Data description >

This PM_2.5 concentration prediction dataset fused the satellite aerosol optical depth (AOD) retrieval, chemical transport model (CMAQ) simulation, and ground PM_2.5 measurement through a machine learning algorithm. By filling the missing satellite data due to cloud cover and high surface reflectance, our method provides spatiotemporally continuous daily PM_2.5 and PM_2.5 composition maps covering China.

The gridded daily average PM_2.5 concentration data are at 0.1° resolution, covering 73.25°-135.25°E and 18.15°-53.75°N according to the centroid of grid. The current dataset covers China during February 25, 2000 – December 31, 2019. Predictions after 2019 will be rolling update.

Data version >

• The predictions during 2000-2019 are from the fixed model trained by data during 2013-2019.
• The predictions after 2019 are from a rolling update model trained by data during a one-year time window that includes the prediction time period. This model will provide near real-time predictions.

Citation >

Xiao, Q., Chang, H. H., Geng, G. & Liu, Y. An Ensemble Machine-Learning Model To Predict Historical PM_2.5 Concentrations in China from Satellite Data. Environmental Science & Technology 52, 13260-13269, doi:10.1021/acs.est.8b02917 (2018)

Download >

We provide customized satellite-derived PM_2.5 and PM_2.5 composition concentration data with the user-specific temporal period and spatial coverage.