What we built
One public platform, from sensor to phone
- Outdoor sensor: rooftop PurpleAir PA-II, every 2 min, US EPA-corrected.
- Indoor sensor: PurpleAir in a lecture hall (Chemical and Environmental Engineering building), since 20 Oct 2026; not yet in the results below.
- Wearable sensors (Atmotube): 5 devices, field sessions across campus.
- One public app: corrected data, an indicative AQI on Türkiye’s HKİ scale (PM2.5 breakpoints as published by İBB, after US EPA 2012), health advice, alerts when campus AQI exceeds 100, a daily summary at 08:00 and a personal exposure assistant.
Results
What the data show
12.7 µg/m³
Period mean PM2.5, May–Oct 2026 (not an annual mean). Peak at 06:00 (16.5 µg/m³), lowest at 18:00 (9.4 µg/m³). 42 of 154 complete days were above the WHO 24-hour guideline level (15 µg/m³); WHO annual guideline: 5 µg/m³.
Rooftop PurpleAir PA-II, US EPA-corrected · 3,751 valid hours · 154 complete days (≥ 18 h) · 1 May – 12 Oct 2026.
−38% PM2.5 inhaled
One 1-hour brisk walk on a typical weekday: 22.9 µg inhaled at 06:00–09:00 versus 14.1 µg at 16:00–19:00.
Median of 107 weekdays (14.7 and 9.0 µg/m³) × 1.56 m³/h (US EPA 2011); outdoor sensor, May–Oct 2026.
2 h > 8 h inhaled dose
Two hours in the wrong places (36.5 µg) exceeded eight hours in the right places (16.2 µg) in the app’s exposure model (2.3×). With the measured personal/rooftop ratios the gap is smaller (about 1.2–1.5×) but remains.
Hourly PM2.5 (May–Oct 2026) × building factor (wearables, 27–30 Apr) × indoor factor 0.5 × breathing rate.
Forecasting personal exposure with machine learning: training data from sensors, wearable sessions and weather; explainable hour-by-hour forecasts; source characterization; planned forecast-based early warning.
Burning fossil fuels drives both climate change and much of the air pollution we breathe: from campus to city, low-cost, publicly shared tools can link clean air and climate action.
Join us
Replicate it on your campus, site or city
- Public app and documented methods
- Corrected-data pipeline
- Exposure model
- Field protocol
- Partner campuses and cities
- Co-location with a reference monitor
- Support for a multi-city pilot
- Industrial sites and organized industrial zones: indoor–outdoor and shift-time PM2.5 patterns for screening and awareness; not a substitute for accredited occupational-hygiene or regulatory measurements.
- Municipalities: time-of-day advice for schools and walking routes; morning summaries and alerts.
- Sensor and smart-city companies: a documented correction pipeline and a campus test bed.
Our campus set-up, as a pilot template: one rooftop and one indoor low-cost sensor, five wearables, four field days and the public app.
How to cite
GTU AirLab (2026). GTU AirLab: Campus Air Quality Monitoring Platform. Gebze Technical University. https://gtuairlab.github.io/frontend_v2/
Methods, data sources and limitations: in the app, “Hakkında · Yöntem · Gizlilik” (About · Methods · Privacy). Source code: github.com/gtuairlab/gtuairlab.github.io.