Estimating hourly ground-level aerosols using Geostationary Environment Monitoring Spectrometer aerosol optical depth: a machine learning approach

Abstract

The Geostationary Environment Monitoring Spectrometer (GEMS) is the world's first ultraviolet-visible instrument for air quality monitoring in geostationary orbit. Since its launch in 2020, GEMS has provided hourly daytime air quality information over Asia. However, to date, validation and applications of these data are largely lacking. Here we evaluate the effectiveness of the first 2 years of GEMS aerosol optical depth (AOD) data in estimating ground-level particulate matter (PM) concentrations at an hourly scale. To do so, we train random forest and XGBoost machine learning algorithms using GEMS AOD data and meteorological variables as input features, then employ the trained models to estimate PM10 and PM2.5 concentrations in South Korea. The model-estimated PM concentrations capture the spatial and temporal variations observed in ground-based measurements well, showing strong correlations. However, they exhibit noticeable biases at the extremes, with a tendency to overestimate concentrations at lower PM levels and underestimate them at higher PM levels. Incorporating locally available data, such as carbon monoxide and nitrogen dioxide measurements, into the model training further enhances performance, improving correlations and reducing errors. Moreover, we demonstrate the feasibility of using machine learning models with neighbouring station data to estimate PM concentrations at ungauged locations where ground PM measurements are not available. Our results will serve as a reference to aid the evaluation of future GEMS AOD retrieval algorithm improvements and also provide initial guidance for data users. Copyright: