After nearly two years of work at the Wits University School of Physics, a team of about 25 scientists has unveiled a cheap way to warn South Africans just how dangerous the air they are breathing is.
Their South African Consortium of Air Quality Monitoring has developed an accurate, cost-effective air-monitoring system dubbed AI_r that samples air quality in real time, communicates the information digitally and analyses it with AI.
It’s a game changer, says team leader Prof Bruce Mellado, a particle physicist, because of the pressing need to know how good or bad our air is. Pollution, ranging from airborne microparticles that directly enter the blood stream to larger pollutants that get stuck in our lungs, is making people sick, sometimes fatally.
Poor air quality has become the norm over wide stretches of the country, lowering the quality of life and causing health problems that range from lung disease to infertility and abnormal foetus development.
But now AI_r, which can at least let people know what precautions they need to take, is ready to be rolled out. About 120 prototypes of the system’s sensor devices have already been deployed around Johannesburg.
“This is vital. People need to know the quality of the air they breathe so they can make good decisions. Air quality information is just as crucial as the weather,” Mellado said.
For example those who exercise in the open air, such as runners and cyclists, need to know if the air pollution is so bad that they will be doing themselves more harm than good.
The haze that blankets cities and industrial areas is a mix of chemicals, dust, exhaust fumes, firewood smoke and hundreds of other pollutants. These include solids, gases and liquid microparticles that can remain in the air for prolonged periods. Major culprits are Sasol and Eskom.
The AI_r system will involve the installation of hundreds of its devices, small boxes fitted with sensors and tiny lasers, which will sample air quality in real time. The information is fed to a cental hub where it is analysed by AI software that identifies hot spots and makes predictions.
A dense network of sensors, Mellado said, would not only give an accurate picture of where air quality is noxious, but would also show the causes — information the Green Scorpions need to take action.
He said this was particularly relevant in areas such as Johannesburg and Soweto where air quality was notoriously bad, pushing the limits set by the World Health Organisation or exceeding them “almost every single day”.
Currently, air quality is measured and monitored by the South African Air Quality Information System (Saaqis), which falls under the forestry, fisheries and the environment department. It operates about 130 testing stations around the country and passes the data on to the South African Weather Service. It is then published with green, orange or red danger ratings. But researchers the Sunday Times spoke to called this system “inaccurate”, “completely pointless” and “meaningless”.
“Traditional monitoring is very expensive,” said Mellado. “You are looking at a cost of about R1.5m-R2m per station.” In comparison, the new devices cost about R1,000 each.
Kristy Langerman, a professor in the University of Johannesburg’s department of geography, environmental management and energy studies, said Gauteng had better air quality monitoring stations than most provinces, but “the data available is not great”.
This was because the monitoring stations required continuous, expensive maintenance and were useless when hit by theft, vandalism, breakages, load-shedding and other challenges. Gaps in the information — such as monitoring downtime during a winter cold snap — rendered the data incomplete and unreliable.
People feel bad. They don't have flu, but they don't want to go to work or cannot work properly. And that is a massive cost to the economy.
— Prof Bruce Mellado
“There are 132 monitoring stations and sensors feeding into Saaqis for the whole country. But in 2022 only 25 of those produced good data,” she said. Without this information it was not possible to determine priority areas and track long-term changes.
Mellado and Langerman said that with the information that is now available it is impossible to say what air quality changes have taken place in cities over the past 10 years or more.
Other shortcomings of Saaqis are that monitoring is done with fixed sensors that cannot adequately factor in wind conditions, Mellado said.
The AI_r system is supported by Canada’s International Development Research Centre, the quantum technology initiative run by the European Council for Nuclear Research (CERN) and South Africa’s National Research Foundation.
The sensor devices can be placed almost anywhere. They consume very little power and run on batteries, so load-shedding is not a problem. They are ready for full-scale production and can be made locally.
Mellado said they had funding for 100 units and were rolling them out this year to demonstrate how a network performs.
"We are working with different stakeholders in goverment to use their facilities to place the sensors. So far, the response has been positive."
Rolled out as a national network, Mellado said, the system would show air quality everywhere across the country. It would pick up where companies are noncompliant with pollution controls and show the development of hot spots. In time the information would be freely available via an app.
“This will save the country billions. Decision-makers cannot act on hypothesis, they need data-driven information. People generally need to be aware of this silent killer and know to plan their lives and know when to wear masks.”
The prototype sensors are showing that Johannesburg’s air “hovers just below the WHO’s recommended safety limit in summer” and gets worse in winter.
In 2019 about 99% of people in South Africa were breathing air that exceeds the quality limits
— Researcher Ryan Mckenzie
Forestry, fisheries and the environment spokesperson Peter Mbelengwa said the department welcomed the additional air quality monitoring efforts to complement the existing network.
"Increased data coverage through sensor networks and forecasting with artificial intelligence can support our improved understanding and management of air pollution.
"The department is engaging the university on the efficacy of the artificial intelligence predictions and to understand if these predictions may be used to assists with data gaps experienced in monitoring stations due to loadshedding."
Ryan Mckenzie, a researcher at the Wits Institute of Collider Particle Physics and at CERN, said that five years ago “about 99% of people in South Africa were breathing air that exceeds the quality limits — particularly in poor areas where people cook and heat their homes with coal and kerosene”.
The impact of breathing poor air has significant health impacts in the long term, but in the short term can cause exhaustion and general malaise. The damage accumulates over time.
“People feel bad. They don’t have flu, but they don’t want to go to work or cannot work properly. And that is a massive cost to the economy,” Mellado said.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.