The Meteorological Satellite Center (MSC) of the Japan Meteorological Agency began retrieving aerosol optical thickness over land in March 2007 using the AVHRR/NOAA visible channel. Exact nonspherical scattering with realistic particle shape distribution is fully considered. We compared the MSC optical thickness with AERONET version 1.5 optical thickness at nine AERONET stations in northeastern Asia. Although we found statistically significant correlations at all stations, the dispersions were large, especially at southern stations. We thus examined the samples with large errors at each station to investigate the causes of the errors. The following suggestions were obtained:
(1) The MSC optical thickness shows overestimation for severe Yellow Sand events and underestimation for very thick gray haze. This suggests that the absorption property of aerosol particles used in our theoretical calculation is unrealistic. A revision of the complex refractive index of aerosol used in making the look-up table may be required.
(2) The influence of small cumuli or cloud edges is seen in the medium range of optical thickness (< ∼2). When cloud-contaminated samples are removed by reference to MODIS high-resolution images, the correlations between the MSC and AERONET optical thickness are notably improved and are as high as 0.6 at the most of the stations. However, systematic underestimations were found at many stations.
(3) These underestimations are likely to be caused by the aerosol-free assumption used in the land surface reflectance modelling because the samples on adjacent ocean grids showed excellent correlations and no systematic bias. This is confirmed by the fact that the monthly minimums of optical thickness at AERONET stations show notable positive values.