Data-driven modelling of traffic and air pollution

Abstract

Human activities release many polluting gases and particles into the atmosphere which diffuse into the environment. These pollutants have damaging effects on both the natural environment and also serious consequences on the health of human beings. Recently, this has lead to a great interest in modelling and in improving the air quality of the human habitat including the Maltese Islands. Modelling of the air quality around us is too complex to be described by known natural or physical laws. In such situations mathematical models can be inferred directly from observed spatio-temporal data. Various methods for inference are available depending on the nature of the data and the dynamics being modelled. Statistical inference methods have proved to be a fundamental tool for such data driven modelling due to their ability to account for unknown dynamics and unobserved factors commonly found in complex geographic systems. The fundamental requirement for the development of accurate spatio-temporal statistical models is the availability of data that describes well the relationships involved. In Malta, the Malta Environment and Planning Authority is collecting monthly average data of various air pollutants from a network of diffusion tubes located across the Maltese territory. This study uses air pollution measurements associated with traffic, that is nitrogen dioxide and benzene. Thus the main objectives of this study are to i) analyze the traffic and air pollution situations in the Maltese Islands, particularly in Malta, ii) develop a statistical based model that estimates the road network traffic assignments in Malta, based on origin-destination flows and traffic counts, observed and collected by Transport Malta, iii) develop statistical based models to describe air pollution behaviour on the Maltese islands while making use of computationally efficient methods to estimate the model parameters, iv) predict future pollutant levels within the analysed sites, v) determine spatial dependencies between different geographically located sites, vi) identify factors that affect pollution characteristics and vii) identify possible future remedies. Analysis of the results will show that there is little spatial dependency among neighbouring sites and that local sources, mainly traffic, have a predominant effect on the measurements.