Data-driven understanding of air pollution in Pretoria
1
School of Health Systems and Public Health, University of Pretoria, Pretoria, South Africa
Popul. Med. 2026;8(Supplement Supplement 1):A467
ABSTRACT
INTRODUCTION:
Fine particulate matter (PM₂.₅) is a major contributor to global disease burden, yet long-term monitoring and chemical characterisation remain limited in low- and middle-income countries, including South Africa. South Africa is one of a handful of African countries that has an air quality act since 2005. The national air quality standard of PM2.5 was only promulgated in 2012 and there was a considerable delay to deploy PM2.5 monitoring equipment. This study addresses persistent air quality data gaps in South Africa by deploying and calibrating low‑cost sensors (LCSs) to characterise PM2.5 while quantifying spatial inequality in population exposure across Pretoria.
METHODS:
Two PurpleAir sensors were collocated with an Oizom reference monitor during a multi‑month 2025 campaign, with daily averages computed after rigorous cleaning and outlier screening. Machine‑learning calibration models incorporating meteorological covariates (temperature, relative humidity, wind speed) were evaluated.
RESULTS:
Random forest regression performed best (R² = 0.90,MSE = 5.32), followed by multiple linear regression (R² = 0.88,MSE = 5.94) and a feed‑forward neural network (R² = 0.73, MSE = 13.93). Inter‑sensor agreement was strong (r > 0.97), and PM2.5 frequently exceeded the WHO daily guideline (15 µg/m³), with peaks in late February, mid‑March, and mid‑April. Calibrated observations were integrated with satellite‑derived concentration surfaces and ward‑level population data to estimate population‑weighted exposure. The resulting Pretoria map shows five exposure classes spanning 28.6-76.5 µg/m³, revealing pronounced intra‑urban disparities where high‑density and socio‑economically vulnerable wards bear disproportionate burdens. These mapped disparities, derived via ward aggregation, inequality indices, and cluster detection, align with spatial segregation patterns and highlight environmental injustice.
CONCLUSIONS:
By combining validated LCS measurements with machine‑learning calibration and population‑weighted mapping, the study delivers policy‑relevant evidence for targeted air‑quality interventions, prioritised monitoring expansion, and equitable urban planning in Pretoria.
Fine particulate matter (PM₂.₅) is a major contributor to global disease burden, yet long-term monitoring and chemical characterisation remain limited in low- and middle-income countries, including South Africa. South Africa is one of a handful of African countries that has an air quality act since 2005. The national air quality standard of PM2.5 was only promulgated in 2012 and there was a considerable delay to deploy PM2.5 monitoring equipment. This study addresses persistent air quality data gaps in South Africa by deploying and calibrating low‑cost sensors (LCSs) to characterise PM2.5 while quantifying spatial inequality in population exposure across Pretoria.
METHODS:
Two PurpleAir sensors were collocated with an Oizom reference monitor during a multi‑month 2025 campaign, with daily averages computed after rigorous cleaning and outlier screening. Machine‑learning calibration models incorporating meteorological covariates (temperature, relative humidity, wind speed) were evaluated.
RESULTS:
Random forest regression performed best (R² = 0.90,MSE = 5.32), followed by multiple linear regression (R² = 0.88,MSE = 5.94) and a feed‑forward neural network (R² = 0.73, MSE = 13.93). Inter‑sensor agreement was strong (r > 0.97), and PM2.5 frequently exceeded the WHO daily guideline (15 µg/m³), with peaks in late February, mid‑March, and mid‑April. Calibrated observations were integrated with satellite‑derived concentration surfaces and ward‑level population data to estimate population‑weighted exposure. The resulting Pretoria map shows five exposure classes spanning 28.6-76.5 µg/m³, revealing pronounced intra‑urban disparities where high‑density and socio‑economically vulnerable wards bear disproportionate burdens. These mapped disparities, derived via ward aggregation, inequality indices, and cluster detection, align with spatial segregation patterns and highlight environmental injustice.
CONCLUSIONS:
By combining validated LCS measurements with machine‑learning calibration and population‑weighted mapping, the study delivers policy‑relevant evidence for targeted air‑quality interventions, prioritised monitoring expansion, and equitable urban planning in Pretoria.
| eISSN: | 2654-1459 |
We process personal data collected when visiting the website. The function of obtaining information about users and their behavior is carried out by voluntarily entered information in forms and saving cookies in end devices. Data, including cookies, are used to provide services, improve the user experience and to analyze the traffic in accordance with the Privacy policy. Data are also collected and processed by Google Analytics tool (more).
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
