ON the morning of Aug 11, 2026, residents in Serian, a town 39km from the Tebedu checkpoint on the Malaysian-Indonesian border, woke up to a smoke-filled sky.
The haze is back, along with the acrid odour created from all the dangerous elements that comprise it.
Over the next few days, as Serian townsfolk and people in the neighbouring districts of Kuching, Lundu, Samarahan, and Sri Aman went about their lives, the air pollutant index (API) steadily climbed higher, driven by the transboundary haze from Indonesia.
At its worst, Serian – known affectionately among locals as Durian Town for its giant fruit landmark – recorded an API reading of 240; almost 600 schools in Sarawak had to be closed due to the haze over the next few days.
API readings of between zero and 50 indicate good air quality, 51 and 100 moderate, 101 and 200 unhealthy, 201 and 300 very unhealthy, while readings above 300 are categorised as hazardous.
API readings in Serian have since crept back below 200 – but the burning that spews the smoke in Kalimantan, Indonesia, has yet to abate.
Unfortunately, this year, the fires in Indonesia, as well as those currently alight in some parts of Malaysia, are happening at a time when the weather is getting warmer and drier, supercharged by an El Nino weather cycle forecast by scientists to be among the strongest in years.
It was reported that, already, over 200,000ha of land across Indonesia burned between January and July even as the country throws more troops, helicopters, and firefighting equipment at the frontlines in Kalimantan. Four companies are also under investigation and over 70 people have been arrested over the worst fires in West Kalimantan.
The bad news is that El Nino is predicted to only peak later this year, so the high temperatures and droughts that are prime conditions for fires will only keep going.
No smoke without fire
According to the World Health Organisation, air pollution – especially smoke – is a major environmental risk factor linked to severe chronic and acute conditions.
While it triggers asthma, eye and throat irritation, and lower respiratory infections almost immediately, long term exposure could also lead to heart disease, stroke, and certain types of cancer. Recent research even found a link between air pollution and certain types of dementia like Alzheimer’s.
Since it first blew up as a public health crisis during 1997’s severe El Nino episode, nothing much has changed in how we deal with transboundary haze.
While civil society groups have long pushed for a specific transboundary haze pollution act to nail down culpability, Indonesia’s neighbours, including Malaysia and Singapore, have traditionally and primarily relied on diplomacy to resolve the issue, particularly the Asean Agreement on Transboundary Haze Pollution.
Hence the flurry of phone calls and meetings every time smoke gets in our eyes.
However, with the vast advancement in artificial intelligence (AI) models lately, can technology’s latest breakout star help spare humanity from the worst of the impacts from haze?
In the age of AI, can we better predict wind and smoke patterns so that nearby communities can prepare or even evacuate before haze clouds out the sky? Could AI use fires, weather, and satellite data to predict which communities will experience unhealthy air several hours – or even a day – before it happens?
Light up my fire alert
In 2023, five researchers from Universiti Putra Malaysia (UPM) aimed to show some of this is possible through a study entitled “Prediction of Peatland Forest Fires in Malaysia Using Machine Learning”.
Machine learning is a type of AI that lets computers learn from data and make choices or predictions without needing humans to write code for every single step.
For the study, the researchers installed an Internet of Things (IoT) monitoring system at Raja Musa Forest Reserve in Kuala Selangor, with sensors collecting environmental measurements once every minute in accordance with several parameters.
Those parameters were then used to predict how conducive weather conditions such as dryness, heat, and the moisture content of organic material were to the outbreak and spread of fires.
Some years later, a similar system was installed at Kuala Baram, Miri, Sarawak.
UPM’s Institute for Mathematical Research director Prof Dr Aduwati Sali says that at Kuala Baram, a Forest Fire Detection System (FFDS) has been implemented to monitor conditions and detect potential peatland fires and give real-time alerts.
The system consists of IoT ground sensors, weather stations, and cameras connected through LoRa (long range wireless communication), 4G, and satellite networks, she explains.
Data is sent to the Internet in real time and fed into the AI engine to be profiled and used to calculate indices from a Fire Danger Rating System (FDRS), such as the fire weather index, drought code, and duff moisture code (which measures moisture levels in the partially decayed organic material called duff that sits on soil).
Basically, the system assesses fire danger before or during fire-prone conditions.
The FDRS is one that has been adopted by Asean; in Malaysia it has been deployed nationwide by the Meteorological Department (MetMalaysia), says Prof Aduwati.
The data is then processed and analysed in real-time for potential peatland fire alerts through a dashboard.
“The local community, emergency responders, and regulatory agencies can access the dashboard using mobile phones or a computer.
“The alert system is also relayed to villages nearby, and the local forestry department through sirens,” says Prof Aduwati
The professor is also with WiPNET (Wireless and Photonics Networks Research Centre) at UPM’s Computer and Communication Systems Engineering Department. She is one of the authors of the 2023 study, along with Lu Li, Nor Kamariah Noordin, Alyani Ismail, and Fazirulhisyam Hashim.
The uniqueness of the Forest Fire Detection System, according to Prof Aduwati, is that it predicts fire weather indices based on on-site weather and ground sensor data, which is more accurate and localised to the area of interest.
“The system sends the data through the Internet from this remote area in real-time, every two minutes.
“The fire alerts indices predicted by our AI engine have an accuracy of 97% for next day prediction,” she says.
‘I see fire’
The system has been deployed and analysed since 2020 in Kuala Selangor and 2025 in Kuala Baram.
When the first IoT system was being deployed, the team had approached the Selangor Forestry Department, recounts Prof Aduwati.
The team had received grants from Japan’s National Institute of Information and Communi-cations Technology through its Asean IVO initiative, an Asean-Japan collaborative research initiative, to develop the system in 2018.
The area of interest is selected based on historical forest fires, and it offers access to a lookout tower, she says.
The system was replicated in Jambi, Sumatra, and Badas, Brunei, with the same project funding.
In 2025, when the system was enhanced with more state-of-the-art features in Kuala Baram, the team and the Sarawak Forestry Department tested several potential sites and found the best area with access to 4G and satellite networks, vegetation cover, and potential for dry season-caused forest fires.
The system is protected behind a fence to avoid human vandalism and damage from wildlife, while a nearby longhouse Iban community and staff from a forest development company help maintain it.
Prof Aduwati explains that with the alerts the team receives in real-time through the system, the local community and government agencies can prepare emergency response initiatives appropriately.
“In Kuala Baram, we work closely with the Sarawak Forestry Department as well as the forest development company and the longhouse Iban community in developing initiatives on what to do should the fire weather index reach high [yellow] or extreme [red] levels,” she says.
MetMalaysia’s guidelines through the FDRS helps explain fire behaviours and the necessary mitigation and response techniques to be deployed accordingly.
Asked if the system will allow nearby communities and authorities enough advance warning to take preventive action before a fire starts, Prof Aduwati says the Forest Fire Detection System can be accessed in real-time anytime and anywhere.
“... And the authorities can issue sirens on-site at the Iban longhouse and at the Forestry Department’s Miri Office.
“This is apart from the system alerts pushed to the mobile phones.
“Since our AI engine can predict with 97% accuracy for the next day’s fire weather index, the alerts can help the local community and the authorities conduct emergency preparedness and preventive actions in a timely manner,” she says.
The team finds the system is very promising in monitoring forest areas which are remote and at high risk for fires due to the dry season or human activities such as slash-and-burn agricultural practices.
Stop the fire
So are we technologically better equipped to predict haze events now?
Prof Aduwati says: “When we invest in deploying the IoT system with sensors in potential forest areas all over the country, then we can better predict hotspots using ground sensors and weather stations’ parameters.”
However, she cautions that we should not become dependent on technology no matter how promising and futuristic-sounding.
It still comes down to people being more responsible at stopping forest fires from happening in the first place, before they become uncontrollable, haze-causing conflagrations.
“More awareness campaigns should be conducted among the rakyat, especially on avoiding slash-and-burn activities during dry seasons, avoiding open burning, or throwing cigarette butts around when participating in nature activities like fishing, hiking, or camping in fire-prone areas.”
So, ultimately, the power to stop fires and haze still lies with us all – though AI can help to an extent.
