Air pollution modelling and forecasting using hybrid machine learning for Craiova City in Romania

Abstract

Abstract Inadequate air quality has adverse impacts on human well-being and contributes to the progression of climate change, leading to fluctuations in temperature. Therefore, it holds great significance to gain a localized comprehension of the interplay between climate variations and air pollution to alleviate the health repercussions of air pollution. This study aims to investigate the associations between meteorological factors, encompassing temperature (T), humidity (H), and air pressure (P), and concentrations of particulate matter (PM1, PM2.5, PM10). Additionally, it explores the correlation between PM1, PM 2.5 and PM10, as well as between noise levels, CO2 emissions, and other variables. To achieve this objective, five hybrid Machine Learning models were employed for predicting PM concentrations and subsequently calculating the Air Quality Index (AQI). The dataset utilized was provided by an independent network of sensors and spans from September 22, 2021, to February 17, 2022. The results indicated that, in general, R² values exceeded 0.96 and, in most instances, approached 0.99. Humidity emerged as the least influential variable on PM concentrations, while the most accurate predictions were achieved by combining pressure with temperature. Moreover, PM10 concentrations exhibited a notable correlation with PM2.5 concentrations and a moderately strong connection with PM1. Nevertheless, the relationship between PM10 concentration and noise levels and CO2 data was relatively weak. Ultimately, this study has established novel relationships for forecasting PM concentrations and AQI based on the most effective combinations of predictor variables identified.