Title: Development of Geospatial AI and Deep Learning Models to Estimate High-Resolution Multi-Pollutant Ground Air Quality Using Satellite Remote Sensing
Abstract: Air pollution is associated with different health issues, including heart and lung problems and even premature death. The World Health Organization reports an alarming figure nearly 7 million deaths per year worldwide due to exposure to ambient air pollution. The prevalent methodology for monitoring air quality heavily relies on a network of ground-based sensors, which, despite their accuracy, are limited by their sparse distribution and high operational costs. This scarcity and the resultant lack of coverage impede the ability to capture the nuanced spatial and temporal variations in air quality, thereby failing to delineate localized zones of heightened pollution within urban settings.
An alternative approach is modeling ground air quality from satellite imageries. With the advancement of computational technologies and the outstanding performance of deep learning (DL) in statistical modeling, DL is being applied in extracting ground air quality indicators from satellite data. This later approach of air quality modeling, with or without applications of DL, mostly focused on country or regional level as they mostly used low resolution images. In addition, the DL models are usually employed without considering the geographical principles, which are pivotal for accurately reflecting the spatial dynamics of air pollution. This research aims to develop geospatial deep learning models to estimate high resolution ground air quality, i.e., six criteria air pollutants utilizing satellite imageries.
