Dynamic Linear Models for Air Pollution

Dynamic Linear Models for Air Pollution

Statistics

Daniel McDonagh

2024

Air pollution is a growing concern in urban environments as it can heavily impact public health, especially in relation to respiratory diseases. This has led to a need for real-time monitoring of pollution levels to quickly identify any significant deviations and inform authorities, allowing for necessary interventions. Here, we propose a dynamic monitoring system that uses Kalman filters and multivariate statistical methods to track pollution trends and seasonality. By analysing the residuals of the model with non- parametric methods, we can detect changes in pollution levels and predict future readings. This gives rise to the anticipation of dangerous pollutant levels, enabling us to forewarn authorities. The approach is demonstrated using data from Athens, focusing on key pollutants such as ozone (O3), nitrogen dioxide (NO2), nitrogen oxide (NO), carbon monoxide (CO) and sulphur dioxide (SO2).

Bayesian Statistics dynamic linear models