Doctoral Research Harnesses Deep Learning and Satellite Data to Boost Air‑Quality Monitoring in Finland

Andrea Porcheddu, a Master of Science graduate in Technical Physics, will defend a doctoral dissertation at the Faculty of Science, Forestry and Technology on Kuopio campus that proposes a new method for tracking air quality using artificial intelligence and space‑borne observations.

The study merges satellite measurements of atmospheric particles with data from ground‑based monitoring stations, training deep‑learning models to predict pollutant concentrations at a finer spatial scale than traditional remote‑sensing techniques allow. By teaching the algorithm to recognise patterns in the satellite imagery, the approach can generate near‑real‑time air‑quality maps even in areas where monitoring infrastructure is sparse.

Improved coverage matters because current air‑quality networks rely heavily on fixed stations that are unevenly distributed, leaving large rural and remote regions under‑represented. Accurate, high‑resolution information is essential for public‑health advisories, urban planning and compliance with European Union air‑quality directives, which aim to reduce exposure to harmful pollutants linked to cardiovascular and respiratory diseases.

The dissertation builds on a broader shift in environmental monitoring, where the limitations of satellite data—such as coarse resolution and cloud contamination—are being addressed through machine‑learning techniques. Researchers worldwide have demonstrated that neural networks can fill data gaps, correct biases and extrapolate measurements, and Porcheddu’s work adds a Finnish perspective to this growing field.

Kuopio’s Faculty of Science, Forestry and Technology hosts a multidisciplinary environment that blends physics, remote sensing and data science, providing the academic infrastructure for projects that bridge theory and application. The findings are expected to stimulate collaborations with national agencies and could inform the integration of the method into the European Copernicus programme, which supplies a steady stream of satellite observations for environmental monitoring.

Looking ahead, the research points to possibilities for extending the model to other pollutants such as nitrogen oxides and ozone, as well as scaling the system to cover larger geographic areas across Europe. If adopted widely, the technique could enhance the timeliness and precision of air‑quality forecasts, supporting policymakers and communities in tackling pollution challenges more effectively.