Remote sensing of trace gases from multiple Chinese satellite instruments and its application combined with AI 

Abstract

Recently, China launched a series of the Environmental Trace Gases Monitoring Instruments (EMI) on satellites with different over-pass times to support global daily multi-temporal atmospheric monitoring. Compared with previous satellites in sun-synchronous orbit, such as OMI and TROPOMI which over passes at 13:30 LT, the series of Chinese EMI satellite instruments with similar performance can be used to investigate the diurnal variations in trace gases. Besides, the series of EMI expanded the areas with diurnal observation in trace gases from the North America, East Asia, and Europe, where are covered by geostationary satellite observations, to the global scales, particularly the South Hemisphere. However, how to eliminate the systematic bias between retrieval results from multiple satellites is the major difficulty. Here we are performed further spectral calibration and inversion algorithm improvement. We use meteorological parameters and a priori atmospheric profiles at different moments in the radiative transfer calculations and inversion. Results from the Tracer Model (version 5, TM5) were also used for background correction. We use secondary radiometric calibrations, i.e. soft calibrations, to further correct the systematic biases between observed spectra from different instruments. Through these algorithm updates, we have successfully retrieved the concentrations of several trace gases, such as Ozone (O3), nitrogen dioxide (NO2), sulfur dioxide (SO2) and formaldehyde (HCHO), at different times of day (10:30 and 13:00 LT). The retrieved trace gas concentrations from the series of EMI were further used to investigate the diurnal variation in air pollutant emissions and ozone formation. Moreover, due to the influence of factors such as cloud cover, satellite observations often have gaps. Here we used artificial intelligence (AI) analysis, such as neural operator methods, to achieve spatial full coverage of remote sensing results. We masked existing satellite observation data, and then used artificial intelligence models to repair and fill in missing areas, with ERA5 meteorological field data, various emission inventories, and geographical information data. The combination with AI analysis improved the usability and application scope of satellite data.