89th American Meteorological Society Annual Meeting

Monday, 12 January 2009
Cloud property variational retrieval for global real-time applications
Hall 5 (Phoenix Convention Center)
Alan E. Lipton, AER, Inc., Lexington, MA; and J. L. Moncet, R. P. d'Entremont, G. B. Gustafson, R. Aschbrenner, J. B. Eylander, and S. Sarasamma
A one-dimensional variational algorithm has been adapted to retrieval of cloud properties from imager data. The work has been oriented toward global, real-time production of cloud products and data assimilation. The variational framework ensures radiometric consistency among the retrieved cloud properties, and thus facilitates conversion between retrieved physical/microphysical properties and optical properties. This framework is also compatible with transition from the current one-dimensional approach to three or four-dimensional assimilation systems, either as a pre-processor or toward inclusion of cloud properties among the assimilation control variables. With the orientation toward real-time production at full spatial resolution of satellite imagers, algorithmic and computational efficiency are essential elements of the approach. The algorithm includes an option for clustering of pixels in images so that a single solution to the non-linear variational problem can be shared among similar pixels to obtain an initial estimate of cloud properties, followed by a local, linear pixel-by-pixel adjustment. The algorithm handles radiative transfer with the Optimum Spectral Sampling (OSS) method for fast and accurate treatment of molecular absorption/emission across each sensor band. OSS uses optimal weighting of monochromatic calculations at optimally selected spectral points. The treatment of multiple scattering is economized by doing scattering calculations at fewer points than are needed for molecular effects, with explicit control over error tolerances. The initial applications of this cloud retrieval approach are for MODIS thermal infrared measurement data. The presentation will include analyses of sensitivity of retrieval errors to assumptions about variables that are not directly retrievable from the information content of the measurements. We will also present examples of cloud retrievals from Aqua MODIS in relation to products from the CloudSat and CALIPSO active sensors, which are approximately coincident with the center of the MODIS swath.

Supplementary URL: