The PhD defence and trial lecture will be held in Auditorium 1 in The Biology Building/Kristine Bonnevies hus. In some cases, it will be possible to attend the trial lecture and dissertation digitally, in that case a link to Zoom will be posted.
Trial lecture
Wednesday 17 September, 10:15-11:00, Aud 1, The Biology Building/Kristine Bonnevies hus
Satellite Retrievals of Atmospheric Composition: Bridging Climate Change and Air Quality Research
Conferral summary (in Norwegian)
Mye svevestøv kan forringe luftkvaliteten og gi betydelige helse- og miljøutfordringer, og er derfor strengt regulert og må reduseres. I sin doktorgradsavhandling har Shobitha Shetty undersøkt hvordan satellittdata, atmosfæremodeller, stasjonsmålinger og lavkostsensorer kan kombineres for å forbedre luftkvalitetsovervåkingen over Europa. Med fokus på svevestøv (PM2.5) og NO2 viser doktorgradsstudien at ved å integrere disse datakildene med maskinlæring gir det mer presise, tidsaktuelle og detaljerte romlige og tidsmessige kart over luftforurensning. Studien viser potensialet i datasynergi og maskinlæring for å styrke overvåking av luftkvalitet og utforming av politikk.
Main research findings
Popular scientific article about Shetty’s dissertation:
Synergy of Earth Observation and Machine Learning for Air Quality Monitoring in Europe
Air pollution poses major health and environmental challenges, especially in urban areas where nitrogen dioxide (NO₂) and fine particulate matter (PM2.5) dominate. Reliable, timely, and high-resolution estimates are essential for exposure assessments and policymaking, but traditional monitoring networks suffer from limited spatial coverage. In her doctoral thesis, Shobitha Shetty develops a Satellite and Machine Learning (ML)-based Estimation of Surface air quality at High resolution (S-MESH) framework, which are capable of integrating data from satellite retrievals, atmospheric models, meteorology, low-cost sensors (LCS) and regulatory stations to improve NO2 and PM2.5 observations and mapping across Europe.
For NO2, daily 1 km estimates show that fusing TROPOMI satellite data with nightlights, vegetation indices, and meteorology achieves robust predictions, with TROPOMI emerging as a critical input for resolving both routine and episodic air pollution events. For PM2.5, downscaling CAMS regional forecasts to 1 km significantly improves estimates, particularly during high-pollution episodes, while also correcting known winter underestimations of Eastern Europe in CAMS reanalysis. LCS integration further enhances model performance with a convolution-based strategy providing the most accurate urban-scale estimates, capturing localized variations that satellites and models miss.
Overall, the findings in this doctoral work concludes that earth observation-based data synergies with ML deliver more accurate, timely, and scalable air quality information, strengthening both scientific assessments and the basis for environmental policy.
Photo and other information:
Press photo: Shobitha Shetty, portrait; 1200px. Photo: Private
Other photo material: Photo with description and credit as specified in the article above, size 3000px.
