A physics-informed methodology to estimate PM2.5 from satellite AOD

Aerosol pollution poses critical risks to human health, yet ground-based PM2.5 monitoring remains insufficient globally, with only 3-4 stations per million inhabitants in Europe. Satellite Aerosol Optical Depth (AOD) offers complementary coverage, but translating columnar AOD to surface PM2.5 requires careful consideration of local aerosol properties and atmospheric conditions. The primary challenge is the measurement mismatch: satellite AOD is retrieved at ambient relative humidity (RH), while standard PM measurements are performed under dry conditions (RH < 50%). Existing methodologies apply empirical hygroscopic corrections whose coefficients depend on local aerosol characteristics, limiting transferability and requiring seasonal calibration. This thesis develops a physics-informed methodology for estimating ground-level PM2.5 from satellite AOD in the Po Valley (Bologna, Italy) during February-September 2023. The approach eliminates empirical hygroscopic corrections by directly comparing ambient RH PM2.5 from a Light Optical Aerosol Counter (LOAC) with satellite AOD. Three key aerosol properties are derived from LOAC measurements: effective radius from particle size distributions, average extinction efficiency from particle type classification combined with Mie theory, and average mass density from type-specific values. Results demonstrate that ambient PM2.5 correlates better with AOD (R = 0.56) than dry PM2.5 (R = 0.16). Incorporating aerosol microphysical properties substantially improves correlation to R = 0.76, nearly doubling explained variance (R² from 0.31 to 0.58). A critical methodological contribution addresses the LOAC detection limit: numerical simulations reveal systematic effective radius overestimation by factors of 1.5-2, which an extrapolation correction reduces to 10-20% error. With corrected effective radius, the linear regression slope reaches the theoretical value of 1, validating the theoretical framework.