The aerosol optical depth retrieved from a satellite is basically provides an inexpensive choice for regular observing of particulate matter (PM) concentration, but there are challenges present in estimation accuracy. This study developed a physically interpretable Principal Component Analysis (PCA) -regression to estimate PM2.5-related dust column mass density (DCMD-PM2.5) and surface PM2.5 levels in Baghdad using the MERRA-2 aerosol and ERA5 reanalysis data from 2014 to 2025. The meteorological data collected from ERA5 consisted of: temperature, dew point, wind components, cloud cover, precipitation, radiation, boundary-layer, and aerosol-related variables from MERRA-2; aerosol extinction, black carbon surface mass concentration, and scattering indicators. These variables were checked to assess their impacts on DCMD-PM2.5 and surface PM2.5 values. Varimax rotation identified five PCs Baghdad, explaining 75.63% of the total variance and representing thermal-radiative mixing, cloud dynamics, aerosol optical loading, precipitation-related removal, and moisture conditions. The PC-based regression models were reconstructed using representative driver variables: temperature, high cloud cover, aerosol optical depth (AOD), total precipitation, and dew point. During model development, the PC-based DCMD-PM2.5 model explained 80.6% of the variance, compared with 53.4% for the surface PM2.5 model. Independent validation showed stronger performance for DCMD-PM2.5, with R = 0.799, RMSE = 4.82 × 10⁻⁵ kg m⁻², MAE = 3.86 × 10⁻⁵ kg m⁻², and a very small negative bias. The surface PM2.5 model showed weaker but still informative performance, with R = 0.448, RMSE = 25.29 µg m⁻³, MAE = 20.27 µg m⁻³, and slight positive bias. The results indicate that DCMD-PM2.5 and surface PM2.5 are related but physically distinct aerosol indicators. The proposed framework provides a transparent approach for representing reanalysis-based dust and PM2.5 variability in data-limited urban environments.