Abstract

Continuous and accurate surface pollutant data can provide data support for health effect analysis. Based on the hourly AOD data of the Himawari-8 satellite as the basic data set, this study collected auxiliary parameters including meteorological reanalysis data and geospatial data to estimate the surface PM2.5 hourly concentration. The random forest (RF) and CatBoost models with superior performance were integrated by linear fitting. The experimental results showed that the sample-CV R2 and RMSE of the integrating model were 0.929 and 9.846 μg/m3; time-CV R2 = 0.903, RMSE = 11.521 μg/m3; station-CV R2 = 0.894, RMSE = 12.05 μg/m3, which had the best validation accuracy among all the comparison models and were also better than the estimation results of many previous studies. The spatial and temporal analysis results of PM2.5 showed that the surface PM2.5 concentration was generally high in winter and spring, and low in summer and autumn during the study period. During COVID-2019, PM2.5 concentration on the surface of China showed a significant decreasing trend. The model with the estimation method used in this study can produce reliable surface PM2.5 data products.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call