Effects of Assimilating Aerosol Optical Depth (AOD) Observations on Dust Forecasts over North Africa and the East Atlantic

Flow-dependent background error covariance is one of major characteristics of the Ensemble Kalman filter (EnKF) data assimilation (DA) method, which is a great advantage under fast-developing weather systems. On the other hand, the EnKF method ignores model bias and long-term statistical error, which are considered in a variational (Var) method. In this study, we used both methods to investigate the impact of assimilating Aerosol Optical Depth (AOD) observations on dust forecasts over North Africa and the East Atlantic. Three experiments with (EnKF and 3D-Var) and without (NoDA) DA cycling were conducted. A dust storm, which was accompanied by clouds, occurred over North Africa between 19 and 23 July 2006, was used for the study. We verified analyses and forecasts of the three experiments using independent observations, such as backscatter data from Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) and Cloud-Aerosol Transport System (CATS) instruments.

Some basic statistics (e.g., observation minus background vs. observation minus analysis) show that all observations (conventional, radiance, and AOD) are appropriately assimilated in both the EnKF and 3D-Var experiments. Both analyses and forecasts are improved in the DA experiments compared to the NoDA experiment. Although the difference between the EnKF and 3D-Var experiments is not significant, both meteorological and AOD fields in the EnKF experiment are slightly closer to the observations than those in the 3D-Var experiment. This is because taking full advantage of AOD observations via flow-dependent background error covariance in the EnKF experiment leads to more accurate description of meteorological fields as well as dust fields via dust-physics interaction.