Low-cost sensors
Research supports low-cost PM networks only when calibration, humidity correction, co-location, and drift monitoring are part of the operating model.
Research-backed design
Built from evidence, not dashboard decoration.
The architecture follows published findings on low-cost sensor calibration, smartphone-image PM estimation, satellite AOD priors, and spatiotemporal air-quality forecasting.
Citizen photos
Smartphone images can contribute PM/haze signals because particles scatter light and degrade outdoor visibility, but this should be treated as soft evidence.
Satellite priors
MAIAC/MODIS AOD and Sentinel/TROPOMI products help with regional context and priors; they are not street-level truth by themselves.
Forecasting
Spatiotemporal graph models are strong long-term candidates, while gradient boosting is the practical hackathon/pilot baseline for sparse data.
Design decision
VayuLens therefore uses a transparent evidence graph: each source has confidence, decay, corroboration rules, and municipal accountability.