Doctoral Study Uses AI and Satellite Imagery to Enhance Air‑Quality Tracking in Finland

Andrea Porcheddu, who holds an M.Sc. in Technical Physics, is set to defend his doctoral thesis at the Faculty of Science, Forestry and Technology on Kuopio’s campus, where he introduces a novel AI‑driven approach for monitoring air quality with satellite observations.

Combining satellite‑derived aerosol data with readings from surface monitoring stations, the research trains deep‑learning networks to estimate pollutant levels at a spatial resolution finer than that achievable by conventional remote‑sensing methods. By enabling the model to detect patterns within the satellite images, it can produce near‑real‑time air‑quality maps even where ground‑based stations are few.

Enhanced coverage is crucial since existing air‑quality networks depend largely on stationary stations that are unevenly spaced, resulting in sparse data for many rural and remote zones. Precise, high‑resolution data are vital for health alerts, city planning and meeting EU air‑quality regulations that seek to lower exposure to pollutants associated with cardiovascular and respiratory illnesses.

The thesis taps into a wider movement in environmental surveillance, wherein the drawbacks of satellite observations—like low resolution and cloud interference—are mitigated via machine‑learning methods. Global studies have shown that neural networks can bridge data gaps, adjust biases and extrapolate readings, and Porcheddu’s contribution brings a Finnish angle to this expanding discipline.

At Kuopio, the Faculty of Science, Forestry and Technology offers a cross‑disciplinary setting that merges physics, remote sensing and data science, furnishing the scholarly support needed for theory‑to‑practice initiatives. The results are anticipated to foster partnerships with national bodies and may guide the incorporation of the technique into the European Copernicus programme, which delivers continuous satellite data for environmental oversight.

Future work envisions adapting the model to additional pollutants like nitrogen oxides and ozone, and expanding its reach to encompass broader European territories. Broad adoption could improve the speed and accuracy of air‑quality forecasts, aiding policymakers and citizens in addressing pollution issues more efficiently.