We analyzed the measurements of aerosol parameters in the surface air layer with sets of Panasonic PM2.5 optical sensors and in throughout the atmospheric column by the photometric method at urban and background observation sites in the Middle Urals for 2016–2019. The features of the intra-annual and daily variations in the aerosol parameters in the surface air layer and in the atmospheric column are compared; also, the relationships between the PM2.5 concentration, AOD, and meteorological parameters in two regions are studied. For the first time for the Middle Urals, we constructed the statistical models for estimating the PM2.5 concentration. Multivariate regression models for estimating the logarithm of PM2.5 concentration are far superior to the single-factor models. The significant predictors are found to be: boundary layer height (blh, m), ln AOD, normalized difference vegetation index (NDVI), relative air humidity (Hu, %), and pressure (P, Pa).