In 2016 The Economist unveiled the results of a year-long comparison of air pollution levels in 15 major cities in Asia, Europe and North America, in which Hong Kong was found to be the second most polluted, just behind Seoul. This result, coupled by growing wealth inequality in Hong Kong, motivated Professor Victor OK Li, Cheng Yu-Tung Professor in Sustainable Development and Chair of Information Engineering, and Dr Jacqueline CK Lam, Associate Professor, to embark on a study that sought to quantify the link between air pollution exposure and social deprivation through artificial intelligence and statistical techniques.
They had just established the interdisciplinary HKU-Cambridge Clean Energy and Environment Research Platform (CEERP) with the University of Cambridge and the air pollution study became HKU-Cambridge CEERP’s first major study. Its results, released in January, 2018, provide the most detailed picture yet of air pollution in Hong Kong and the populations that are most affected.
“The Government has only 16 air quality monitoring stations throughout the city, but air quality can actually change across just a few city blocks,” Professor Li said. “We wanted to provide a more accurate and fine-grained estimate of air quality.”
He and his team combined relevant information– including not only the Government’s air quality readings, but traffic flow data, meteorological information and urban morphology – and input that into an artificial intelligence model that created 110,000 different virtual ‘stations’ across Hong Kong, each covering an area measuring 100m x 100m. The model estimated the air quality for each of these areas in real time and had an accuracy of 82 per cent compared to the government readings, which was significantly better than the 50 to 60 per cent accuracy achieved by other data-driven models.
To measure the human impact of air quality, the researchers collected data on social deprivation, including income, education level, profession, and housing ownership or rental, and found that people living in the most socially-deprived areas of Hong Kong were exposed to higher levels of PM2.5 (Particulate Matter of width 2.5 microns or less), a particularly harmful type of pollutant. “There is a statistically significant correlation between social deprivation and PM2.5 exposure, leading to environmental injustice,” Dr Lam said. The districts of Yuen Long, Kwun Tong, Wong Tai Sin and Sham Shui Po were the hardest hit.
In addition to defining the problem, the researchers also suggested remedies, such as funding tree-planting, pedestrianisation, and other schemes that reduce air pollution in the most affected areas.
Personalised air pollution readings
They are also in the process of taking their research to a new level through a HK$50 million Theme-based Research Scheme project, co-led by Professor Li and Dr Lam, called ‘Big Data for Smart and Personalised Air Pollution Monitoring and Health Management’. The project began in 2018 and is funded for five years and it will improve the quality of data collected and make the readings widely accessible to the public.
Additional data will be gathered and fed into a deep-learning artificial intelligence model to sharpen accuracy and develop forecasting capabilities. New apps will be developed to tell people their air pollution exposure at any place and time in Hong Kong. Users would also have the option of inputting information from their Fitbits or other wearable devices about their personal activity levels and other health-related data, to obtain personalised air quality information. They may opt to receive alerts about avoiding active sports in certain areas with high pollution, or be shown how to avoid heavily-polluted areas as they move from one part of the city to another.
A more intensive health study is also part of the project. Young asthma patients will be given electronic devices to measure individual air pollution exposure, lung capacity and inhaler use to understand how these factors may be correlated. Professor Lau Yu-lung and Dr Lee So-lun of the Department of Paediatrics and Adolescent Medicine are involved in this part of the study, which is among the first of its kind internationally. Furthermore, Professor John Bacon-Shone of the Faculty of Social Sciences is also helping the team develop statistical techniques for data analysis.
Dr Lam said they also intend to develop a happiness index to correlate air quality levels with emotional well-being.
“We will make some of this information freely available to the public,” Professor Li said, adding that their work will benefit the healthcare and information technology industries, too, because they will be able to access the project’s technologies and database through a licensing arrangement and develop new apps and other products from them.
The work also has relevance beyond Hong Kong. The air quality model from the first study has been applied in Shenzhen using that city’s data. It is ready for adoption by other cities that want to obtain a detailed picture of air pollution exposure for their residents.
Professor Li’s team is currently recruiting postdocs and research associates, and anyone interested is welcome to contact him at email@example.com
From left: Dr Wilton WT Fok, Professor Victor OK Li, Dr Jacqueline CK Lam, and PhD student Mr Yang Han introduce their research on air pollution-induced environmental injustice.
Visualisation of PM2.5 pollution estimates and Social Deprivation Index (SDI) at the constituency area level in Hong Kong. Darker colours correspond to higher values of air pollution and SDI.
THE INJUSTICE OF
HKU researchers show how poorer people are exposed to worse pollution than others by using an interdisciplinary approach that combines artificial intelligence, big data analytics and environmental management policy.
The Government has only 16 air quality monitoring stations throughout the city, but air quality can actually change across just a few city blocks.
Professor Victor OK Li
Constituency areas (shaded in red) with both the highest level of PM2.5 pollution concentration and the highest level of Social Deprivation Index.