A Data-Driven Review of Air Pollution: Causes, Impacts and Prediction Models

Abstract

Air pollution remains a major global environmental problem driven by rapid urbanisation, industrialisation, fossil fuel combustion, and large-scale deforestation, resulting in elevated levels of key pollutants including nitrogen dioxide, carbon monoxide, sulfur dioxide, and particulate matter. These pollutants have a tremendous effect on human health, climate systems and ecosystems, directly impacting multiple Sustainable Development Goals (SDGs), including health, sustainable cities, and climate action. This study presents a comprehensive review of recent literature on air pollution, with keen interest in data-driven approaches for monitoring and prediction. A narrative review analyzed re-cent studies from major scientific databases, with emphasis on modelling techniques, datasets, and methodological frameworks. The findings reveal a clear progression from classical statistical models to more sophisticated machine learning and deep learning methods, which demonstrate improved capability in handling nonlinear, high-dimensional air quality data. The study also highlights the growing importance of multi-source data integration and preprocessing techniques in enhancing predictive performance. Despite these progressions, there are still several challenges to over-come, including data quality, model interpretability, and scalability for real-time applications. The review concludes by identifying critical research gaps and providing strategic insights for researchers, policymakers, and urban planners to aid in the development of effective, data-driven air quality management systems aligned with sustainability goals.