnt indicated a further attempt to contain the fire. The Expert for the Plaintiff (who had previously been contracted by Kentucky’s neighbour to prepare a report on the fire) mentioned the “application of a counter fire along the land owner’s boundary closing up to the main plantation road.”

Clearly there were numerous attempts to introduce counter fires and fire-breaks along this boundary under extreme weather conditions.

At the same time as fire C was burning in the old pine compartment on Kentucky farm, a power failure at the landing strip where the water bombers were refuelling delayed their return to the fire. A further complication arose when weather conditions became too dangerous to fly. Eventually the fire exited the old pine compartment and was driven, out of control, by the extreme weather conditions across many farms before entering the Plaintiff’s commercial pine plantations where extensive damage was experienced.

Final hypothesis
It now becomes possible to develop a final hypothesis describing the fire. This hypothesis is summarised in the diagramme that follows.

Initially as mentioned a preliminary hypothesis was considered which was represented by an “Escalating” archetype: Discarded ash catching alight and resulting in the fire spreading onto Kentucky Farm from whence it enters the slash in the old pine compartment and later spreads to neighbouring farms and forest plantations (fire B).

However, on further investigation a second archetype emerged – a ‘Fixes that Fail’ archetype where an unintended consequence of the use of counter fires or fire-breaks set in an effort to contain fire B, results in the fire escaping and entering the old pine compartment on Kentucky farm from whence, after a delay (indicated by the parallel lines) the fire enters farms and commercial forests causing immense damage (fire C).

An interesting addition to the overall picture is the loop linking Curry’s Post road to the main fire. Pictures of the windrowed plantation slash on a property some 1.5 km away from the Defendant’s farm across the Curry’s Post road from the Defendant was presented as evidence of negligence on the part of the Defendant. This was shown as being clearly incorrect!

After careful consideration of the evidence and interviews with people involved it became clear that the Defendant did not act negligently.
The unintended consequence of trying to “fix” the problem of fire B exiting Kentucky by burning counter fires and fire-breaks on the Defendant’s property is the probable cause of the subsequent damage to farms and forests downwind of Kentucky. This counter fire activity was never reported.

The evidence presented of the extreme fire danger posed by the plantation slash on Kentucky plus the photographic evidence of windrowed slash burning in support of this claim was refuted.
In the end the Plaintiff’s summons was deemed to have no substance and was subsequently withdrawn. The Plaintiff ended up having to pay the Defendant the cost of suit.

KIM, D., H., ANDERSON, V. 1998. Systems Archetypes Basics. From Story to Structure. Waltham Massachusetts, Pegasus Communications Inc.