A novel spatiotemporal-enhanced aerosol retrieval model based on GNSS-derived PWV

Aerosol optical depth (AOD) is a core indicator for quantifying atmospheric pollution, and its spatiotemporal variation is strongly influenced by meteorological factors, including water vapor, a critical driver of atmospheric moisture transport and energy balance. To investigate the role of water vapor in AOD modeling, this study integrates global navigation satellite system (GNSS), precipitable water vapor (PWV), and ERA5 meteorological data to develop a high-resolution spatiotemporal AOD model over Europe, at 0.25° spatial and 1-hour temporal resolution. The accuracy of GNSS PWV and MERRA-2 AOD is first validated, yielding R2 values of 0.988 and 0.843, respectively; the correlation coefficient between MERRA-2 AOD and GNSS PWV is 0.47. Wavelet coherence analysis reveals a 256–512-day periodicity shared by AOD, GNSS PWV, and selected meteorological indices. Three models using multiple linear regression, support vector machine, and random forest (RF) are developed. Comparative evaluation demonstrates that the RF model achieves superior performance in internal and external consistency checks, spatial validation, and generalization tests. These findings confirm the reliability of GNSS PWV-based high-resolution AOD retrieval, offering valuable reference and application potential for climate research.