Abstract:
Air pollution is a serious concern all over the world. It is primarily caused by human activities, such as fuel combustion, motor vehicle exhaust and industrial waste. Besides that, natural sources (volcanic activity, dust and smoke) can emit air pollutants as well. Pollution has adverse effects not only on human health but also on agricultural yield. One of the main air pollutants is particulate matter (PM). There are two major types of PM, PM₂.₅ and PM₁₀, which refer to the particles with a diameter equal to or smaller than 2.5 micrometers and 10 micrometers, respectively. Because monitoring the concentration of PM₂.₅ across the entire Auckland is expensive, in this thesis we investigate how the measurements can be replaced with estimations which are based on levels of pollutants collected at specific locations in the city and meteorological data. To this end, we propose three different approaches for the predictive modelling of PM₂.₅ -level: (i) linear models; (ii) decision trees and (iii) random forests. We discuss in detail the advantages and disadvantages of each method and provide guidance on how they can be used in practical applications. We also propose scenarios for stopping the measurements at some sites where the pollutants are currently monitored and evaluate the estimation errors in each case. Air pollution; particulate matter; linear model; decision trees; random forests