AI-Driven Forecasting of PM10 Concentrations in the Republic of Ireland: Integrating Machine Learning for Urban Air Quality Prediction

Abstract

This study introduces artificial intelligence models for the first time in the Republic of Ireland to forecast particulate matter (PM₁₀). By integrating local environmental data with cutting-edge predictive algorithms, this research specifically targets high-impact urban areas in Ireland. The models not only account for complex air quality temporal patterns and meteorological variables but also introduce region-specific calibrations that have not been explored in previous air quality studies. Several state-of-the-art machine learning models: such as XGBoost, CatBoost, ANN, LSTM, Bi LSTM, and an adaptable Ensemble model are trained using available data. Interestingly, climatic variables have a minimum impact on the PM₁₀ concentrations in the tested areas of Ireland. However, they show good correlations with PM_2.5 and NO₂ concentration levels. The performance of the models is tested using widely accepted performance indicators (Coefficient of Determination, Root Mean Square Error, Mean Squared Error, and Mean Absolute Error). Based on the outcome of the models, the PM₁₀ forecast for the next 24 h is presented. The findings of this research can be effectively used to enhance policy decisions to minimize the expected rising pollution levels in the future.