Urban Air Quality Monitoring, Mapping and Modelling to determine the Main Drivers of Air Pollution
Air pollution is a growing concern for human health, biodiversity and natural environment in large urban areas in the UK and elsewhere. Therefore, it is vital to monitor and model air quality (AQ) in urban areas to understand its spatiotemporal variabilities and its main drivers, which can lead to effective AQ management. Due to the high levels of nitrogen dioxides (NO2) and particular matter (PM10), most of the urban area of Sheffield has been declared as an air quality management area (AQMA).
This PhD project intends to improve air quality monitoring network (AQMN) in Sheffield by structuring a multi-criteria dense AQMN based on economic, social and environmental indicators. The network will be made of several layers of sensors including reference sensors (most accurate sensors recommended by EU and DEFRA), low-cost sensors (e.g., AQMesh and Envirowatch E-MOTEs) and IoT (internet of things) sensors. The static AQMN will be supported by MOBIle Urban Sensing (MOBIUS) vehicle which is mounted with AQ sensors. Furthermore, land-use regression (LUR) and dispersion modelling will be carried out for predicting air pollutant levels, analysing spatiotemporal variability of air pollution levels, quantifying emission sources and identifying the main AQ controlling factors, such as land-use, meteorological parameters and road traffic characteristics. In contrast to previous studies, LUR model developed here will be using nonlinear regression approaches that can handle non-normal AQ data distribution and nonlinear association between response and explanatory variables. Furthermore, nonlinear time series analysis (e.g., NARX and NARMAX) will be employed to analyses temporal variability of air pollution concentrations. Various data fusion techniques will be used to integrate AQ data collected by reference, low-cost and IoT (internet of things) sensors.
Furthermore, monitored and modelled data will be fused for developing high resolution spatiotemporal maps in Sheffield. So far, in this project several tasks have been completed, which include: (i) aims and objectives of this research project are defined; (ii) literature review into AQ monitoring and modelling has been conducted; (iii) methodology, data and tools used in this project have been described, however this section will require revisiting from time to time as the project matures; (iv) criteria for AQMN have been defined and locations for AQMS deployment identified (this is also written in a paper format and submitted to a journal for publication (see appendix 7.1); (v) data collected by low-cost sensors (Envirowatch E-MOTES) during 2016 and 2017 are analysed employing graphical presentation, linear regression and nonlinear regression approaches (submitted for journal publication, see appendix 7.2). Future work in year two to four include: (a) deployment of AQ sensors according to the defined criteria to provide high resolution spatiotemporal data; (b) modelling air pollution levels using dispersion modelling approaches, for instance AEROMOD, ADMS-Urban or AirViro to estimate air pollution levels using emissions, meteorological, boundary layer and geographical characteristics; (c) developing a land-use regression (LUR) model using nonlinear regression approach that can handle non-normal distributed AQ data and nonlinear association between response and explanatory variables. LUR model will utilises land use, population, emissions and meteorology data; (d) Fusion of AQ data from various sources, such as reference sensors, low-cost sensors and modelling prediction; (e) developing high resolution spatiotemporal AQ maps using modelled, measured and integrated AQ data in Sheffield; and (f) assessing the performance of low-cost sensors against reference sensors over a range of temporal and spatial scales and improving their performance through better calibration and experimental designs.