Predictive Modelling of Air Quality in Madrid.

Abstract

As global concerns about climate change and deteriorating air quality intensify, the European Environment Agency (EEA) and other international organizations are making copious efforts to undo the damage that so many human activities and industries have done to our ecosystems, especially to the air we breathe. The Barcelona Institute for Global Health (Instituto de Salud Global de Barcelona) annually publishes a ranking that studies mortality attributable to air pollution in more than 1,000 European cities. The Spanish capital, Madrid, leads the ranking associated with deaths caused by nitrogen dioxide. This end-of-degree dissertation provides a holistic assessment of Madrid City Council’s current air quality system. It is demonstrated that this system is rather rudimentary and needs urgent actualization. Not only is this air quality control system only composed of 24 static measurement stations, but also, the data is vastly incomplete. Furthermore, two predictive models have been developed (an ARIMA Time Series and an LSTM recurrent neural network) to study how time series models adapt to this type of data. These models highlight the importance for Madrid’s City Council to have a robust air quality control system. The results of both predictive models are used to make recommendations to the City Council on improving its air quality system. A stronger air quality system will allow Madrid’s City Council to act proactively in reducing pollution and making efficient energy use.