The Assimilation of Hyperspectral Satellite Radiances in Global Numerical Weather Prediction

Abstract

Hyperspectral infrared radiance data present opportunities for significant improvements in data assimilation and Numerical Weather Prediction (NWP). The increase in spectral resolution available from the Atmospheric Infrared Sounder (AIRS) sensor, for example, will make it possible to improve the accuracy of temperature and moisture fields. Improved accuracy of the NWP analyses and forecasts should result. In this thesis we incorporate these hyperspectral data, using new assimilation methods, into the National Centers for Environmental Prediction's (NCEP) operational Global Data Assimilation System/Global Forecast System (GDAS/GFS) and investigate their impact on the weather analysis and forecasts.

The spatial and spectral resolution of AIRS data used by NWP centers was initially based on theoretical calculations. Synthetic data were used to determine channel selection and spatial density for real time data assimilation. Several problems were previously not fully addressed. These areas include: cloud contamination, surface related issues, dust, and temperature inversions.

In this study, several improvements were made to the methods used for assimilation. Spatial resolution was increased to examine every field of view, instead of one in nine or eighteen fields of view. Improved selection criteria were developed to find the best profile for assimilation from a larger sample. New cloud and inversion tests were used to help identify the best profiles to be assimilated in the analysis. The spectral resolution was also increased from 152 to 251 channels. The channels added were mainly near the surface, in the water vapor absorption band, and in the shortwave region.

The GFS was run at or near operational resolution and contained all observations available to the operational system. For each experiment the operational version of the GFS was used during that time.

The use of full spatial and enhanced spectral resolution data resulted in the first demonstration of significant impact of the AIRS data in both the Northern and Southern Hemisphere. Experiments were performed to show the contribution to the improvements in global weather forecasts from the increase in spatial and spectral resolution. Both spatial and spectral resolution increases were shown to make significant contributions to forecast skill. New methods were also developed to check for clouds, inversions and for estimating surface emissivity. Overall, an improved methodology for assimilating hyperspectral AIRS data was achieved.