SPATIO-TEMPORAL ANALYSIS OF AIR QUALITY AND NOISE POLLUTION: ADVANCED STATISTICAL METHODS AND PREDICTIVE MODELING

Abstract

Urban environments face escalating challenges from air pollution, which poses significant risks to public health and urban sustainability. Airborne pollutants such as PM2.5 and NO2 contribute to respiratory and cardiovascular diseases, emphasizing the need for high-resolution monitoring and predictive analysis. This study employs mobile sensor networks, specifically data collected from postal vans in Antwerp, Belgium, to analyze spatio-temporal patterns of air pollution over a five-year period (2018–2023). By integrating advanced statistical techniques and machine learning models, specifically Long Short-Term Memory (LSTM) networks, this study identifies pollution hotspots, uncovers temporal dynamics, and predicts future pollution levels. The findings reveal significant seasonal and spatial variations, with industrial zones exhibiting the highest concentrations. Predictive modeling achieved high accuracy, with LSTM models attaining an R² of 0.92 for PM2.5 predictions. This research highlights the utility of mobile sensors in urban environmental monitoring and provides actionable insights for policymakers to mitigate urban air pollution.