Retrieving global sources of aerosols from MODIS observations by inverting GOCART model
1Laboratoire de Optique Atmosphérique, Université de Lille 1/CNRS, Villeneuve d 'Ascq, France
2Major part of this study was done while worked at: Laboratory for Terrestrial Physics, NASA Goddard Space Flight Center, Greenbelt, MD, USA
3Laboratory for Terrestrial Physics, NASA Goddard Space Flight Center, Greenbelt, MD, USA
4Science Systems and Applications, Inc., Lanham, MD, USA
5Laboratory for Atmospheres, NASA Goddard Space Flight Center, Greenbelt, MD, USA
6Geophysical Fluid Dynamics Laboratory, NOAA, Princeton, NJ, USA
Abstract. Knowledge of the global distribution of tropospheric aerosols is important for studying the effects of aerosols on global climate. Chemical transport models rely on archived meteorological fields, accounting for aerosol sources, transport and removal processes can simulate the global distribution of atmospheric aerosols. However, the accuracy of global aerosol modeling is limited. Uncertainty in location and strength of aerosol emission sources is a major factor in limiting modeling accuracy. This paper describes an effort to develop an algorithm for retrieving global sources of aerosol from satellite observations by inverting the GOCART aerosol transport model.
To optimize inversion algorithm performance, the inversion was formulated as a generalized multi-term least-squares-type fitting. This concept uses the principles of statistical optimization and unites diverse retrieval techniques into a single flexible inversion procedure. It is particularly useful for choosing and refining a priori constraints in the retrieval algorithm. For example, it is demonstrated that a priori limitations on the partial derivatives of retrieved characteristics, which are widely used in atmospheric remote sensing, can also be useful in inverse modeling for constraining time and space variability of the retrieved global aerosol emissions. The similarities and differences with the standard "Kalman filter" inverse modeling approach and the "Phillips-Tikhonov-Twomey" constrained inversion widely used in remote sensing are discussed. In order to retain the originally high space and time resolution of the global model in the inversion of a long record of observations, the algorithm was expressed using adjoint operators in a form convenient for practical development of the inversion from codes implementing forward model simulations.
The inversion algorithm was implemented using the GOCART aerosol transport model. The numerical tests we conducted showed successful retrievals of global aerosol emissions with a 2°×2.5° resolution by inverting the GOCART output. For achieving satisfactory retrieval from satellite sensors such as MODIS, the emissions were assumed constant within the 24 h diurnal cycle and aerosol differences in chemical composition were neglected. Such additional assumptions were needed to constrain the inversion due to limitations of satellite temporal coverage and sensitivity to aerosol parameters. As a result, the algorithm was defined for the retrieval of emission sources of fine and coarse mode aerosols from the MODIS fine and coarse mode aerosol optical thickness data respectively. Numerical tests showed that such assumptions are justifiable, taking into account the accuracy of the model and observations and that it provides valuable retrievals of the location and the strength of the aerosol emissions. The algorithm was applied to MODIS observations during two weeks in August 2000. The global placement of fine mode aerosol sources retrieved from inverting MODIS observations was coherent with available independent knowledge. This was particularly encouraging since the inverse method did not use any a priori information about the sources and it was initialized under a "zero aerosol emission" assumption. The retrieval reproduced the instantaneous global MODIS observations with a standard deviation in fitting of aerosol optical thickness of ~0.04. The optical thickness during high aerosol loading events was reproduced with a standard deviation of ~48%. Applications of the algorithm for the retrieval of coarse mode aerosol emissions were less successful, mainly due to the currently existing lack of MODIS data over high reflectance desert dust sources.
Possibilities for enhancing the global satellite data inversion by using diverse a priori constraints on the retrieval are demonstrated. The potential and limitations of applying our approach for the retrieval of global aerosol sources from aerosol remote sensing are discussed.