Mapping Predictive Ambient Concentration Distribution of Particulate Matter and Sulfur Dioxide for Air Quality Monitoring Using Remote Sensing

Abstract

Particulate matter (PM) and sulfur dioxide (SO2) are harmful to human health, especially at exposure levels of 30 μg/m3 of PM and 150 μg/m3 of SO2 for 24 hours. Agriculture and transportation are primary sources of PM while industries are the primary sources of SO2. A network of air quality monitoring stations was established to measure ambient PM and SO2 concentrations continuously but they are not enough to cover remote areas of Saskatchewan. To provide a cost-effective solution to this problem, principles of remote sensing for predicting ambient concentrations of air pollutants and mapping their distributions over the areas of interest were applied. Concentration prediction can be done in an almost real-time fashion depending on how fast the satellite data can be obtained. The Landsat TM/ETM+ data were used in this study because of its fine resolution (30m), availability, and easy to obtain compared to other data. A lot of previous studies used the aerosol optical thickness (AOT) as PM or SO2 predictors, but this study used the atmospheric path radiance as the PM or SO2 predictors instead. Calculating atmospheric path radiance is easier than calculating AOT because it requires only one equation to calculate while AOT requires six equations. Two concepts were used as a basis to develop predictive ambient concentration algorithms of PM or SO2: dark pixel and atmospheric correction concepts. The dark pixel concept was used as a guide for selecting suitable Landsat TM/ETM+ bands only. Although it is possible to develop predictive ambient concentration algorithms of PM or SO2 concentrations, but the developed algorithms are limited to be used to estimate PM or SO2 concentrations over the water surfaces only. They cannot be used over the land surfaces. The atmospheric correction concept is based on the energy conservation law. The atmospheric path radiance, which was used as PM and SO2 predictors, was calculated by using the atmospheric correction equation. The relationships between the atmospheric path radiance and PM or SO2 concentrations were developed and used anywhere unlimited as long as the spectral reflectance of the surface is known. After entering the derived relationships and Landsat TM/ETM+ data into the PCI Geomatica Program, the PM10, PM2.5, and SO2 concentration distribution maps were created. The results agree very well with those actually occur in the study area. For example, the map illustrates areas with high PM concentrations due to forest fires. The PM concentrations are shown on the maps to be higher in summer than those in spring and winter due to increasing agricultural activities. The SO2 concentrations in the southern part are higher than that in the central and northern parts of the study area due to emissions from the coal-fired power plant in Estevan. All results confirm that the developed algorithms work well.