Traditional studies have relied on fixed-site monitors to quantify concentration of particulate matter (PM). Such monitors, however, cannot fully capture the large spatial variability of pollution. Due to large spatial coverage, satellite remote sensing of PM concentration from aerosol optical depth (AOD) has been actively investigated over the past decade. Conversion from the satellite-based AOD to the ground-level PM concentration, however, remains a great challenge.

Relationship between AOD and PM concentration is affected by aerosol vertical structure, relative humidity, and aerosol characteristics. Observation-based and simulation-based AOD-PM models have been developed to retrieved PM concentration from satellite-observed AOD. However, while there is general recognition of the si...[ Read more ]

Traditional studies have relied on fixed-site monitors to quantify concentration of particulate matter (PM). Such monitors, however, cannot fully capture the large spatial variability of pollution. Due to large spatial coverage, satellite remote sensing of PM concentration from aerosol optical depth (AOD) has been actively investigated over the past decade. Conversion from the satellite-based AOD to the ground-level PM concentration, however, remains a great challenge.

Relationship between AOD and PM concentration is affected by aerosol vertical structure, relative humidity, and aerosol characteristics. Observation-based and simulation-based AOD-PM models have been developed to retrieved PM concentration from satellite-observed AOD. However, while there is general recognition of the significant impact of aerosol characteristics on the AOD-PM_2.5 relationship, and these characteristics have been considered in simulation-based methods, they have not been considered in any observational-based methods. On the other hand, simulation-based methods are limited by model biases and uncertainties, and it is always important to have model independent observational-based methods for comparison.

In this thesis, the development of a new observation-based AOD-PM algorithm that takes into account the effect of the aforementioned aerosol characteristics will be described. Different from simulation-based AOD-PM models, the algorithm is based only on observational data, including MODIS AOD measurements, and meteorological data from surface weather stations. Furthermore, the algorithm is based on physical understanding of the light extinction process in a moist environment, rather than just based on statistical regressions. The new algorithm was first applied to estimate the PM_2.5 distribution in China at 1-km resolution in 2013. Verification against ground observation shows a high spatial correlation of around 0.90 for annual PM_2.5, and a very respectable temporal correlation of >0.70 for hourly PM_2.5 in China.

Air quality studies have been hindered by the lack of long-term PM data in China. With our new observational-based algorithm, we can now reconstruct high-resolution PM₁₀ and PM_2.5 concentration time-series for more than a decade period. Evaluation of the long-term satellite-retrieved PM₁₀ and PM_2.5 concentration against the ground observation also shows accuracy comparable to the best of other studies that made use of satellite data at 10-km resolutions.

The reconstruction of the long-term PM data now allows the assessment of long-term trend, population exposure, and health impacts of PM in China for more than a decade period. It is encouraging to find that PM_2.5 concentration decreased over the eastern China especially after 2007. Air pollution, however, remains in high level with population-weighted mean PM_2.5 concentration exceeds 50 μg/m³ over the eastern China in recent years. Annual mortality attributable to PM_2.5 is estimated at around 1.6 million people. In addition, the improvement in air quality has led to reduction of mortality attributable to PM_2.5 after 2007. With the use of the high-resolution satellite data, we also assess the difference in long-term exposure using data at different spatial resolutions. Results show that the use of simple spatial average or data at low spatial resolutions can lead to systematic underestimation of the population exposure.

On the basis of our new model, effects of temporal variation in aerosol characteristics will also be discussed. Results show that the monthly variation in hygroscopic growth is determined more by sulfate, whereas the monthly variation in extinction efficiency is determined more by organic matter. Empirical relationships are also built to address effects of these aerosol chemical compositions on the relationship between aerosol extinction and PM_2.5 concentration. Evaluations show that PM_2.5 estimate from aerosol extinction coefficient largely improves after addressing the effects of temporal variation in aerosol characteristics.