 |
| Titel |
Multi-Angle Aerosol Retrievals Without Lookup Tables: A Feasibility Study for MISR |
| VerfasserIn |
David Diner, Rachel Hodos, Anthony Davis, Michael Garay, Suniti Sanghavi, John Martonchik, Paul von Allmen, Alexander Kokhanovsky, Pengwang Zhai |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250057675
|
|
|
|
|
|
| Zusammenfassung |
| Currently, most satellite-based aerosol retrievals make use of lookup tables (LUTs) to store
precomputed quantities and avoid expensive calculations. However, they inherently limit the
extent and granularity of the solution space. The current retrieval algorithm for
the Multi-angle Imaging SpectroRadiometer (MISR), for example, compares the
observations to a discrete set of 74 aerosol mixtures, each composed of aerosol
models having prescribed optical properties and size distributions. In a recent study
comparing the performance of satellite retrieval algorithms against simulated data
over a black surface [Kokhanovsky et al. (2010), Atmos. Meas. Tech. 3, 909-932],
the MISR algorithm performed reasonably well in recovering the “true” aerosol
optical depths, but because the simulated aerosol model was not matched by the
contents of the MISR LUT, the results were biased. This implies that some biases
between instruments can be attributed to the use of different LUTs. These findings
motivated an investigation of whether a retrieval approach that replaces the LUT-based
aerosol models with a parametric approach, solved using least-squares optimization,
could improve the results. Preliminary testing on the synthetic data demonstrates the
viability of this approach. Superior results relative to the LUT-based retrievals were
obtained using a successive orders of scattering (SOS) radiative transfer code and
Levenberg-Marquardt optimization. These results were obtained using only intensity data.
Use of polarization measurements from future missions is expected to further enhance the
information content of the measurements, thereby improving the constraints on the inverse
problem. Next steps include (1) further testing for sensitivity and robustness with
more general aerosol models and surface reflectance, (2) improving computational
efficiency to eventually satisfy the demands of a global, production-level algorithm, and
(3) testing on actual MISR data. We will present the latest results of our study. |
|
|
|
|
|
|