The burning issue: fighting forest fires with technology

This is a guest post by Bakhtiar Talhah, chief operating officer, Roundtable on Sustainable Palm Oil

Many living in Southeast Asia – and even those outside the region – will remember the 2015 transboundary haze crisis. The World Bank put the economic cost to Indonesia at more than $16bn, while researchers from Harvard and Columbia universities estimated that the smoke caused upwards of 100,000 premature deaths across Indonesia, Malaysia and Singapore.

These fires – which were lit to prepare land for agriculture and then burned out of control – were exacerbated by lower than average rainfall during the dry season, a result of the El Niño. Scientists predict that extreme El Niño and La Niña events will become more frequent in the 21st century, intensifying existing hazards such as forest fires. While neither El Niño nor La Niña conditions are currently prevailing, many fear that in the wake of Covid-19 this dry season will be just as deadly.

Understanding forest fires

For millennia, fire has been used by hunter-gatherers and in slash-and-burn agriculture with few detrimental effects on biological diversity. The threat we face today is largely a result of land use policies that encourage logging, plantation establishment and large-scale ranching, underpinned by weak governance.

People light forest fires for many reasons. Some are started for practical and beneficial purposes (seed germination and regeneration, for example), some are accidental, and others are deliberately lit to cause damage. All have the potential to harm forest ecosystems and surrounding communities.

In Indonesia and Malaysia – the source of 85% of the world’s palm oil – fire is often used as a cheap and simple land management tool. For poorer people, it is often the only option. This year, as farmers struggle to make up financial shortfalls, there is a real risk that many will resort to land clearance by burning.

These fires are particularly dangerous because many of them occur on drained peatlands. Dry peat ignites easily and can smoulder underground long after flames on the surface have been extinguished. Moreover, damaged peatlands are a major source of greenhouse gas emissions. The fires in Indonesia released between 15-20 million tonnes of CO2 per day over September and October 2015 – more than the emissions of the entire US economy over the same timeframe.

Technology shows promise

Over the last few years, however, a number of technological innovations have emerged to help us prevent and respond to these fires more effectively.

Internet of Things (IoT) and wireless sensors

IoT devices and sensors measure environmental changes – such as atmospheric temperature, relative humidity, and carbon dioxide levels – allowing for early-stage detection and warning. They also provide valuable forest and climate parameters for fire propagation models.

Wilmar adopts remote sensing technologies to monitor practices on-the-ground and identify fire spots. They also provide concession maps to the World Resources Institute for inclusion in their Global Forest Watch (GFW) platform. GFW uses satellite imagery and data analytics to identify fires to the precision of one square kilometre, allowing Wilmar to mobilise follow-up action on the ground.


When it comes to halting the spread of fires and limiting damage, timing is critical. There’s only a small containment window between the fire starting and raging out of control. Drones give firefighters a bird’s eye view of the terrain and real-time information on the transmission path. Quick to launch and more cost-effective than helicopters, they help emergency services develop plans to contain the blaze.

Many palm oil growers are now integrating drones into their fire-fighting arsenal. Golden Agri-Resources (GAR), for example, deploys drones across all its plantations. Together with satellite surveillance, they enable faster detection and confirmation and, in turn, faster response by GAR’s Emergency Response team.

Machine-learning and artificial intelligence (AI)

Machine-learning and AI are increasingly being applied to large datasets to identify environmental threats, such as deforestation, which increase the risk of major fires.

Deforestation by illegal logging is a perennial problem in the industry – not least because these areas are where fires burn at the highest severity and highest rate of spread. To date, producers and consumers have relied upon monitoring tools which use satellite imagery but detection can be delayed when clouds obstruct the view of plantations.

A new, publicly-available radar-based forest monitoring system represents a significant breakthrough. Funded by a coalition of ten major palm oil producers and buyers, Radar Alerts for Detecting Deforestation (RADD) uses advances in radar and machine-learning methods to help companies and other stakeholders detect deforestation happening in near real-time and with greater accuracy. Preliminary results from pilots in Indonesia and Malaysia indicate that the system can identify tropical deforestation several weeks earlier than optical-based systems.

AI also opens up the possibility of mining social media for insights that inform haze disaster management at different stages of an emergency. Artificial Intelligence for Digital Response (AIDR) is an open source software platform that uses supervised machine learning and artificial intelligence to filter and classify thousands of social media messages per minute. Although this has mainly been used by humanitarian agencies during natural disasters, it’s easy to see how it could be used to prioritise haze disaster response activities by public authorities.


Global Forest Watch Fire, ESRI ArcGIS GIS Software Suite, PlanetScope High Resolution satellite imagery, and NASA Fire Information for Resource Management System (Firms). These allow us to quickly identify potential non-compliance. When a hotspot is detected, we contact the member to encourage them to conduct a field verification and take appropriate action.

Since January 2018, we have made information about members’ concessions, land cover, and active hotspots publicly available on our website through our interactive map application GeoRSPO. This further strengthens supply chain traceability and transparency.

Last year, we detected 1,403 hotspots within RSPO member concessions from a total of 463,952 hotspots across Malaysia and Indonesia. Hotspots in RSPO concessions therefore represent approximately 0.30% of all hotspots detected. This is encouraging but also signals the need for stronger national monitoring and investigation frameworks.

Good governance

In 2015, Indonesia’s National Agency for Aviation and Space calculated that 500,000 ha of palm oil concessions were burnt between June to October, representing 20% of the total burnt area.

Technology can go some way to reducing the incidence of these fires but ultimately it must be coupled with a long-term commitment to sustainable land management and good governance of natural resources.

For the palm oil sector, we need a coherent governance architecture that fosters complementarities and resolves disconnects between public regulations and private standards. Governments have the coercive power to mandate sustainable production practices through national policies and regulations; while private standards, such as the RSPO, can influence behaviour through market access and premium pricing incentives.

To truly eliminate forest fires and haze, we need both governance and technology – and public and private actors – working together in harmony.

Data Center
Data Management