Improved Hourly and long-term PM2.5 Prediction Modeling Based on MODIS in Bangkok

Abstract

This study used an innovative combination of ground-based, satellite, climate variables (temperature (T), relative humidity (RH), wind speed (WS)), normalized difference vegetation index (NDVI), and Height of the Planetary Boundary Layer (HPBL) data at the site for generating PM2.5 model. Interestingly, a new parameter called “ventilation rate (VRA)” was utilized in the model. Aerosol optical depth (AOD) data were obtained from the MODIS satellite data. Monthly and daily data were combined for filling non-retrieval days. The linear mixed effect model (LMEM) was applied as a tool for prediction. Validation of the PM2.5 model is handled by comparing ground based PM2.5 concentrations at monitoring sites. The funding shows that the optimal subset LMEM model with the other factors can substantially enhance the precision of predicting ground based hourly PM2.5 concentrations. The coefficient (R2) rises from 0.64 to 0.87, and the root mean square error (RMSE) reduces from 10 to 6 μg/m3. The optimal subset LMEM were generated for all seasons. The hourly R2 values were above 0.65, with a high value of R2 in the summer season (R2 = 0.78) and a low value of R2 in rainy (R2 = 0.65). PM2.5 patterns between observed and predicted show a similar representative and are a slight increase from 2018 to 2022. Therefore, our study provides MODIS AODs data with a potentially helpful estimation of PM2.5 concentrations, giving more information for urban scale studies.