Identification of particulate matter (PM10 and PM2.5) sources using bivariate polar plots and k-means clustering in a South American megacity: Metropolitan Area of Lima-Callao, Peru

Abstract

The identification of different air pollution sources is essential to effectively control atmospheric pollution, particularly in megacities of emerging countries with rapid economic development, such as the Metropolitan Area of Lima-Callao (MALC). The objective of this research was to identify the main sources of particulate matter pollution by applying bivariate polar plots and the k-means clustering algorithm. These statistical techniques were applied to hourly in situ data of four variables collected over a 5-year period (2015–2019) by the Automatic Air Quality Monitoring Network of the MALC: wind direction, wind speed, PM₁₀, and PM_2.5 concentrations. Average PM₁₀ concentrations ranged from 34 μg m⁻³ (CDM station) to 126.7 μg m⁻³ (VMT station), while average PM_2.5 concentrations ranged from 16.8 μg m⁻³ (CDM station) to 41.2 μg m⁻³ (ATE station). The diurnal variation of PM presented two peaks, one in the morning (from 0800 to 1000 h) and the other at night (from 1900 to 2300 h), with the highest concentrations of PM₁₀ recorded at the ATE (0800 h: 155.8 μg m⁻³) and VMT (2100 h: 154.6 μg m⁻³) stations, and PM_2.5 at the ATE station (0800 h: 60.3 μg m⁻³ and 2300 h: 37.5 μg m⁻³). The results showed that the contributions of PM₁₀ are directly related to emissions from industrial activities, automotive fleet, construction, demolition, wind erosion, and the suspension and resuspension of particulates from unpaved roads. Meanwhile, high concentrations of PM_2.5 are mainly attributed to vehicle exhaust emissions, industrial emissions, secondary particulate formation, and drag by the action of the winds. The major source of particulate matter contamination is the vehicle fleet, and within this, automobiles, station wagons, combi vans, and 2 and 3-wheel motorcycles are those that have the greatest contribution. These results were supported by non-parametric statistical tests such as Kruskal–Wallis and Mann–Whitney U and validated by the conditional bivariate probability function. The findings of this work may help to implement pollution prevention and control strategies in the future within this South American megacity.