AtmoSTEM: A high-resolution spatiotemporal emission model for urban air quality applications

As growing environmental pressures challenge urban resilience, sustainability, and human well-being, the lack of high-resolution geospatial information at the urban or intra-urban scale remains a critical limitation for effective and targeted decision-making. In the context of air quality and associated health impacts, this study addresses this gap by developing the Atmospheric SpatioTemporal Emissions Model (AtmoSTEM), a high-resolution spatiotemporal framework for representing atmospheric emissions, pollutant concentrations, and population exposure at 1 km² resolution. The application focuses on the Ioannina basin in Greece, where residential biomass burning (BB) constitutes a dominant emission source, especially during the cold season, frequently leading to pollution levels exceeding the World Health Organization (WHO) Air Quality Guidelines (AQG) and EU thresholds and underscoring the need for targeted interventions.

For this purpose, a high spatiotemporal emission inventory for Residential Heating is developed. The Copernicus Atmosphere Monitoring Service (CAMS) regional emission inventory, structured according to the GNFR classification and provided at a spatial resolution of 0.05° × 0.1°, serves as the baseline dataset. The downscaling process is based on publicly available, open-access, GNFR-dependent high-resolution spatial proxies, including the Coordination of Information on population density data from Global Human Settlement (GHSL), land-use classifications from the Copernicus Land Monitoring Service (CLC 2018), the OpenStreetMap (OSM) road network, and, where applicable in coastal and maritime domains, marine traffic density from the European Marine Observation and Data Network (EMODNet). Particular emphasis is placed on refining pollutant fields that are more relevant to BB activities, thereby improving the spatial representativeness of BB emissions within urban and peri-urban environments.

To capture the temporal variability of the emissions, CAMS is combined with CAMS temporal Regional Profiles (CAMS-TEMPO), enabling the generation of analytically resolved, hourly emission estimates. Pollutant concentrations are then estimated using a Random Forest machine learning model that integrates AtmoSTEM’s high-resolution emissions, with meteorological, satellite-derived, and spatial data, as well as in-situ air quality measurements provided by the University of Ioannina. The resulting high-resolution concentration fields are evaluated against independent in-situ measurements. Additionally, BB-related PM_2.5 fields are derived and analyzed, enabling improved source-specific characterization of residential heating contributions and providing a physically consistent basis for air-quality and exposure assessments.