An Examination of Satellite Aerosol Product Assimilation in an Aerosol Transport Model Using Ensemble Versus Variational Methods

Tuesday, 16 December 2014
Juli I Rubin1, Jeffrey S. Reid2, Jim A. Hansen2, Jeffrey L Anderson3, James R Campbell2, Nancy Collins3, Timothy J Hoar3 and Jianglong Zhang4, (1)NRC Postdoctoral Fellow, Naval Research Laboratory, Monterey, CA, United States, (2)Naval Research Laboratory, Marine Meteorology Division, Monterey, CA, United States, (3)NCAR, Boulder, CO, United States, (4)U of N Dakota-Atmos Sciences, Grand Forks, ND, United States
Many of the world’s operational forecast centers are taking advantage of aerosol data assimilation as a means of improving aerosol forecasts and climatological reanalyses. However, each center has pursued its own methodology, ranging from 2 to 4D variational and Ensemble Kalman Filter (EnKF) methods. This diversity in methods and aerosol modeling approaches has challenged the development of best practices in utilizing satellite aerosol data, particularly for narrow swath or high information content data streams such as from MISR and CALIOP. In this work, a series of experiments are conducted to examine the performance of the EnKF from the Data Assimilation Research Testbed (DART) against 2D and 3D variational counterparts within the US Navy’s Navy Aerosol Analysis and Prediction System (NAAPS) framework. Examples are drawn from the boreal summertime 2013 period to take advantage of field datasets provided by the Studies of Emissions and Atmospheric Composition, Clouds and Climate Coupling by Regional Surveys (SEAC4RS) and Southern Oxidant Aerosol Study (SOAS) campaigns. The EnKF experiments consider ensemble draws from meteorology fields and source functions, as well as multiplicative inflation and adaptive inflation for overcoming sampling errors. The experiment results indicate that source function draws are necessary for producing ensemble spread near-source regions and for producing realistic correlation fields for spreading information, especially in the vertical. The meteorology ensembles were found to be important for transport events, indicating a need for both source and meteorology ensembles. Relative to the 2DVar, EnKF had comparable performance in the midlatitudes at AERONET sites with the added benefit of spreading information in a more realistic manner. In the tropics, the 2DVar outperformed the EnKF, even with the inclusion of inflation methods, due to limited spread in the meteorology ensembles. This points to the need for a hybrid assimilation system that takes advantage of the power of the variational methods to impact the state fields combined with the ability of the EnKF to realistically spread information. In addition to AOD assimilation, the defined EnKF setup is used to explore the inclusion of lidar datasets for constraint of aerosol vertical profiles and comparison to 3DVar.