Nitrogen dioxide (NO2) is an air pollutant identified as a public health concern. Exposure to NO3 is associated with a number of adverse respiratory health effects, and ultimately with premature mortality. It also contributes as a precursor to formation of tropospheric ozone (O3) and ammonium nitrate fine particulate matter (PM2.5, particles with aerodynamic diameters < 2.5 μm). Rapid economic growth, industrialization, and urbanization in China are leading to substantial adverse air quality issues, including high levels of annual mean NO2 concentrations. It is important to quantify human exposure to NO2 to evaluate its health impacts and to assess the effectiveness of mitigation approaches. Since 2013, the China National Environmental Monitoring Centre (CNEMC) has been implementing a nationwide monitoring network for the routine measurement of ambient air pollutant concentrations. Previous studies into population exposure used the monitor data as a proxy for human exposure. However, NO2 concentrations within cities have shown high spatial variations. The monitoring network only provides concentrations at a limited number of discrete points, which is inadequate to describe the spatial variability of urban air pollution. New methods need to be developed to tackle these challenges. The overall aim of this PhD project is to explore modelling approaches for better estimating intra-urban variability of NO2 for human exposure research in China, given the obstacles in data availability of monitored data, emission inventories, and other highly spatially resolved data in China. Guangzhou is chosen as an exemplar geographic domain. It is the third largest city in China, with a population of 14 million and an area of 7,433 km2, and does not currently meet the Chinese air quality standard (GB 3095-2012) for NO2, which is set as 40 μg m-3 as an annual average. The Guangzhou local government has an air quality compliance plan that aspires to annual average NO2 concentrations of 40 μg m−3 by 2020. Two modelling methods are widely used to simulate pollutant concentrations at relatively high spatial resolution within urban areas: dispersion modelling and land-use regression (LUR) modelling. Dispersion modelling aims to simulate the physical chemical processes that link the emissions of pollutants from sources and their transport and dispersion. Recently, urban dispersion models have been developed in Beijing, Shanghai, Chongqing, Hangzhou, Kunming, Hong Kong, Harbin, Lanzhou, Urumqi, Liaoning province, Jinan, Fushun, and Macao using ADMS and AERMOD. Substandard modelling results can arise due to insufficient monitor data and incomplete or inaccurate emission inventories. LUR relies on existing measurements to derive the statistical relationship between pollutant concentrations at a given location and predictor variables representing the emission and dispersion of air pollutants. An appropriately sized and designed monitoring network is an important component for the development of a robust LUR model. LUR models are now being applied to simulate pollutant concentrations with high spatial resolution in Chinese urban areas. Current challenges and future needs in employing LUR approaches were identified first in this PhD work. Details of twenty-four recent LUR models for NO2 and PM2.5/PM10 (particles with aerodynamic diameters < 10 μm) were reviewed. LUR modelling in China is currently constrained by a scarcity of input data, especially air pollution monitor data. There is an urgent need for accessible archives of qualityassured measurement data and for higher spatial resolution proxy data for urban emissions, particularly in respect of traffic-related variables. The rapidly evolving nature of the Chinese urban landscape makes maintaining up-to-date land-use and urban morphology datasets essential for LUR models. Given the limited number of monitoring sites in Guangzhou and the geographical scale of the domain, an integrated modelling approach combining dispersion modelling with ADMS-Urban and LUR has been developed in this PhD work. ADMS-Urban was applied in Guangzhou using input data including emissions from the Multi-resolution Emission Inventory for China (MEIC), road geometry from OpenStreetMap, and hourly meteorological data from the European Centre for Medium-Range Weather Forecasts (ECMWF). Concentrations of NO2 were simulated by ADMS-Urban at 83 ‘virtual’ monitoring sites spanning the six districts in Guangzhou and weighted according to population (since the overall focus is estimation of population NO2 exposure). The LUR model was validated against both the 83 virtual sites (adj R2: 0.96, RMSE: 5.48 μg m−3; LOOCV R2: 0.96, RMSE: 5.64 μg m−3) and, independently, against available observations (n = 11, R2: 0.63, RMSE: 18.0 μg m−3). The modelled population-weighted long-term average concentration of NO2 across Guangzhou in 2017 was 52.5 μg m−3, which contributes an estimated 7,270 (6,960−7,620) attributable deaths. This hybrid modelling approach is then applied to explore the scale of emissions reductions necessary within the Guangzhou domain to achieve compliance with a number of different interpretations of an NO2 concentration target of 40 μg m−3. (The Guangzhou Ambient Air Quality Compliance Plan does not explicitly state how to practically assess compliance.) The modelling results show that achieving compliance requires different levels of emission reductions, depending on how the concentration target was defined; for example, to reduce the average concentration at all monitoring sites below 40 μg m−3, requires a 60% reduction of emissions from all source sectors. In contrast, to attain ≤40 μg m−3 concentration across the whole of Guangzhou requires a 90% emissions reduction. The impacts of the emissions reductions on NO2-attributable premature mortality are also calculated and illustrate that use of a concentration value as a target does not fully convey the underlying health gains even when the target is not met. In the final part of this thesis, the findings and implications from the modelling studies are discussed in the context of current air quality management system in China. Whilst the results are based on detailed and consistent model results for the specific situation in Guangzhou, they are relevant for, and can provide evidence to, decision makers designing effective air pollution control policies in other fast-growing megacities in China and elsewhere globally. The challenges and limitations for the development of a highly spatial revolved model for human exposure are discussed.