Advances in remote sensing and machine learning techniques for air quality monitoring

Various techniques are used to assess air quality. Data for basic parameters, including particulate matter and major air pollutants, can be easily obtained from local meteorological stations, whereas obtaining more detailed information, such as heavy metal concentrations, requires laboratory analysis of collected samples. The structural limitations of regulatory monitoring systems limit their ability to provide continuous coverage across both the spatial and temporal domains. Satellite imagery provides vital information about atmospheric conditions and surface data. Each year, new missions with advanced sensors further enhance remote sensing capabilities. Sentinel-5 precursor, mounted with a tropospheric monitoring instrument, currently provides ready-to-use air quality data, including measurements of gases and aerosols, whereas Sentinel-4 and Sentinel-5, part of the Copernicus program, will extend these capabilities. Public satellite missions, such as Landsat and Sentinel, and instruments such as the moderate resolution imaging spectroradiometer, are widely used and provide high-resolution data with frequent updates. Integrating in situ measurements with satellite data and machine learning (ML) techniques enhances the accuracy and comprehensiveness of air quality monitoring. Modeling serves an important role in bridging data gaps, producing detailed assessments of particular areas, and supporting the partial automation of environmental control systems. This review assesses satellite-based programs, data processing tools, and project realization methods that enable efficient air quality estimation. The review demonstrates how ML-based remote sensing technology is effective for monitoring air quality, discusses commercial satellite missions, presents firsthand experience, and outlines future directions for advancing air quality monitoring technologies.