8.4
An Albedo Dataset Derived from MODIS for Use in Climate Models Over Northern Africa and the Arabian Peninsula

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 1 February 2006: 4:45 PM
An Albedo Dataset Derived from MODIS for Use in Climate Models Over Northern Africa and the Arabian Peninsula
A405 (Georgia World Congress Center)
Liming Zhou, Georgia Institute of Technology, Atlanta, GA; and R. E. Dickinson and Y. Tian

Current climate models generally represent soil albedos by a limited number of prescribed values without consider solar zenith angle (SZA) dependence. This paper develops a new dataset to characterize spatial and spectral variability and SZA dependence of soil albedos over Northern Africa and the Arabian Peninsula from the MODerate resolution Imaging Spectroradiometer (MODIS) albedo products for use in climate models. A climatology of 21 albedo kernels of MODIS 7 spectral bands at 1 km resolution is first composited from the best quality pixels during dust-free seasons from 2000 to 2005. A minimum noise fraction (MNF) rotation transforms is then performed to determine the inherent spatial patterns at various spatial scale, segregate noise, and reduce the dimensionality of the MODIS data. By including the first 1-7 transformed MNF bands, we create a dataset that provides a set of grid mean MNF amplitudes for each model grid cell at spatial resolutions from 0.5° to 5°. These amplitudes can be used to formulate direct albedos at any SZA and diffuse albedos either for MODIS 7 spectral bands or for climate model broadbands. The grid mean albedos of MODIS based are assessed versus those of the MNF based by a linear fit using ordinary least squares to quantify the proportion of the MODIS based total variance explained by the MNF based method at four spatial resolution (square) grids of 50, 100, 150, and 200 km. Results indicate that the new dataset significantly improves characterization of spatial and spectral variability and SZA dependence of soil albedo relative to simple means from MODIS data. The new dataset successfully captures most of the MODIS based albedo variance with much fewer parameters and is able to extract large-scale spatial structures of albedo patterns from the original MODIS data.