Optimizing prevention strategies for PM2.5-related health risks in Nakhon Ratchasima

Abstract

The detrimental effects of PM2.5 on the health of urban populations, especially in developing countries, are being carefully addressed by governments worldwide. To mitigate the adverse health effects of air pollution, this paper presents a case study and suggests a three-phase plan for implementing a prediction and control approach in the Thai metro region of Nakhon Ratchasima (NRM). A data-driven assessment of the region's air pollution situation is conducted in the initial phase, which is then followed by an analysis of the health risks related to PM2.5 concentrations. It is therefore imperative to determine which machine learning model has the best capabilities to estimate PM2.5 concentration. With the help of this advanced predictive capability, decision-makers will have an important window of opportunity to put preventive measures in place to safeguard the population from the adverse effects of air pollution. The study showed that there were respiratory diseases in the NRM area because of high concentrations of PM2.5 particles, primarily from vehicle exhaust, which far exceeded Thailand's threshold of 15 μg m−3. A total of 10,379 people, or 16.91% of the affected population, lived in zones with PM2.5 concentrations >38 μg m−3 in 2019. Another 741 people, or 1.21% of the population, lived in zones with PM2.5 values between 36 and 37 μg m−3. The study finds that Extra-Trees regression combined with AdaBoost has the best-predicting capacity for PM2.5 ambient concentration in the NRM area out of the 13 machine-learning regression models examined. This case study presents an additional tool for policymakers to employ in controlling adverse health effects in urban areas.

Keywords

• Respiratory hospital admissions
• Incidence rate ratio
• Predictive models
• Machine learning
• PM2.5 prediction