Integration of Satellite Remote Sensing and Ground-based Measurement for Modelling the Spatiotemporal Distribution of Fine Particulate Matter at a Regional Scale

Loading...
Thumbnail Image
Date
2009-11-18T15:26:17Z
Authors
Tian, Jie
Keyword
Remote Sensing , Particulate Matter , Air Quality , Spatiotemporal Distribution , Modelling
Abstract
Accurate information on the spatial-temporal distributions of air pollution at a regional scale is crucial for effective air quality control, as well as to impact studies on local climate and public health. The current practice of mapping air quality relies heavily on data from monitoring stations, which are often quite sparse and irregularly spaced. The research presented in this dissertation seeks to advance the methodologies involved in spatiotemporal analysis of air quality that integrates remotely-sensed data and in situ measurement. Aerosol optical depth (AOD) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) is analyzed to estimate fine particulate matter (PM2.5) concentrations as the target air pollutant. The spatial-temporal distribution of columnar aerosol loading is investigated through mapping MODIS AOD in southern Ontario, Canada throughout 2004. Clear distribution patterns and strong seasonality are found for the study area. There is a detectable relationship between an AOD level and underlying land use structure and topography on the ground. MODIS AOD was correlated with the ground-level PM2.5 concentration (GL-[PM2.5]) at various wavelengths. The AOD-PM2.5 correlation is found to be sensitive to spatial-temporal scale changes. Further, a semi-empirical model has been developed for a more accurate prediction of GL-[PM2.5]. The model employs MODIS AOD data, assimilated meteorological fields, and ground-based meteorological measurements and is able to explain 65% of the variability in GL-[PM2.5]. To achieve a more accurate and informative spatiotemporal modelling of GL-[PM2.5], a method is proposed that integrates the model-predictions and in situ measurements in the framework of Bayesian Maximum Entropy (BME) analysis. A case study of southern Ontario demonstrates the procedures of the method and support for its advantages by comparison with conventional geostatistical approaches. The BME estimation, coupled with BME posterior variance, can be used to depict GL-[PM2.5] distribution in a stochastic context. The methodologies covered in this work are expected to be applicable to the modelling or analysis of other types of air pollutant concentrations.
External DOI