Using linear mixed effect model to estimate ground-level PM2.5: case study for Tehran

Background and Objective: In the recent decade, critical condition of particulate matters (PMs) concentration is considered as one of the most important issues in Tehran megacity. Due to sparse spatial distribution of air quality monitoring stations and economic considerations, researchers proposed remote sensing technique as a fast and economical way to obtain complete spatial and temporal coverage of PM concentrations. Materials and Methods: In this study, aerosol optical depth (AOD) retrieved by MODIS along with meteorological parameters were used to develop statistical linear mixed effect (LME) model and estimating ground-level PM2.5 concentrations. AOD data with a spatial resolution of 3 km from 13 monitoring stations and meteorological data from 5 synoptic stations were extracted over Tehran during 2013. Results: The results showed that the proposed model was able to explain about 57%-72% of daily PM2.5 concentration variations. Temporal analysis of predicted PM2.5 concentrations could follow the curve trend which was obtained from the observed PM2.5 measurements with a reasonable level of accuracy. Best performance of the model was in May 2013 during a model-fitting and cross-validation practice. Also, the spatial distribution of the estimated PM2.5 concentrations was consistent with the measured values in the monitoring stations. Conclusion: Based on the spatial distribution map of the estimated PM2.5, central and northern parts of Tehran were the most polluted areas in the study region. The result showed that the LME model using the satellite-derived AOD and meteorological variables could provide an accurate prediction of ground-level PM2.5 concentrations