Algorithm for Retrieval of Aerosol Optical Properties over the East Asia from Geostationary Environment Monitoring Spectrometer (GEMS)

The Geostationary Environment Monitoring Spectrometer (GEMS) is an instrument primarily designed for remote sensing of trace-gas and aerosol concentrations over the Asia using hyper-spectral channels from 300 nm to 500 nm. In this study, we present a developed GEMS aerosol retrieval algorithm, based on optimal estimation method to provide aerosol optical depth (AOD), single scattering albedo (SSA) at 443 nm, and aerosol loading height (ALH) as products. Look up table (LUT) approaches were used in the algorithm with the consideration of aerosol optical properties obtained from extensive AERONET dataset located in East Asia for three aerosol types of heavy-absorbing fine aerosol (HAF), dust and non-absorbing aerosol (NA). Non-sphericity of dust aerosol models from Go-bi desert are also considered based on T-matrix method using a Vector Linearized Discrete Ordinate Radiative Transfer code (VLIDORT). Six wavelengths from UV to Vis are carefully selected and used in the retrieval algorithm to relieve the ocean interference signal, such as water-leaving radiance. Taking the importance of a priori, the developed GEMS aerosol algorithm retrieved a priori states of AOD and SSA before the OE approach based on the two-channel approach which has been used to the OMAERUV algorithm, instead of using a climatological state.

We tested the GEMS aerosol retrieval algorithm using OMI level-1B data for GEMS measurement, and evaluate the results using ground-based AERONET level 2.0 products obtained from 28 sites located in East Asia, and using OMAERUV data for 3 years from January 2005 to December 2007. Comparisons of the results show that a correlation coefficient between GEMS and AERONET AODs at 440 nm channel is 0.84, and root-mean-square error (RMSE) is 0.28 with regression line slope 0.78 and offset 0.22. Correlation coefficient between GEMS and OMI AODs at 443 nm channel is 0.83, and root-mean-square error (RMSE) is 0.3 with regression line slope 1.1 and offset 0.11. This algorithm, when applied to GEMS satellite data, will provide aerosol products in spatial resolution of 3.5 km x 8 km on hourly basis during daytime.