Air quality forecasting using artificial intelligence techniques

Abstract

Air pollution is among the most challenging issues modern societies are currently facing with direct impacts on society, economy, politics, as well as on the health of the population and the environment (European Environmental Agency, 2020). This study aims to introduce a state-of-the-art air quality forecasting model using a Deep Long-Short Memory Network integrating data from various sources, from in-situ measurements coming from official ground-based stations to Earth Observation from the Copernicus Atmosphere Monitoring Service (CAMS). The overall aim for the development of this model is the production of air quality forecasts with higher spatial and temporal resolution than those offered from Copernicus, capturing air pollution dynamics at a city level, something that will significantly affect decision-making allowing the adoption and implementation of better measures by policy makers to minimise the effects of the degraded air quality. The air quality forecasting model presented in this study uses a novel method based on artificial neural networks in order to produce forecasts of future air quality conditions with high spatial and temporal resolution, by leveraging data from a wide range of sources. Even though it was developed in the context of addressing the challenging issue of air pollution, the model presented in this study could also be used, with a few modifications, to produce forecasts of thermal comfort and the probability of forest fires. Keywords: air pollution, artificial intelligence, artificial neural networks, forecast