• aerosol;
  • data assimilation

[1] Currently, the Moderate-resolution Imaging Spectroradiometers (MODIS) level II aerosol product (MOD04/MYD04) is the best aerosol optical depth product suitable for near-real-time aerosol data assimilation. However, a careful analysis of biases and error variances in MOD04/MYD04 aerosol optical depth product is necessary before implementing the MODIS aerosol product in aerosol forecasting applications. Using 1 year's worth of Sun photometer and MOD04/MYD04 aerosol optical depth (τ) data over global oceans, we studied the major biases in MODIS aerosol over-ocean product due to wind speed, cloud contamination, and aerosol microphysical properties. For τ less than 0.6, we found similar uncertainties in the mean MOD04/MYD04 τ as suggested by the MODIS aerosol group, while biases are nonlinear for τ larger than 0.6. We showed that uncertainties in MOD04/MYD04 data can be reduced, and the correlation between MODIS and Sun photometer τ can be improved by reducing the systematic biases in MOD04/MYD04 data through empirical corrections and quality assurance procedures. By removing noise and outliers and ensuring that only the highest-quality data were included, we created a modified aerosol optical depth product that removes most massive outliers and ultimately reduced the absolute error (MODIS–Sun photometer) in MODIS τ at 0.55 μm (τ0.55) by 10–20%. Averaged over 1 year's worth of Terra MODIS aerosol product over global oceans, we found a 12% reduction in MODIS τ0.55 with extremes of 30% over the southern midlatitudes and the North Pacific due to a reduction in cloud contamination. This modified aerosol optical depth product will be used operationally.