Explainable AI for Atmospheric Temperature Prediction Using Air Quality and Meteorological Parameters at Dublin International Airport, Ireland

Airports are dynamic but pose complex environmental concerns because of their air quality. The air quality in the vicinity of the airports can deteriorate due to heavy vehicular traffic and aircraft handling. The air quality interacts with meteorological conditions; thus, understanding these links would benefit any mitigation activities in the search for sustainable solutions. However, these interactions are complex and nonlinear. Therefore, interpretable machine learning techniques can effectively be used to understand complex environmental scenarios. This study examines atmospheric temperature prediction at Dublin Airport, Ireland, using combined air quality and meteorological observations. Hourly data resulting in 3319 observations were used in training, testing, and validating the developed machine learning prediction models. Among the tested machine learning algorithms, namely K-Nearest Neighbours (KNN), Support Vector Regression (SVR), Random Forest (RF), Multi-Layer Perceptron (MLP), and Decision Tree (DT), KNN was selected as the best model based on its validation performance, achieving R2 = 0.85, RMSE = 1.73 °C and MAE = 1.19 °C in the testing stage. This accuracy mainly reflects the estimation of temperature within the observed period. Shapley Additive Explanations (SHAP) revealed the black-box nature and found that CO, SO2, O3, and wind direction were the most influential variables contributing to atmospheric temperature predictions. The findings can then be used as a foundation for any future studies relating to aviation-related emissions, greenhouse gases, air quality, and atmospheric temperature. This study demonstrates the potential for developing similar prediction models for other busy airports and then incorporating necessary policy decisions to mitigate any adverse effects.