Spatio-temporal prediction of aerosol optical thickness using machine learning and spatial analysis techniques

Emission sources, meteorology, and topography are the major factors that make it difficult to predict aerosols in space and time. In this study, the moderate resolution imaging spectro-radiometer (MODIS) aerosol optical thickness (AOT) and the surface meteorology observed in Korea have been used to predict spatio-temporal AOT by using the machine learning with spatial analysis techniques. This method enables timeseries based prediction and spatial distribution modeling, and allows modeling values where there are no observation points. The model results show root mean square error (RMSE) 0.33 which is smaller than the standard deviation of the observed value 0.43. Using this technique, the trend of aerosol change in the future was estimated, and it was found that the aerosol in the area of interest decreased by about 7.4%. The methodology will be useful to analyze the regional scale aerosol evaluations, air quality, and climate study.

 

Acknowledgement

This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education(NRF-2019R1I1A3A01062804)”