A Comprehensive Review of Ensemble and Hybrid Air Pollution Forecasting Models

Modern research increasingly employs deep learning architectures to capture spatial-temporal pollutant patterns. The study has examined deep learning and hybrid modelling methods for the prediction and management of air pollution. The identified models, such as CNN-LSTM, GCN-LSTM and ensemble frame-works, are able to capture spatialtemporal dependencies more effectively com-pared to higher forecasting accuracies using traditional means. The IoT and satellite technologies closely capture monitoring urban and industrial areas in real-time and high resolution. Multi-output and ensemble learning methods are used to improve the simultaneous prediction of multiple pollutants. It boosts forecast accuracy as well as meteorological, traffic, and remote sensing data. This works aims at a comprehensive study of AI/ ML models used in air pollution detection.