Banff National Park, Alberta, Canada - wildfire

Are wildfires getting worse?

Image credit: London School of Economics

Following what is the most destructive and deadliest wildfire season on record in California last November, there has been increasing concern about the severity of wildfires in the years to come.

2018 was a bad year for wildfires. Regions as diverse as the Mediterranean  and the Arctic Circle were exceptionally hard hit, and the wildfires in California caught the world’s headlines. Even moorlands near Manchester  were ravaged by flames and  declared a major incident. Frightened residents who were evacuated from nearby villages described the fire as an “apocalypse” and a “towering inferno”, and thick plumes of smoke were visible from space.

A warming planet has been blamed for the ‘mega fires’ of recent years, which have raged with an intensity fire crews say they’ve never seen before. Deadly Californian fires broke new records last year, tearing through towns and leaving dozens dead in their wake. Last summer, northern Europe experienced unprecedented level of blazes through countryside desiccated by the heatwave. In 2017, Spain and Portugal suffered tragic losses, while Greece experienced its worst ever wildfire last summer. Fire seasons around the world are becoming longer and starting earlier. For the first time in the UK, severe wildfire was included in the national risk register in 2013.

“Our wildfires around the globe are much more intense and much more severe than in the past,” says Andy Elliott, whose decades of firefighting expertise have led him to become one of 12 wildfire tactical advisers for the UK. He also runs his own wildfire consultancy service, WildfireTaC. “They burn everything in their path, often destroying the soil itself, such is their severity.” Wildfires can burn at more than 1,000°C – hot enough to melt gold, generate hurricane-force winds, and move twice as fast as the average person can run.

Fires have played a crucial role in ecosystems for thousands of years. But if the planet continues to warm, and we do nothing to manage fuels that build up in the absence of natural fires – less than 1 per cent of US fires are allowed to burn – then we are in trouble. With forests and vegetation growing drier, and more houses built in forest land, there’s more fuel than ever ready to ignite.

“Fire will become an uncontrollable, dominant factor in our landscape,” Elliott warns. As a firefighter, he witnessed the Valley fire sweep through Lake County in California during 2015. “Despite all the resources that California had available to throw at it, it just grew and tore through communities and open landscapes,” he notes. Last year, wildfires in California were the most destructive ever recorded.

Clearly, say fire scientists, we need to harness the potential of technologies to help forecast and predict how and when fires will rage, and analyse data in the wake of a wildfire to improve tactics and training.

The burning issue


■ The UK spends around £55m a year on fighting wildfires.

■ As the climate has been heating up over the last 40 years, the number of fires worldwide has grown. Globally the length of the wildfire season had grown by nearly 19 per cent between 1978 and 2013.

■ Worldwide, an area equivalent to about 20 times the size of Great Britain is burnt by wildfires on average each year.

■ By July 2018, the area burned in the UK that year was more than four times the average of the past decade.

■ In Europe, the vast majority of fires are caused by humans.

Fundamental fire science is an area crying out for better research, says Guillermo Rein, professor of fire science at Imperial College London, who initially trained as a mechanical engineer. “We still fight fire as we fought it 100 years ago.”

Rein says what’s missing is a deep scientific understanding of how fire spreads. Because of this, the applied science is weaker than it should be. “Fire is complex. It’s the intersection of many different sciences: heat transfer, thermodynamics, chemistry, atmospheric science, ecology, structural engineering. All these sciences must develop. We have ideas, thoughts, experiments and hypotheses, but not the fundamental theory. We still can’t produce a model to predict accurately what might happen.” And the rate of progress is glacial, he says. “It’s painful. It takes time to develop. Authorities ask for help as California, the Mediterranean or Australia burn. But by then it’s too late – we should have begun five years ago.”

Firefighters still use tools deployed for decades; US crew attending a wildfire, for example, will clear away fuel with a trusted Pulaski – a long-handled adapted axe.

Firefighters aren’t luddites, but they need to have faith in their tools when they’re risking their lives. Fighting wildfires can be physically exhausting – while planes and helicopters can drop water and chemical suppressants, this is a last resort and not particularly effective. Although northern Europe has a shorter fire season, intense blazes can overwhelm fire services, which mostly don’t have dedicated helicopters to hand.

“New suppression technologies are a little like using a better glue to stick a plaster on an arterial bleed – simply not good enough,” says Elliott. What firefighters really need is more and better information, he says.

Scientists are looking at computer modelling of fire scenarios, to understand the effects of fuel, weather and landscape on wildfires.Fire modelling programs are used in both Canada and the US, but they have limitations. “Their forecasting ability is too poor currently for firefighters to trust,” says Rein, because there’s a lack of empirical data to inform these models.

US services work with ‘off the shelf’ modelling and simulation program Farsite, while Canadian fire services use a program known as Prometheus. These projects work with geographic information systems that combine spatial data with topography, vegetation and transport networks. But fire centres are far from using real-time data to predict how fast a wildfire will spread and with what intensity, and modelling isn’t yet used operationally.

“Until recently, wildfire-spread modelling has been used only for training and planning prescribed fires,” says Thomas Smith, assistant professor in environmental geography at the London School of Economics and Political Science.

Planned fires allow dry vegetation and other natural fuels to be burned in a controlled way during benign weather, preventing landscapes from becoming a tinderbox during drier windy conditions. Unlike random wildfires, the behaviour of planned fires is obviously easier to predict. Once started, flames advance with the wind, burn uphill faster and will rip across south-facing slopes where vegetation is drier. This information helps experts compute how fast a fire will move and with what intensity. But fires can also lurch, change direction, jump across crowns of trees or even valleys.

'These advances mean that today’s satellite remote sensing data feeds typically provide very rapid information for applications far beyond meteorology – data useful for detecting landscape fires, monitoring their evolution, and characterising their behaviour’

Professor Martin Wooster, Kings College London

In the US, some fire services are starting to experiment with emerging technology. Stations in California are trialling the use of drones equipped with thermal cameras to report back on remote fires. “Somewhere as large as California, it makes sense to deploy drones,” says Smith. New sensors and upgraded mountain-top cameras in North America now enable detection of even tiny smoke plumes, which reveal fires in their early stages. At the University of Nebraska, researchers have experimented with ‘fire drones’ that drop fuel fire bombs to set prescribed burns in the wild.

Some US state agencies are using an app called Collector, which allows fire crews to see geographic information data and local details such as water sources and structures and use this to map out fire attack plans. Crews can also upload pictures and video and record their location.

Intelligence from planes, satellites and sentinel drones could eventually deliver information faster and more often. “But at the moment, we are just on the edge of connecting it all up,” says Smith. “In the future we’ll be able to have a data stream from a drone to people on the ground, and plug this in to other data on topography, weather, fuel, windspeed, direction and so on, and be able to predict how a fire will behave.” With this improved situational awareness, firefighters could make more informed decisions. “These technologies already exist,” Smith says. “They are just not operational yet.”

In an ideal world, fire services will be able to forecast how a fire will behave in the same way meteorology predicts the weather, but academics estimate that this scenario is possibly a decade away. “Current forecasts are one hour behind at best,” says Rein. “But if we could predict six hours ahead or more, that would be a revolution.”

Knowing, for instance, how high flames will reach would help services decide what tactics to deploy. Below one metre, fire­fighters attack manually with beaters and hoses, says Smith. “Between one to three metres, you might start thinking about creating fuel breaks, maybe with a chainsaw or bulldozers.”

Off-the-shelf modelling programs don’t travel well. Vegetation and conditions vary hugely from country to country, and more data is required to refine them. The heather that covers much of Dorset is unique and burns differently from Canadian grass and forest. Peatland fires smoulder underground and are difficult to extinguish, particularly if surface vegetation is drier than usual. So, while the UK’s Met Office now publishes a Fire Severity Index to help authorities assess how severe a potential fire might become, it’s based on Canada’s fire rating system, which was created for different landscapes.

In Rein’s Hazelab at Imperial, his team conduct mostly laboratory-based fire experiments to help inform larger-scale computer simulations. During his research, Rein has felt the heat of huge wildfires first-hand and witnessed how easy it is for flames to leap across mountains as the wind changes. He’s recreated the phenomenon of a fire ‘tornado’ – literally a whirl of flames which springs up in extreme temperatures.

Working with fire modelling programs such as Farsite, his team are analysing data from satellites and sensors to try to refine forecasting.

False-colour image of Winter Hill Fire, Juluy 2018

Image credit: Envirosar

They’re also looking at developing models for an even more complex problem – when to evacuate towns and villages in the fire’s path – using data from satellites, fire service and transport modelling. “This is essential,” he says. “You can’t order an evacuation once the fire has arrived.” But to give the order, authorities need to know how the fire will advance, how people will behave, whether there are schools and hospitals nearby, whether fires might cut across escape routes and how congestion will build – all of which his team are trying to compute and model. Many victims of Portugal’s 2017 fires died tragically after being trapped in cars as they fled the forest fire north-east of Lisbon.

“The need for these models is so acute, it’s one of the fastest transfers of knowledge I’ve experienced,” says Rein. Authorities worldwide are keen to get their hands on his team’s research as soon as it’s conducted.

Wildfire risk maps used by authorities are nearly 20 years out of date, says Attila Toth, chief executive of, a California-based technology company for the insurance sector. In the last couple of years these maps performed poorly – great swathes of properties which burned sat beyond California’s map of fire hazard zones.

“It completely blindsided firefighters and the whole insurance industry,” says Toth. Working with former academics and data scientists, deploys computer vision, deep learning and high-resolution images from satellites and aircraft. The platform can then analyse at scale the fire risk of an individual property. By assessing elements such as the pitch and quality of a roof, and proximity to flammable vegetation or other houses, homeowners and insurers can obtain a realistic assessment of risk in the event of a wildfire. State losses from the fires of 2018 are estimated at a record US$19bn, making it the worst fire season in history. “The tab is rising every year,” says Toth.

Further north, in the remote forests of Canada, fire is a natural and important process for wildlife and regeneration. While it’s generally people who start the wildfires of Europe and populated areas, in nature they might ignite naturally and burn unchecked across hundreds of thousands of remote miles, with ash and smoke drifting across continents. Scientists expect climate change will increase the frequency and scale of these fires and throw more greenhouse gases into the atmosphere.

Last summer, Professor Martin Wooster of King’s College London spent time crossing northern Canada in a small survey aircraft, collecting detailed visible and thermal infrared imagery of natural forest fires, as orbiting remote-sensing satellites passed overhead. This work aims to enhance the ability of remote-sensing satellites to provide accurate, real-time information on global landscape fires.

This high-resolution airborne imagery gathered by Wooster and his team is now being used to validate the coarser, but much more frequent, information derived from satellites. “A lot of our work with satellite data is used to support understanding of what fires are doing around the world,” says Wooster, who’s also a director at NERC’s National Centre for Earth Observation (NCEO). “Where they are burning, how intense they are, how much fuel and carbon they are consuming, and how the gases and particulates they are releasing is affecting the atmosphere.”

Wooster leads the Wildfire Research Group at King’s, which uses a variety of remotely sensed data collected by planes and helicopters, along with ground-based fieldwork and laboratory studies, to develop and validate methods for extracting wildfire information from the data collected by satellites – both those in near-polar orbits some hundreds of miles above the Earth, and in more distant geostationary orbits more than 30,000km away. “We primarily focus on methods using thermal infrared electromagnetic radiation measures,” he says. “The fires produce very strong infrared signatures due to their high temperatures, and this can be used to provide very up-to-date information on the fire situation.”


Radar data

California’s clear skies mean optical satellites more easily capture images of burned areas after wildfires than they do in the UK, where cloud cover often obscures land, says Dr Gail Millin-Chalabi, founder of the University of Manchester start-up EnviroSAR. But radar data could be used alongside optical images – when available – to map the impact of wildfires on UK heathland and moorland.

Using satellite SAR (synthetic aperture radar) data, scientists can monitor the Earth’s surface, day or night, regardless of cloud coverage. This is important, says Millin-Chalabi. Moorland wildfires east of Manchester in 2018 left peat open to the elements and prone to erosion. Organic carbon from peat dissolves into the runoff during downpours of rain and discolours drinking water. If moorlands aren’t restored – protected in heavy rainfall and reseeded to prevent erosion – then sediment and heavy metals in north-west England’s peaty soils could enter into the water course.

“We’re keen to work with authorities and utility companies and identify where it’s best to target restoration efforts,” she says. “Our vision is to do this nationwide and develop a platform to provide this information for any wildfires in the UK. Nationally things need to be more integrated when it comes to what is collected on the ground and the information we can extract from satellite data.”

Satellite-based instrumentation is growing ever more sophisticated and versatile. Historically it collected data used for weather forecasting. Now, meteorological and environmental monitoring satellites are typically equipped with sensors measuring across many more bands of the electromagnetic spectrum, and in more detail. These measurements are coupled to data processing chains that can enact the algorithms on the collected data very quickly,  producing almost real-time information.  

“Together these advances mean that today’s satellite remote-sensing data feeds typically provide very rapid information for applications far beyond meteorology – data useful for detecting landscape fires, monitoring their evolution, and characterising their behaviour,” says Wooster.

This type of rapidly available data can help decision-making, being used to work out what burning fires are putting into the atmosphere. Atmospheric transport models can be used to forecast where smoke might travel and how this could affect air quality en route, and allows authorities to warn at-risk populations.

Better earth observation will help identify areas of risk and can record the progress of large wild fires, says Elliott. Even massive fires start as a single flame or a small ember. “The sooner the responders detect it and take appropriate action the better. Early detection systems have proven effective but are only used in a few locations around the world.”

Training simulations and 3D visualisation can do much to prepare firefighters, particularly in countries such as the UK where they may never have fought a wildfire before. “But we should reflect back. These large uncontrollable wildfires are a relatively recent phenomenon. The indigenous peoples of the world had a good grasp of this. They had the knowledge and the skills to manage the landscape in a way that fire was a positive, not negative, influence.

Further reading on wildfire management is available from and



Due to factors like climate change, wildfires on the scale of the recent blazes in California are growing in number and severity.

In the US alone, seven million acres of forest are lost each year to uncontrolled fires, a faster rate than that at which they can be replaced. The severity of the blazes means they are burning deep into the soil, destroying the seed bank that would normally repopulate the forests, and current methods of replanting are slow and laborious.

Enter DroneSeed, a Seattle-based start-up that intends to automate the solution with a combination of AI, drones and bio-engineered seed packages. The idea arose when CEO Grant Canary was looking for ways to tackle climate change that industry professionals would be interested in. Talking to a friend, he was jokingly advised to plant trees. “I started looking into how trees were getting planted and was baffled that the technology had gone nowhere since the 1800s,” he says.

The main problem was targeting microsites (places a tree would grow well), which required a human mind. Canary quickly realised that a combination of modern technologies could now do what previously only humans had been capable of.

DroneSeed harnesses these technologies into a powerful multi-level solution. Drones are used to scan the affected area, using lidar and multi-spectral cameras. This produces a 3D map of the terrain at a high enough resolution to identify vegetation as well as small-scale features such as stumps, roots, new trees and other features that help DroneSeed’s machine-learning program identify the perfect spots for reseeding.

The next step is herbicide spraying to eliminate unwanted plants. The maps enable spraying to be targeted to specific areas instead of blanket spraying, which is much more cost-effective and environmentally friendly.

The final stage is the seeds. These are bio-engineered in tiny packages that surround the seed in nutrients tailored exactly to match the soil content and acidity of the terrain. The whole thing is held together with biodegradable fibre and coated with capsaicin, the chemical that gives spicy food its kick and which also deters predators. The seed packages are shot from special dispensers attached to the drones at just the right speed and position to land them within a few centimetres of the microsites. The packages sit on the surface “like pinecones”, says Canary. The material surrounding the seed absorbs and holds moisture, helping newly grown roots avoid drying out and boosting survival.

The aircraft tasked with reseeding are no ordinary drones. DroneSeed takes off-the-shelf 2.4m UAVs and retrofits them to its needs. “Our drones are special as we’ve got exacting standards to carry a heavy payload and be reliable for thousands of flights a year and to work with our software,” says Canary. One of these special characteristics is load bearing. DroneSeed is the first US company to obtain a multi-craft, over-55lb (25kg) unmanned aerial vehicle licence from the Federal Aviation Administration.

The drones fly in swarms of five and can switch in and out of formation. This, together with the added load capacity, means they can cover larger areas and operate over longer time spans. It also adds to the challenge. Coordinating five drones is no simple task. Add to this operating in areas with no phone or Wi-Fi signal and you get the feel for some of the technical issues the team had to overcome.

The company has been testing with some of the US’s top timber companies. In October last year, it trialled its first post-wildfire environment – a 5,000-acre burn site for a large timber company that didn’t have enough seedlings to reforest the area itself. “This is happening all over the US,” says Canary. “This moment where it’s ‘Oh hell, we had X thousand acres burn. We don’t have X thousand acres of seedlings ready to go. We don’t have several hundred or thousand labourers. What do we do?’ DroneSeed provides that solution.”

The company is now in talks with stakeholders in Northern California about the re-establishment of rangelands and forests, according to Canary.

By Lee Williams

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles